diff --git "a/data/train-00000.jsonl" "b/data/train-00000.jsonl"
new file mode 100644--- /dev/null
+++ "b/data/train-00000.jsonl"
@@ -0,0 +1,5000 @@
+{"id": "train_00000", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_00001", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00002", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00003", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00004", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_00005", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00006", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00007", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_00008", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00009", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tooling", "security_gates"]}}
+{"id": "train_00010", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_00011", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "documentation"]}}
+{"id": "train_00012", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00013", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00014", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_00015", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_00016", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_00017", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00018", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_00019", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00020", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_00021", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_00022", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_00023", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00024", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_00025", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00026", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_00027", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00028", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00029", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_00030", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "tooling"]}}
+{"id": "train_00031", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_00032", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00033", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00034", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00035", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00036", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00037", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00038", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_00039", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00040", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_00041", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_00042", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00043", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00044", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00045", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00046", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_00047", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00048", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00049", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "documentation"]}}
+{"id": "train_00050", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00051", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00052", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "security_gates"]}}
+{"id": "train_00053", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00054", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_00055", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00056", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00057", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00058", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_00059", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00060", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_00061", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_00062", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00063", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00064", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_00065", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00066", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_00067", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_00068", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_00069", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00070", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00071", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_00072", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00073", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_00074", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00075", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00076", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_00077", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_00078", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00079", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_00080", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_00081", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00082", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00083", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_00084", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00085", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_00086", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_00087", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_00088", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_00089", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "evaluation_metrics"]}}
+{"id": "train_00090", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_00091", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00092", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_00093", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "governance"]}}
+{"id": "train_00094", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_00095", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00096", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_00097", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00098", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00099", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00100", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00101", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00102", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00103", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00104", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_00105", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00106", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_00107", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00108", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_00109", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00110", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00111", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00112", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_00113", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_00114", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00115", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00116", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00117", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_00118", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_00119", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "tooling"]}}
+{"id": "train_00120", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_00121", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_00122", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_00123", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00124", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_00125", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_00126", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00127", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00128", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_00129", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00130", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00131", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_00132", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00133", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00134", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00135", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_00136", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00137", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00138", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00139", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_00140", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00141", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "security_gates"]}}
+{"id": "train_00142", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00143", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_00144", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00145", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "governance", "documentation"]}}
+{"id": "train_00146", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00147", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00148", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_00149", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_00150", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00151", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00152", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_00153", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_00154", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00155", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00156", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "security_gates"]}}
+{"id": "train_00157", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00158", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00159", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_00160", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_00161", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00162", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_00163", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00164", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00165", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00166", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_00167", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00168", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00169", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_00170", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_00171", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_00172", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00173", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_00174", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_00175", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_00176", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_00177", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_00178", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00179", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00180", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "tooling"]}}
+{"id": "train_00181", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00182", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tooling", "tests_are_truth"]}}
+{"id": "train_00183", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00184", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_00185", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_00186", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_00187", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00188", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00189", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_00190", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_00191", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00192", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_00193", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_00194", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00195", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00196", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "ci_integration"]}}
+{"id": "train_00197", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00198", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "governance"]}}
+{"id": "train_00199", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_00200", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_00201", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_00202", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00203", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00204", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00205", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_00206", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "governance", "documentation"]}}
+{"id": "train_00207", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00208", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00209", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "tooling"]}}
+{"id": "train_00210", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_00211", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_00212", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_00213", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00214", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_00215", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00216", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_00217", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_00218", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "tooling"]}}
+{"id": "train_00219", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00220", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00221", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00222", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_00223", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_00224", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00225", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "documentation"]}}
+{"id": "train_00226", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_00227", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00228", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_00229", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00230", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_00231", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00232", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_00233", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00234", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00235", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00236", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00237", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00238", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00239", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00240", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00241", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00242", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00243", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_00244", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00245", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00246", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_00247", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00248", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_00249", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_00250", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00251", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_00252", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00253", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_00254", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_00255", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_00256", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00257", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00258", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00259", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_00260", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00261", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00262", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00263", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00264", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_00265", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "reproducibility", "security_gates"]}}
+{"id": "train_00266", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00267", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_00268", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00269", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00270", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "reproducibility"]}}
+{"id": "train_00271", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00272", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00273", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00274", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_00275", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00276", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_00277", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00278", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "tests_are_truth"]}}
+{"id": "train_00279", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00280", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00281", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_00282", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00283", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00284", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00285", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_00286", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_00287", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_00288", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_00289", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_00290", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00291", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00292", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_00293", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00294", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "ci_integration", "documentation"]}}
+{"id": "train_00295", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00296", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "documentation", "evaluation_metrics"]}}
+{"id": "train_00297", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00298", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_00299", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_00300", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00301", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_00302", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00303", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00304", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00305", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00306", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00307", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_00308", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00309", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00310", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_00311", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00312", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_00313", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_00314", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00315", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00316", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "governance", "documentation"]}}
+{"id": "train_00317", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00318", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_00319", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tooling", "documentation"]}}
+{"id": "train_00320", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00321", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00322", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00323", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00324", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_00325", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_00326", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00327", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_00328", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00329", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00330", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00331", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00332", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00333", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "documentation"]}}
+{"id": "train_00334", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00335", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_00336", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tooling"]}}
+{"id": "train_00337", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00338", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_00339", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00340", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_00341", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00342", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00343", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00344", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00345", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "evaluation_metrics"]}}
+{"id": "train_00346", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00347", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_00348", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "tooling"]}}
+{"id": "train_00349", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00350", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00351", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00352", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_00353", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_00354", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00355", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00356", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_00357", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00358", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_00359", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_00360", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00361", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_00362", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_00363", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00364", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "governance"]}}
+{"id": "train_00365", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "documentation"]}}
+{"id": "train_00366", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00367", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_00368", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tooling"]}}
+{"id": "train_00369", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "tooling"]}}
+{"id": "train_00370", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "security_gates"]}}
+{"id": "train_00371", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "reproducibility"]}}
+{"id": "train_00372", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_00373", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00374", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "tooling"]}}
+{"id": "train_00375", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00376", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00377", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_00378", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_00379", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "documentation"]}}
+{"id": "train_00380", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_00381", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00382", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_00383", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00384", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_00385", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00386", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00387", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00388", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00389", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00390", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00391", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_00392", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00393", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00394", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_00395", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00396", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_00397", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tooling", "governance"]}}
+{"id": "train_00398", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "security_gates"]}}
+{"id": "train_00399", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00400", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00401", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "tests_are_truth"]}}
+{"id": "train_00402", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00403", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00404", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00405", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00406", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_00407", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00408", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_00409", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00410", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "security_gates"]}}
+{"id": "train_00411", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_00412", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_00413", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_00414", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00415", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00416", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_00417", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00418", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00419", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00420", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00421", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_00422", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00423", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00424", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_00425", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00426", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_00427", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00428", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_00429", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00430", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_00431", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_00432", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "reproducibility"]}}
+{"id": "train_00433", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "governance"]}}
+{"id": "train_00434", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_00435", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_00436", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00437", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00438", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "evaluation_metrics"]}}
+{"id": "train_00439", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_00440", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_00441", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_00442", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00443", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_00444", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00445", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "security_gates"]}}
+{"id": "train_00446", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00447", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tooling", "security_gates"]}}
+{"id": "train_00448", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_00449", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_00450", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_00451", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_00452", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00453", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_00454", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_00455", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00456", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_00457", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00458", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00459", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_00460", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_00461", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00462", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00463", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_00464", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00465", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00466", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00467", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00468", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00469", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00470", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "governance", "ci_integration"]}}
+{"id": "train_00471", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00472", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_00473", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_00474", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_00475", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00476", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00477", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00478", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00479", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_00480", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_00481", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00482", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_00483", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00484", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_00485", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_00486", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "ci_integration"]}}
+{"id": "train_00487", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "reproducibility"]}}
+{"id": "train_00488", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_00489", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00490", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_00491", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_00492", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_00493", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00494", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "tooling"]}}
+{"id": "train_00495", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_00496", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00497", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00498", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_00499", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00500", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00501", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00502", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00503", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_00504", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_00505", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00506", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_00507", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00508", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00509", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_00510", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "documentation"]}}
+{"id": "train_00511", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_00512", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00513", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00514", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00515", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_00516", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tooling", "ci_integration"]}}
+{"id": "train_00517", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00518", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00519", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "governance", "reproducibility"]}}
+{"id": "train_00520", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_00521", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00522", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_00523", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_00524", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00525", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "security_gates"]}}
+{"id": "train_00526", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00527", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_00528", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_00529", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_00530", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_00531", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00532", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_00533", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00534", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00535", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_00536", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00537", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00538", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_00539", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_00540", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "tests_are_truth"]}}
+{"id": "train_00541", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00542", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00543", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_00544", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_00545", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00546", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "documentation"]}}
+{"id": "train_00547", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "governance"]}}
+{"id": "train_00548", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_00549", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_00550", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_00551", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_00552", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_00553", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00554", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_00555", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_00556", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00557", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_00558", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_00559", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00560", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_00561", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00562", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00563", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00564", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_00565", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_00566", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_00567", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00568", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00569", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_00570", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00571", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00572", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_00573", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00574", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_00575", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "tooling"]}}
+{"id": "train_00576", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_00577", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00578", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_00579", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00580", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_00581", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00582", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_00583", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00584", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00585", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_00586", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00587", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00588", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "security_gates"]}}
+{"id": "train_00589", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_00590", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_00591", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00592", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00593", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00594", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_00595", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00596", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00597", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_00598", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_00599", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00600", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00601", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "governance", "ci_integration"]}}
+{"id": "train_00602", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00603", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00604", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_00605", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00606", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00607", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_00608", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00609", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00610", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00611", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00612", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00613", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_00614", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00615", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00616", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00617", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00618", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_00619", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00620", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_00621", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00622", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00623", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_00624", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_00625", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_00626", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00627", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00628", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00629", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00630", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_00631", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00632", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00633", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00634", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_00635", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00636", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00637", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_00638", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00639", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00640", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00641", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_00642", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00643", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_00644", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00645", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_00646", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_00647", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_00648", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00649", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_00650", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_00651", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00652", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_00653", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00654", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00655", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_00656", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00657", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_00658", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_00659", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00660", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00661", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_00662", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_00663", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_00664", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00665", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00666", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "documentation"]}}
+{"id": "train_00667", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00668", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_00669", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00670", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00671", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00672", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00673", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00674", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00675", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_00676", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_00677", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00678", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_00679", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_00680", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00681", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00682", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00683", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_00684", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_00685", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00686", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00687", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00688", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00689", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_00690", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_00691", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_00692", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00693", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "ci_integration"]}}
+{"id": "train_00694", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00695", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_00696", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "reproducibility"]}}
+{"id": "train_00697", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_00698", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00699", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00700", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_00701", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_00702", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_00703", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00704", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00705", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00706", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_00707", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00708", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00709", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00710", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00711", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00712", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00713", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "reproducibility"]}}
+{"id": "train_00714", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00715", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_00716", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00717", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00718", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00719", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00720", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_00721", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_00722", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00723", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_00724", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00725", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00726", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00727", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_00728", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00729", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_00730", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "ci_integration", "documentation"]}}
+{"id": "train_00731", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_00732", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_00733", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00734", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00735", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_00736", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_00737", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_00738", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00739", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_00740", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00741", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_00742", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_00743", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_00744", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00745", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_00746", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00747", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "tests_are_truth"]}}
+{"id": "train_00748", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00749", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_00750", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00751", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_00752", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00753", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00754", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_00755", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_00756", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_00757", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00758", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00759", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_00760", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "security_gates"]}}
+{"id": "train_00761", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_00762", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_00763", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00764", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00765", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00766", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "governance", "reproducibility"]}}
+{"id": "train_00767", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00768", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "security_gates"]}}
+{"id": "train_00769", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_00770", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_00771", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_00772", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_00773", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00774", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00775", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00776", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_00777", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "documentation"]}}
+{"id": "train_00778", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00779", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_00780", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00781", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00782", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00783", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00784", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00785", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00786", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00787", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00788", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00789", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00790", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00791", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00792", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_00793", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00794", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_00795", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "reproducibility", "tooling"]}}
+{"id": "train_00796", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_00797", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_00798", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00799", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_00800", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_00801", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_00802", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_00803", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00804", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_00805", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00806", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "security_gates"]}}
+{"id": "train_00807", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_00808", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00809", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00810", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00811", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_00812", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_00813", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00814", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00815", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00816", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_00817", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00818", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_00819", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_00820", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_00821", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00822", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00823", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_00824", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_00825", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_00826", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00827", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_00828", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_00829", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_00830", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_00831", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00832", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00833", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00834", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00835", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_00836", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00837", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00838", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "security_gates", "governance"]}}
+{"id": "train_00839", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_00840", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00841", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_00842", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00843", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00844", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_00845", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00846", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_00847", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_00848", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "ci_integration", "documentation"]}}
+{"id": "train_00849", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00850", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00851", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_00852", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_00853", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "security_gates"]}}
+{"id": "train_00854", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_00855", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_00856", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_00857", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_00858", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_00859", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_00860", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_00861", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00862", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_00863", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00864", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_00865", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_00866", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00867", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_00868", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_00869", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00870", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_00871", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_00872", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_00873", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_00874", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_00875", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_00876", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_00877", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_00878", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_00879", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_00880", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_00881", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_00882", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_00883", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_00884", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_00885", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_00886", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_00887", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_00888", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_00889", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00890", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_00891", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_00892", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "documentation", "security_gates"]}}
+{"id": "train_00893", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00894", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00895", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00896", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "security_gates"]}}
+{"id": "train_00897", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_00898", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00899", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00900", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "security_gates"]}}
+{"id": "train_00901", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_00902", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_00903", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "ci_integration", "tooling"]}}
+{"id": "train_00904", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00905", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_00906", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_00907", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_00908", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_00909", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00910", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_00911", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00912", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00913", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_00914", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00915", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00916", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00917", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_00918", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_00919", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_00920", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00921", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00922", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_00923", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_00924", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_00925", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_00926", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00927", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00928", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_00929", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00930", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_00931", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_00932", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_00933", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_00934", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00935", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00936", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_00937", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_00938", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_00939", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "documentation"]}}
+{"id": "train_00940", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00941", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_00942", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_00943", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_00944", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_00945", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_00946", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_00947", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_00948", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00949", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00950", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_00951", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_00952", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_00953", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_00954", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_00955", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_00956", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "governance"]}}
+{"id": "train_00957", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00958", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_00959", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00960", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tooling"]}}
+{"id": "train_00961", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_00962", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00963", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_00964", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_00965", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_00966", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_00967", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_00968", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_00969", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_00970", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00971", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00972", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "reproducibility"]}}
+{"id": "train_00973", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00974", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_00975", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_00976", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_00977", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_00978", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_00979", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_00980", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_00981", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00982", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_00983", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "reproducibility"]}}
+{"id": "train_00984", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_00985", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_00986", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00987", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_00988", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_00989", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_00990", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_00991", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_00992", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_00993", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_00994", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_00995", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_00996", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_00997", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_00998", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_00999", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01000", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_01001", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "tests_are_truth"]}}
+{"id": "train_01002", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_01003", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_01004", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01005", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01006", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_01007", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01008", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_01009", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01010", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01011", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01012", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_01013", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_01014", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01015", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01016", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01017", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_01018", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01019", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_01020", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_01021", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01022", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01023", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01024", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01025", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01026", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "security_gates", "documentation"]}}
+{"id": "train_01027", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_01028", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "reproducibility"]}}
+{"id": "train_01029", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01030", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_01031", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01032", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_01033", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01034", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01035", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01036", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01037", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01038", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_01039", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_01040", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01041", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_01042", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_01043", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01044", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01045", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_01046", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01047", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01048", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "reproducibility"]}}
+{"id": "train_01049", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01050", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01051", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_01052", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01053", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_01054", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01055", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_01056", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01057", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01058", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01059", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01060", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01061", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01062", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01063", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01064", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01065", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_01066", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01067", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01068", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_01069", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01070", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_01071", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01072", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_01073", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_01074", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01075", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01076", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_01077", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01078", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_01079", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_01080", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01081", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01082", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_01083", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01084", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01085", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01086", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01087", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01088", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01089", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01090", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_01091", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_01092", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_01093", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01094", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01095", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01096", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_01097", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01098", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "tests_are_truth"]}}
+{"id": "train_01099", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_01100", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01101", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_01102", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01103", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01104", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "governance", "security_gates"]}}
+{"id": "train_01105", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_01106", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_01107", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_01108", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01109", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_01110", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01111", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_01112", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01113", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01114", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01115", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01116", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01117", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "documentation"]}}
+{"id": "train_01118", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01119", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01120", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_01121", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_01122", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01123", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "reproducibility"]}}
+{"id": "train_01124", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01125", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01126", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01127", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_01128", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_01129", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01130", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01131", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "ci_integration", "governance"]}}
+{"id": "train_01132", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_01133", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01134", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01135", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_01136", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01137", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_01138", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_01139", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "ci_integration", "documentation"]}}
+{"id": "train_01140", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01141", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01142", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_01143", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_01144", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01145", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01146", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01147", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_01148", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01149", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01150", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01151", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "evaluation_metrics"]}}
+{"id": "train_01152", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01153", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01154", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01155", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01156", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_01157", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_01158", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01159", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_01160", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_01161", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_01162", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_01163", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01164", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_01165", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01166", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tooling", "documentation"]}}
+{"id": "train_01167", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_01168", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_01169", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_01170", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_01171", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_01172", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01173", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01174", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_01175", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_01176", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01177", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_01178", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_01179", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01180", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01181", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "security_gates"]}}
+{"id": "train_01182", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_01183", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01184", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_01185", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "evaluation_metrics"]}}
+{"id": "train_01186", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01187", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_01188", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01189", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01190", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01191", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01192", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_01193", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01194", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01195", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_01196", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_01197", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_01198", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_01199", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "tests_are_truth"]}}
+{"id": "train_01200", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01201", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01202", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "governance"]}}
+{"id": "train_01203", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01204", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_01205", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_01206", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_01207", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_01208", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_01209", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_01210", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "governance"]}}
+{"id": "train_01211", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_01212", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01213", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_01214", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_01215", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_01216", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "ci_integration"]}}
+{"id": "train_01217", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01218", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01219", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01220", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_01221", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_01222", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01223", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_01224", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01225", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_01226", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_01227", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01228", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01229", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_01230", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_01231", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01232", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01233", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01234", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01235", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_01236", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_01237", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tooling"]}}
+{"id": "train_01238", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_01239", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "security_gates"]}}
+{"id": "train_01240", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_01241", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "ci_integration", "tooling"]}}
+{"id": "train_01242", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01243", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_01244", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_01245", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_01246", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_01247", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01248", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_01249", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01250", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_01251", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_01252", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01253", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01254", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01255", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01256", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01257", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_01258", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01259", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_01260", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "documentation"]}}
+{"id": "train_01261", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_01262", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01263", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01264", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "tooling", "governance"]}}
+{"id": "train_01265", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_01266", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_01267", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01268", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01269", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01270", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_01271", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_01272", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01273", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_01274", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01275", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01276", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_01277", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01278", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01279", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01280", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_01281", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_01282", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_01283", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01284", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "documentation"]}}
+{"id": "train_01285", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01286", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "tooling"]}}
+{"id": "train_01287", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_01288", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_01289", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_01290", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01291", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01292", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_01293", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_01294", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_01295", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01296", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01297", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_01298", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01299", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_01300", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_01301", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01302", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "documentation", "reproducibility"]}}
+{"id": "train_01303", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "ci_integration", "security_gates"]}}
+{"id": "train_01304", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01305", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01306", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tooling"]}}
+{"id": "train_01307", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_01308", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_01309", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_01310", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01311", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01312", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_01313", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "ci_integration"]}}
+{"id": "train_01314", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_01315", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_01316", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01317", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01318", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_01319", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01320", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_01321", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01322", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01323", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01324", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_01325", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01326", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_01327", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_01328", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_01329", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "security_gates"]}}
+{"id": "train_01330", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01331", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01332", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "tooling", "tests_are_truth"]}}
+{"id": "train_01333", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01334", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_01335", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01336", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "reproducibility", "documentation"]}}
+{"id": "train_01337", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01338", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_01339", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01340", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_01341", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_01342", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01343", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_01344", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01345", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_01346", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_01347", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01348", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01349", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01350", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01351", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01352", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01353", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "security_gates"]}}
+{"id": "train_01354", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_01355", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01356", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_01357", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01358", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01359", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01360", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01361", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_01362", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01363", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_01364", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01365", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01366", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_01367", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01368", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_01369", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01370", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01371", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01372", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01373", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_01374", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01375", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_01376", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_01377", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_01378", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_01379", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01380", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_01381", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01382", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_01383", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_01384", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_01385", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01386", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_01387", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01388", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_01389", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01390", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01391", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "reproducibility"]}}
+{"id": "train_01392", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01393", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_01394", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_01395", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01396", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01397", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_01398", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_01399", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01400", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_01401", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "documentation", "reproducibility"]}}
+{"id": "train_01402", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "governance"]}}
+{"id": "train_01403", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01404", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01405", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_01406", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01407", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "evaluation_metrics"]}}
+{"id": "train_01408", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01409", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_01410", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01411", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_01412", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_01413", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_01414", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_01415", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01416", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01417", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01418", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01419", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "governance", "ci_integration"]}}
+{"id": "train_01420", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01421", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_01422", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_01423", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01424", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01425", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01426", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_01427", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01428", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_01429", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01430", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01431", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01432", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_01433", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_01434", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_01435", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01436", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_01437", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01438", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01439", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_01440", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01441", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_01442", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_01443", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "reproducibility"]}}
+{"id": "train_01444", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_01445", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "reproducibility", "security_gates"]}}
+{"id": "train_01446", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01447", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_01448", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01449", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_01450", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01451", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01452", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_01453", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01454", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01455", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01456", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01457", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01458", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_01459", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01460", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_01461", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_01462", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_01463", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01464", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01465", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01466", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_01467", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_01468", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_01469", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_01470", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_01471", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01472", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01473", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01474", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_01475", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01476", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_01477", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01478", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01479", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_01480", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_01481", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "evaluation_metrics"]}}
+{"id": "train_01482", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_01483", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01484", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01485", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_01486", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01487", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_01488", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01489", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01490", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01491", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_01492", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01493", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_01494", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_01495", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01496", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01497", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_01498", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01499", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_01500", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01501", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "reproducibility"]}}
+{"id": "train_01502", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_01503", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_01504", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_01505", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01506", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_01507", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_01508", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01509", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01510", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_01511", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01512", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_01513", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_01514", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01515", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_01516", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01517", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01518", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_01519", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_01520", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "documentation"]}}
+{"id": "train_01521", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01522", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_01523", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01524", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_01525", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01526", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_01527", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01528", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01529", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01530", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_01531", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_01532", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01533", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01534", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_01535", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "security_gates"]}}
+{"id": "train_01536", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01537", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "documentation"]}}
+{"id": "train_01538", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01539", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01540", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01541", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_01542", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01543", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_01544", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01545", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01546", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_01547", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_01548", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01549", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01550", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_01551", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_01552", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "governance"]}}
+{"id": "train_01553", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_01554", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_01555", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01556", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_01557", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_01558", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_01559", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01560", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01561", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_01562", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01563", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01564", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01565", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01566", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01567", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_01568", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01569", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_01570", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_01571", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_01572", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01573", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "ci_integration", "reproducibility"]}}
+{"id": "train_01574", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01575", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_01576", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01577", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01578", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01579", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_01580", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01581", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_01582", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01583", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01584", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_01585", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_01586", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "tooling"]}}
+{"id": "train_01587", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "reproducibility"]}}
+{"id": "train_01588", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_01589", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_01590", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_01591", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_01592", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_01593", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01594", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01595", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01596", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01597", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01598", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_01599", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_01600", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01601", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_01602", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "governance"]}}
+{"id": "train_01603", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01604", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01605", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_01606", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01607", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_01608", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01609", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_01610", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "tooling"]}}
+{"id": "train_01611", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01612", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_01613", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "documentation"]}}
+{"id": "train_01614", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01615", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_01616", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_01617", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01618", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01619", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "documentation"]}}
+{"id": "train_01620", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01621", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01622", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "tooling"]}}
+{"id": "train_01623", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01624", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_01625", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01626", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "ci_integration", "security_gates"]}}
+{"id": "train_01627", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_01628", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01629", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_01630", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_01631", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01632", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01633", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01634", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01635", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01636", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01637", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "governance"]}}
+{"id": "train_01638", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01639", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01640", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01641", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01642", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01643", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_01644", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_01645", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01646", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01647", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_01648", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_01649", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01650", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01651", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "governance"]}}
+{"id": "train_01652", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_01653", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_01654", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01655", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01656", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01657", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01658", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_01659", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_01660", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_01661", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_01662", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_01663", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01664", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_01665", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_01666", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01667", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01668", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_01669", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "governance", "reproducibility"]}}
+{"id": "train_01670", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01671", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01672", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01673", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_01674", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01675", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01676", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01677", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "documentation", "governance"]}}
+{"id": "train_01678", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_01679", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01680", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_01681", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_01682", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_01683", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01684", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_01685", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01686", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "security_gates"]}}
+{"id": "train_01687", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01688", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01689", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_01690", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "documentation", "governance"]}}
+{"id": "train_01691", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "tooling"]}}
+{"id": "train_01692", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01693", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01694", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_01695", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01696", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_01697", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01698", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_01699", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01700", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01701", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_01702", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01703", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01704", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_01705", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01706", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01707", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01708", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01709", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_01710", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_01711", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01712", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01713", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_01714", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_01715", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_01716", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01717", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_01718", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01719", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_01720", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01721", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01722", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01723", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tooling"]}}
+{"id": "train_01724", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_01725", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01726", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_01727", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01728", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_01729", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_01730", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_01731", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01732", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_01733", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01734", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01735", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_01736", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_01737", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "reproducibility"]}}
+{"id": "train_01738", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01739", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "reproducibility", "tooling"]}}
+{"id": "train_01740", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01741", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_01742", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_01743", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_01744", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01745", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_01746", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_01747", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01748", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_01749", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_01750", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_01751", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01752", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_01753", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_01754", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01755", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01756", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01757", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01758", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01759", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01760", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_01761", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01762", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "reproducibility"]}}
+{"id": "train_01763", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_01764", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01765", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_01766", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01767", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01768", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_01769", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01770", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_01771", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01772", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_01773", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_01774", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01775", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01776", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "documentation"]}}
+{"id": "train_01777", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_01778", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01779", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01780", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "reproducibility"]}}
+{"id": "train_01781", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_01782", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "documentation"]}}
+{"id": "train_01783", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01784", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01785", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_01786", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "reproducibility"]}}
+{"id": "train_01787", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "reproducibility"]}}
+{"id": "train_01788", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01789", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "documentation", "reproducibility"]}}
+{"id": "train_01790", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_01791", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_01792", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_01793", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01794", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01795", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01796", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_01797", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_01798", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01799", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_01800", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_01801", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_01802", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_01803", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01804", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_01805", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "governance", "security_gates"]}}
+{"id": "train_01806", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01807", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01808", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_01809", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_01810", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tooling", "documentation"]}}
+{"id": "train_01811", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_01812", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01813", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_01814", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_01815", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01816", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01817", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_01818", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_01819", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_01820", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01821", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_01822", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_01823", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01824", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_01825", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01826", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01827", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01828", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01829", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_01830", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_01831", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_01832", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_01833", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_01834", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01835", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_01836", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01837", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_01838", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "reproducibility"]}}
+{"id": "train_01839", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_01840", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_01841", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_01842", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01843", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_01844", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_01845", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01846", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "evaluation_metrics"]}}
+{"id": "train_01847", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_01848", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01849", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01850", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_01851", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "reproducibility"]}}
+{"id": "train_01852", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_01853", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01854", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_01855", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_01856", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01857", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "ci_integration"]}}
+{"id": "train_01858", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01859", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01860", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01861", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01862", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_01863", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_01864", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_01865", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_01866", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_01867", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_01868", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01869", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01870", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_01871", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01872", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_01873", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_01874", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01875", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_01876", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01877", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_01878", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_01879", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01880", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_01881", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_01882", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01883", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_01884", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_01885", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01886", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_01887", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "documentation", "evaluation_metrics"]}}
+{"id": "train_01888", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "tests_are_truth"]}}
+{"id": "train_01889", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_01890", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_01891", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_01892", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_01893", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_01894", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_01895", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tests_are_truth", "tooling"]}}
+{"id": "train_01896", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01897", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_01898", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "reproducibility"]}}
+{"id": "train_01899", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_01900", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_01901", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_01902", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "evaluation_metrics"]}}
+{"id": "train_01903", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_01904", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_01905", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_01906", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_01907", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01908", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "documentation"]}}
+{"id": "train_01909", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_01910", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_01911", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_01912", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_01913", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_01914", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01915", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_01916", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_01917", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_01918", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_01919", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01920", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "ci_integration"]}}
+{"id": "train_01921", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_01922", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_01923", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_01924", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01925", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_01926", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_01927", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_01928", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_01929", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01930", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_01931", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_01932", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_01933", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_01934", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01935", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_01936", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01937", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01938", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "security_gates"]}}
+{"id": "train_01939", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_01940", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01941", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01942", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_01943", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_01944", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01945", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_01946", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_01947", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_01948", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "evaluation_metrics", "security_gates"]}}
+{"id": "train_01949", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01950", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_01951", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_01952", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "documentation"]}}
+{"id": "train_01953", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01954", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_01955", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_01956", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_01957", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_01958", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_01959", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_01960", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_01961", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "tooling"]}}
+{"id": "train_01962", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_01963", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_01964", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_01965", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_01966", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_01967", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01968", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_01969", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "security_gates", "governance"]}}
+{"id": "train_01970", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_01971", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_01972", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_01973", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "tooling"]}}
+{"id": "train_01974", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_01975", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "security_gates"]}}
+{"id": "train_01976", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_01977", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01978", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_01979", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_01980", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "governance"]}}
+{"id": "train_01981", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_01982", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_01983", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tests_are_truth", "governance"]}}
+{"id": "train_01984", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01985", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "documentation"]}}
+{"id": "train_01986", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "ci_integration", "security_gates"]}}
+{"id": "train_01987", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_01988", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_01989", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_01990", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_01991", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_01992", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_01993", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_01994", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "tooling"]}}
+{"id": "train_01995", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_01996", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_01997", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_01998", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_01999", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02000", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_02001", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02002", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02003", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tests_are_truth", "tooling"]}}
+{"id": "train_02004", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02005", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_02006", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02007", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_02008", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "documentation"]}}
+{"id": "train_02009", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02010", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02011", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02012", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_02013", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02014", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_02015", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_02016", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02017", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_02018", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02019", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02020", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02021", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_02022", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_02023", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_02024", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02025", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_02026", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "governance"]}}
+{"id": "train_02027", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02028", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02029", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "documentation"]}}
+{"id": "train_02030", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02031", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02032", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02033", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "governance", "tooling"]}}
+{"id": "train_02034", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02035", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02036", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_02037", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02038", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_02039", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02040", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02041", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02042", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_02043", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02044", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_02045", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02046", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_02047", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02048", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_02049", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02050", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_02051", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_02052", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_02053", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "governance"]}}
+{"id": "train_02054", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02055", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02056", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02057", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_02058", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02059", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02060", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_02061", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02062", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02063", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_02064", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02065", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02066", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02067", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02068", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02069", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_02070", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02071", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_02072", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_02073", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02074", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02075", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_02076", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_02077", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02078", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02079", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02080", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02081", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_02082", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_02083", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_02084", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02085", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02086", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02087", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02088", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02089", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_02090", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_02091", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02092", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_02093", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_02094", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_02095", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02096", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02097", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02098", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02099", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02100", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02101", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "security_gates", "documentation"]}}
+{"id": "train_02102", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_02103", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02104", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02105", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_02106", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tests_are_truth", "documentation"]}}
+{"id": "train_02107", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02108", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02109", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "documentation"]}}
+{"id": "train_02110", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_02111", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02112", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02113", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02114", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02115", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_02116", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02117", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02118", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02119", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "governance"]}}
+{"id": "train_02120", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02121", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_02122", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_02123", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02124", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_02125", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02126", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_02127", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02128", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_02129", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02130", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02131", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02132", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02133", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02134", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02135", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tooling", "tests_are_truth"]}}
+{"id": "train_02136", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_02137", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02138", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02139", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02140", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02141", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02142", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_02143", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02144", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "documentation", "reproducibility"]}}
+{"id": "train_02145", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02146", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02147", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_02148", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_02149", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_02150", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02151", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_02152", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02153", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02154", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02155", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_02156", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_02157", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_02158", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_02159", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_02160", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02161", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_02162", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_02163", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_02164", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02165", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02166", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02167", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_02168", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02169", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02170", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02171", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_02172", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_02173", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02174", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_02175", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_02176", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "security_gates"]}}
+{"id": "train_02177", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "documentation", "tooling"]}}
+{"id": "train_02178", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02179", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_02180", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02181", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_02182", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_02183", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02184", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_02185", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02186", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02187", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02188", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02189", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02190", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_02191", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_02192", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02193", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02194", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_02195", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_02196", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_02197", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02198", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02199", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02200", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_02201", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02202", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "ci_integration"]}}
+{"id": "train_02203", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_02204", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02205", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_02206", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_02207", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_02208", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02209", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02210", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02211", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02212", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "tests_are_truth"]}}
+{"id": "train_02213", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02214", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02215", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02216", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02217", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_02218", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_02219", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02220", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_02221", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "security_gates"]}}
+{"id": "train_02222", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_02223", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_02224", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_02225", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_02226", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02227", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02228", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02229", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_02230", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_02231", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02232", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02233", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02234", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_02235", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_02236", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02237", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02238", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02239", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_02240", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_02241", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02242", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02243", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_02244", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02245", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02246", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_02247", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_02248", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_02249", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_02250", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_02251", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02252", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_02253", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02254", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02255", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02256", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_02257", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02258", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_02259", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_02260", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02261", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_02262", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_02263", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02264", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02265", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_02266", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_02267", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "documentation"]}}
+{"id": "train_02268", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_02269", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_02270", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02271", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02272", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_02273", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_02274", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_02275", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02276", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02277", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02278", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02279", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_02280", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_02281", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02282", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02283", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_02284", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02285", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02286", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02287", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_02288", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02289", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_02290", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_02291", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02292", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02293", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02294", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_02295", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02296", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02297", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_02298", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02299", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02300", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02301", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02302", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02303", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_02304", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02305", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02306", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02307", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02308", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02309", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02310", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_02311", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02312", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02313", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "governance", "ci_integration"]}}
+{"id": "train_02314", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_02315", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02316", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_02317", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02318", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02319", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02320", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_02321", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02322", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02323", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02324", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_02325", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tooling"]}}
+{"id": "train_02326", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_02327", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_02328", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_02329", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02330", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02331", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02332", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02333", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02334", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_02335", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "governance", "tooling"]}}
+{"id": "train_02336", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_02337", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_02338", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_02339", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tooling"]}}
+{"id": "train_02340", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02341", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "reproducibility"]}}
+{"id": "train_02342", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02343", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02344", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_02345", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "ci_integration", "security_gates"]}}
+{"id": "train_02346", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02347", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02348", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_02349", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_02350", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02351", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02352", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_02353", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02354", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02355", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_02356", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_02357", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "security_gates"]}}
+{"id": "train_02358", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_02359", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_02360", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02361", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02362", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02363", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "governance"]}}
+{"id": "train_02364", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02365", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_02366", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02367", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_02368", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02369", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_02370", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02371", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tooling", "security_gates"]}}
+{"id": "train_02372", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02373", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_02374", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02375", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_02376", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02377", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "tests_are_truth"]}}
+{"id": "train_02378", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_02379", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_02380", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02381", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_02382", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02383", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_02384", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_02385", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_02386", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_02387", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02388", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02389", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_02390", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02391", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_02392", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02393", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_02394", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02395", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02396", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02397", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02398", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02399", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_02400", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02401", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_02402", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02403", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_02404", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02405", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02406", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02407", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02408", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "security_gates"]}}
+{"id": "train_02409", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02410", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "documentation", "tooling"]}}
+{"id": "train_02411", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_02412", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_02413", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02414", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_02415", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "documentation", "security_gates"]}}
+{"id": "train_02416", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02417", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_02418", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tests_are_truth", "tooling"]}}
+{"id": "train_02419", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02420", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "reproducibility"]}}
+{"id": "train_02421", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_02422", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02423", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_02424", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02425", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02426", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02427", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02428", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_02429", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02430", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_02431", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "ci_integration"]}}
+{"id": "train_02432", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_02433", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02434", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_02435", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_02436", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02437", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02438", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_02439", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_02440", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_02441", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "governance"]}}
+{"id": "train_02442", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02443", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02444", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02445", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02446", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02447", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_02448", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_02449", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_02450", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02451", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02452", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "reproducibility"]}}
+{"id": "train_02453", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "security_gates"]}}
+{"id": "train_02454", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "ci_integration"]}}
+{"id": "train_02455", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_02456", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02457", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_02458", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02459", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02460", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02461", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02462", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_02463", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02464", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_02465", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02466", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_02467", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02468", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "governance"]}}
+{"id": "train_02469", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02470", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_02471", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02472", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_02473", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02474", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_02475", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02476", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_02477", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02478", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_02479", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_02480", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_02481", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_02482", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02483", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "tooling"]}}
+{"id": "train_02484", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02485", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_02486", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_02487", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_02488", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_02489", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "security_gates", "tooling"]}}
+{"id": "train_02490", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_02491", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02492", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_02493", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02494", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_02495", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02496", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "documentation"]}}
+{"id": "train_02497", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_02498", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02499", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02500", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02501", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_02502", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02503", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02504", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02505", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02506", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_02507", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02508", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02509", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02510", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_02511", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02512", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02513", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02514", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02515", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_02516", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_02517", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02518", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02519", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "security_gates"]}}
+{"id": "train_02520", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02521", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "governance"]}}
+{"id": "train_02522", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02523", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02524", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02525", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_02526", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02527", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_02528", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_02529", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_02530", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_02531", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02532", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_02533", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_02534", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02535", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_02536", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02537", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_02538", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_02539", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_02540", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_02541", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_02542", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02543", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_02544", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02545", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_02546", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02547", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_02548", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02549", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02550", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02551", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02552", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02553", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "ci_integration", "governance"]}}
+{"id": "train_02554", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02555", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_02556", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02557", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_02558", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02559", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02560", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02561", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_02562", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_02563", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02564", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_02565", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_02566", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02567", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02568", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_02569", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02570", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02571", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02572", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02573", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_02574", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02575", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02576", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_02577", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02578", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "governance", "reproducibility"]}}
+{"id": "train_02579", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02580", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02581", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02582", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02583", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "reproducibility"]}}
+{"id": "train_02584", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02585", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_02586", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02587", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02588", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tests_are_truth", "security_gates"]}}
+{"id": "train_02589", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_02590", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02591", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02592", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_02593", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02594", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02595", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02596", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02597", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02598", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02599", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02600", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02601", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02602", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02603", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_02604", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02605", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02606", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "security_gates", "governance"]}}
+{"id": "train_02607", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_02608", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02609", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_02610", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02611", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02612", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_02613", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_02614", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_02615", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02616", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_02617", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02618", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "governance"]}}
+{"id": "train_02619", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_02620", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02621", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_02622", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_02623", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_02624", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_02625", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_02626", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02627", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02628", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "ci_integration"]}}
+{"id": "train_02629", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_02630", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_02631", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_02632", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_02633", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02634", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02635", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_02636", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_02637", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "evaluation_metrics"]}}
+{"id": "train_02638", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02639", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_02640", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_02641", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02642", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02643", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_02644", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02645", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_02646", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_02647", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "documentation"]}}
+{"id": "train_02648", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02649", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02650", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_02651", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02652", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02653", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_02654", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_02655", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_02656", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_02657", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tooling"]}}
+{"id": "train_02658", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_02659", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02660", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_02661", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tooling", "ci_integration"]}}
+{"id": "train_02662", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02663", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02664", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_02665", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_02666", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_02667", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02668", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02669", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_02670", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02671", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_02672", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02673", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_02674", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "governance"]}}
+{"id": "train_02675", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "security_gates"]}}
+{"id": "train_02676", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_02677", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_02678", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02679", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_02680", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02681", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_02682", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_02683", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "tooling"]}}
+{"id": "train_02684", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_02685", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_02686", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02687", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02688", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02689", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02690", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02691", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02692", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02693", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02694", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02695", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02696", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "evaluation_metrics"]}}
+{"id": "train_02697", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02698", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_02699", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_02700", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02701", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02702", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_02703", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_02704", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_02705", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02706", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tooling"]}}
+{"id": "train_02707", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02708", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02709", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_02710", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02711", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "governance"]}}
+{"id": "train_02712", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_02713", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02714", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "security_gates"]}}
+{"id": "train_02715", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02716", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02717", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02718", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02719", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_02720", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_02721", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02722", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_02723", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_02724", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_02725", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_02726", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02727", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_02728", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_02729", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_02730", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02731", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "governance", "security_gates"]}}
+{"id": "train_02732", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02733", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02734", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02735", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02736", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_02737", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02738", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_02739", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02740", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02741", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02742", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02743", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_02744", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_02745", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_02746", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_02747", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02748", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02749", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_02750", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02751", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_02752", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02753", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_02754", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_02755", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02756", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02757", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_02758", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_02759", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02760", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_02761", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02762", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_02763", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02764", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_02765", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tests_are_truth", "tooling"]}}
+{"id": "train_02766", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_02767", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02768", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02769", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02770", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_02771", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02772", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02773", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_02774", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "security_gates"]}}
+{"id": "train_02775", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02776", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "ci_integration", "governance"]}}
+{"id": "train_02777", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_02778", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02779", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_02780", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02781", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "ci_integration", "security_gates"]}}
+{"id": "train_02782", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_02783", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_02784", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02785", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02786", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_02787", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_02788", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02789", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02790", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_02791", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02792", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02793", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_02794", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02795", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02796", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02797", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_02798", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_02799", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_02800", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_02801", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02802", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02803", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02804", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02805", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "governance", "reproducibility"]}}
+{"id": "train_02806", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02807", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_02808", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "reproducibility"]}}
+{"id": "train_02809", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02810", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_02811", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "security_gates"]}}
+{"id": "train_02812", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "security_gates", "governance"]}}
+{"id": "train_02813", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02814", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_02815", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_02816", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02817", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02818", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02819", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_02820", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_02821", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "documentation"]}}
+{"id": "train_02822", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tooling"]}}
+{"id": "train_02823", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_02824", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02825", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_02826", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02827", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tooling", "tests_are_truth"]}}
+{"id": "train_02828", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_02829", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02830", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_02831", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_02832", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02833", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02834", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "documentation"]}}
+{"id": "train_02835", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02836", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02837", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_02838", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_02839", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "governance", "ci_integration"]}}
+{"id": "train_02840", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_02841", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_02842", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_02843", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_02844", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_02845", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_02846", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "security_gates"]}}
+{"id": "train_02847", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02848", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02849", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "reproducibility"]}}
+{"id": "train_02850", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02851", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02852", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02853", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_02854", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02855", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_02856", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02857", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02858", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_02859", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02860", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_02861", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02862", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_02863", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_02864", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02865", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_02866", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_02867", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "tests_are_truth"]}}
+{"id": "train_02868", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02869", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_02870", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_02871", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_02872", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_02873", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_02874", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_02875", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_02876", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_02877", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_02878", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_02879", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02880", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02881", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02882", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_02883", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_02884", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02885", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_02886", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02887", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_02888", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_02889", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_02890", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_02891", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_02892", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02893", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tooling"]}}
+{"id": "train_02894", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02895", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02896", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02897", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02898", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02899", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_02900", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_02901", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_02902", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_02903", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_02904", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_02905", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_02906", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_02907", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_02908", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02909", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_02910", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_02911", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02912", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_02913", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_02914", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_02915", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02916", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_02917", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02918", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_02919", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_02920", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_02921", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02922", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_02923", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02924", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02925", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_02926", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_02927", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_02928", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02929", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02930", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_02931", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_02932", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02933", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02934", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_02935", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02936", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_02937", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "security_gates"]}}
+{"id": "train_02938", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_02939", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_02940", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_02941", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_02942", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_02943", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02944", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_02945", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_02946", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02947", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_02948", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "tooling"]}}
+{"id": "train_02949", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_02950", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "security_gates"]}}
+{"id": "train_02951", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "ci_integration", "governance"]}}
+{"id": "train_02952", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_02953", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_02954", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_02955", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_02956", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_02957", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_02958", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_02959", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "tooling"]}}
+{"id": "train_02960", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_02961", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02962", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_02963", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_02964", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_02965", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_02966", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_02967", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_02968", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_02969", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_02970", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_02971", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "governance"]}}
+{"id": "train_02972", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02973", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02974", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02975", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_02976", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_02977", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_02978", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02979", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_02980", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_02981", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_02982", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_02983", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "ci_integration"]}}
+{"id": "train_02984", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "governance"]}}
+{"id": "train_02985", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_02986", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_02987", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_02988", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_02989", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_02990", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_02991", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_02992", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "reproducibility", "ci_integration"]}}
+{"id": "train_02993", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tooling"]}}
+{"id": "train_02994", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02995", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "governance", "tooling"]}}
+{"id": "train_02996", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_02997", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_02998", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_02999", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_03000", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_03001", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_03002", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03003", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03004", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03005", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03006", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03007", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tests_are_truth", "tooling"]}}
+{"id": "train_03008", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "tooling"]}}
+{"id": "train_03009", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03010", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03011", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03012", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_03013", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03014", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03015", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03016", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03017", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03018", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03019", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03020", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03021", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03022", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_03023", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03024", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "documentation"]}}
+{"id": "train_03025", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03026", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "evaluation_metrics"]}}
+{"id": "train_03027", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_03028", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "documentation"]}}
+{"id": "train_03029", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03030", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03031", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_03032", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03033", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_03034", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03035", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03036", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03037", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_03038", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "governance"]}}
+{"id": "train_03039", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_03040", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03041", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03042", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03043", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03044", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "documentation"]}}
+{"id": "train_03045", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03046", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03047", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_03048", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_03049", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03050", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03051", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03052", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_03053", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_03054", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03055", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03056", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03057", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03058", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03059", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_03060", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_03061", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "documentation"]}}
+{"id": "train_03062", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03063", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03064", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03065", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "ci_integration", "reproducibility"]}}
+{"id": "train_03066", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_03067", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03068", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_03069", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03070", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03071", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03072", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_03073", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03074", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03075", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tests_are_truth"]}}
+{"id": "train_03076", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03077", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03078", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03079", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03080", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03081", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03082", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_03083", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_03084", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_03085", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_03086", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_03087", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_03088", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_03089", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03090", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_03091", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_03092", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_03093", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_03094", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03095", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03096", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_03097", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_03098", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_03099", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03100", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03101", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03102", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03103", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03104", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_03105", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03106", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_03107", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "security_gates"]}}
+{"id": "train_03108", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03109", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03110", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03111", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_03112", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_03113", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_03114", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_03115", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03116", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_03117", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_03118", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_03119", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_03120", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03121", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_03122", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03123", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_03124", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03125", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03126", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03127", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03128", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_03129", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03130", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03131", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03132", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03133", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03134", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03135", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_03136", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_03137", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03138", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_03139", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_03140", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_03141", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_03142", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_03143", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_03144", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_03145", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03146", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_03147", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03148", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03149", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03150", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03151", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03152", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_03153", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03154", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03155", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03156", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03157", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_03158", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03159", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "tooling"]}}
+{"id": "train_03160", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_03161", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_03162", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "governance"]}}
+{"id": "train_03163", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03164", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03165", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03166", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_03167", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03168", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03169", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_03170", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tooling", "evaluation_metrics"]}}
+{"id": "train_03171", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_03172", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_03173", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_03174", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "tests_are_truth"]}}
+{"id": "train_03175", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_03176", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "tooling"]}}
+{"id": "train_03177", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_03178", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03179", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "reproducibility"]}}
+{"id": "train_03180", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03181", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_03182", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03183", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_03184", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03185", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_03186", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03187", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03188", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03189", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_03190", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_03191", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_03192", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03193", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_03194", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "governance", "ci_integration"]}}
+{"id": "train_03195", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03196", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03197", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03198", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03199", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03200", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03201", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_03202", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "security_gates"]}}
+{"id": "train_03203", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_03204", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03205", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_03206", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_03207", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_03208", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03209", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_03210", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "documentation"]}}
+{"id": "train_03211", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tooling", "ci_integration"]}}
+{"id": "train_03212", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_03213", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tooling", "evaluation_metrics"]}}
+{"id": "train_03214", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_03215", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_03216", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03217", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_03218", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03219", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03220", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03221", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03222", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_03223", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03224", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_03225", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_03226", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "governance", "documentation"]}}
+{"id": "train_03227", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03228", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_03229", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_03230", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03231", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_03232", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03233", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03234", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_03235", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03236", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03237", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "evaluation_metrics"]}}
+{"id": "train_03238", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03239", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03240", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03241", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03242", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_03243", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03244", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_03245", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03246", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_03247", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03248", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tests_are_truth", "tooling"]}}
+{"id": "train_03249", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03250", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03251", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_03252", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03253", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_03254", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_03255", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_03256", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "ci_integration", "reproducibility"]}}
+{"id": "train_03257", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_03258", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "reproducibility"]}}
+{"id": "train_03259", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_03260", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03261", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03262", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_03263", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03264", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_03265", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_03266", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03267", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03268", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_03269", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_03270", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03271", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03272", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_03273", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_03274", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_03275", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03276", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_03277", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03278", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03279", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03280", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03281", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03282", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03283", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03284", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03285", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03286", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03287", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03288", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03289", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_03290", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_03291", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_03292", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_03293", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03294", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_03295", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_03296", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03297", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03298", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03299", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03300", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03301", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_03302", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03303", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03304", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03305", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03306", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_03307", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03308", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_03309", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03310", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_03311", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_03312", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_03313", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03314", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03315", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_03316", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_03317", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_03318", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_03319", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03320", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_03321", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03322", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03323", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03324", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_03325", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_03326", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03327", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_03328", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03329", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_03330", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_03331", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03332", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03333", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_03334", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03335", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_03336", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03337", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03338", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_03339", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_03340", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_03341", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03342", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_03343", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "documentation", "security_gates"]}}
+{"id": "train_03344", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03345", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tooling", "governance"]}}
+{"id": "train_03346", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03347", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03348", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_03349", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03350", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_03351", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03352", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_03353", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_03354", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03355", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03356", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_03357", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "tests_are_truth"]}}
+{"id": "train_03358", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "ci_integration"]}}
+{"id": "train_03359", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_03360", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03361", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_03362", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03363", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_03364", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03365", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_03366", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03367", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_03368", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_03369", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_03370", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03371", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tooling", "ci_integration"]}}
+{"id": "train_03372", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03373", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03374", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_03375", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03376", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03377", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "reproducibility"]}}
+{"id": "train_03378", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03379", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03380", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "tooling"]}}
+{"id": "train_03381", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_03382", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tooling"]}}
+{"id": "train_03383", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_03384", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_03385", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_03386", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03387", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03388", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03389", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_03390", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "documentation"]}}
+{"id": "train_03391", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03392", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03393", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03394", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_03395", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_03396", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03397", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_03398", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "tooling"]}}
+{"id": "train_03399", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03400", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03401", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_03402", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03403", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03404", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "reproducibility", "security_gates"]}}
+{"id": "train_03405", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03406", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03407", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_03408", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_03409", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03410", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "evaluation_metrics"]}}
+{"id": "train_03411", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "security_gates"]}}
+{"id": "train_03412", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_03413", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03414", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03415", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_03416", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03417", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_03418", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_03419", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "documentation", "security_gates"]}}
+{"id": "train_03420", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03421", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_03422", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_03423", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_03424", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_03425", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_03426", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_03427", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03428", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_03429", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03430", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03431", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03432", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "reproducibility", "ci_integration"]}}
+{"id": "train_03433", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_03434", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_03435", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_03436", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_03437", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03438", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_03439", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03440", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03441", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03442", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03443", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_03444", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03445", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "documentation"]}}
+{"id": "train_03446", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03447", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03448", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_03449", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03450", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03451", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03452", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03453", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "ci_integration"]}}
+{"id": "train_03454", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03455", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03456", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_03457", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_03458", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03459", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03460", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03461", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "documentation"]}}
+{"id": "train_03462", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03463", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03464", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03465", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03466", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03467", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_03468", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03469", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03470", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_03471", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03472", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03473", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_03474", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_03475", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_03476", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03477", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03478", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_03479", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03480", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_03481", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_03482", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03483", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_03484", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_03485", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03486", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_03487", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03488", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_03489", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_03490", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03491", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03492", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_03493", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "governance", "security_gates"]}}
+{"id": "train_03494", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03495", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03496", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "documentation"]}}
+{"id": "train_03497", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03498", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03499", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03500", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03501", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03502", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03503", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "ci_integration", "governance"]}}
+{"id": "train_03504", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_03505", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_03506", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_03507", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03508", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_03509", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "documentation", "tooling"]}}
+{"id": "train_03510", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03511", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03512", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_03513", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03514", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03515", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03516", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_03517", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_03518", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_03519", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03520", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "reproducibility"]}}
+{"id": "train_03521", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_03522", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "tooling"]}}
+{"id": "train_03523", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_03524", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_03525", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "tooling"]}}
+{"id": "train_03526", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03527", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "tooling", "reproducibility"]}}
+{"id": "train_03528", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03529", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03530", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_03531", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "tests_are_truth"]}}
+{"id": "train_03532", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03533", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03534", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_03535", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03536", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03537", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03538", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03539", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03540", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03541", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_03542", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03543", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03544", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03545", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_03546", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_03547", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03548", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03549", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_03550", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_03551", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "evaluation_metrics"]}}
+{"id": "train_03552", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03553", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03554", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_03555", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "evaluation_metrics", "governance"]}}
+{"id": "train_03556", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_03557", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03558", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "governance"]}}
+{"id": "train_03559", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_03560", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_03561", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03562", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_03563", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03564", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_03565", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03566", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03567", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03568", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "tests_are_truth"]}}
+{"id": "train_03569", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "governance"]}}
+{"id": "train_03570", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_03571", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03572", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_03573", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03574", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03575", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03576", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_03577", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "documentation"]}}
+{"id": "train_03578", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03579", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03580", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03581", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_03582", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03583", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03584", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_03585", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_03586", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_03587", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03588", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03589", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03590", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03591", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03592", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03593", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03594", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_03595", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_03596", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03597", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03598", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "documentation", "evaluation_metrics"]}}
+{"id": "train_03599", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_03600", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03601", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_03602", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03603", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03604", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_03605", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_03606", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03607", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03608", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_03609", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "security_gates"]}}
+{"id": "train_03610", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "documentation", "reproducibility"]}}
+{"id": "train_03611", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_03612", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03613", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_03614", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_03615", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_03616", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "ci_integration"]}}
+{"id": "train_03617", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03618", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03619", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03620", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03621", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03622", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_03623", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_03624", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03625", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03626", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03627", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03628", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03629", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03630", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03631", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03632", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_03633", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_03634", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_03635", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03636", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_03637", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03638", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "tooling"]}}
+{"id": "train_03639", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_03640", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03641", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03642", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03643", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tooling", "ci_integration"]}}
+{"id": "train_03644", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03645", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03646", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03647", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_03648", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03649", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03650", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "documentation", "ci_integration"]}}
+{"id": "train_03651", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "ci_integration"]}}
+{"id": "train_03652", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_03653", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03654", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03655", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "governance", "ci_integration"]}}
+{"id": "train_03656", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_03657", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_03658", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03659", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03660", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03661", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "ci_integration"]}}
+{"id": "train_03662", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_03663", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03664", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_03665", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_03666", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03667", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03668", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03669", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03670", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03671", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03672", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_03673", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_03674", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tooling"]}}
+{"id": "train_03675", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03676", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_03677", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "security_gates"]}}
+{"id": "train_03678", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "reproducibility", "documentation"]}}
+{"id": "train_03679", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03680", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "governance"]}}
+{"id": "train_03681", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03682", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_03683", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "tests_are_truth"]}}
+{"id": "train_03684", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03685", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03686", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_03687", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_03688", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "security_gates"]}}
+{"id": "train_03689", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03690", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03691", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tooling", "documentation"]}}
+{"id": "train_03692", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_03693", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_03694", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "documentation"]}}
+{"id": "train_03695", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_03696", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03697", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03698", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03699", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_03700", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_03701", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "governance", "documentation"]}}
+{"id": "train_03702", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03703", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03704", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03705", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03706", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03707", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03708", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "documentation"]}}
+{"id": "train_03709", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03710", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03711", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_03712", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_03713", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03714", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03715", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03716", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_03717", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "security_gates"]}}
+{"id": "train_03718", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_03719", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03720", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "governance", "reproducibility"]}}
+{"id": "train_03721", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03722", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_03723", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03724", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03725", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03726", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03727", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03728", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03729", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_03730", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_03731", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "reproducibility"]}}
+{"id": "train_03732", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03733", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03734", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "reproducibility"]}}
+{"id": "train_03735", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_03736", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_03737", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_03738", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03739", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03740", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03741", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03742", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "governance", "evaluation_metrics"]}}
+{"id": "train_03743", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03744", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03745", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_03746", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03747", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03748", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_03749", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_03750", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03751", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_03752", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03753", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_03754", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03755", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03756", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03757", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_03758", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03759", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03760", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_03761", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_03762", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03763", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_03764", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03765", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03766", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03767", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03768", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03769", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_03770", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03771", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_03772", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03773", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03774", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "governance", "ci_integration"]}}
+{"id": "train_03775", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03776", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "documentation"]}}
+{"id": "train_03777", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_03778", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_03779", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "evaluation_metrics"]}}
+{"id": "train_03780", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_03781", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03782", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03783", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03784", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03785", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_03786", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "ci_integration", "tooling"]}}
+{"id": "train_03787", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03788", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03789", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_03790", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_03791", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "security_gates", "documentation"]}}
+{"id": "train_03792", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_03793", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03794", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03795", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_03796", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03797", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "reproducibility"]}}
+{"id": "train_03798", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_03799", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_03800", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03801", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_03802", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03803", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tests_are_truth"]}}
+{"id": "train_03804", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03805", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "reproducibility", "tooling"]}}
+{"id": "train_03806", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "ci_integration"]}}
+{"id": "train_03807", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03808", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03809", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03810", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_03811", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03812", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03813", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03814", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03815", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_03816", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03817", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03818", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03819", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_03820", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03821", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_03822", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_03823", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03824", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03825", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_03826", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_03827", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03828", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03829", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03830", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_03831", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_03832", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03833", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_03834", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03835", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03836", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03837", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_03838", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_03839", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03840", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "documentation", "governance"]}}
+{"id": "train_03841", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_03842", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_03843", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_03844", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_03845", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03846", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "ci_integration"]}}
+{"id": "train_03847", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03848", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03849", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03850", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_03851", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03852", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "governance", "reproducibility"]}}
+{"id": "train_03853", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_03854", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03855", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_03856", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03857", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_03858", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03859", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "security_gates"]}}
+{"id": "train_03860", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_03861", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_03862", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_03863", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_03864", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_03865", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_03866", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tooling", "ci_integration"]}}
+{"id": "train_03867", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03868", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "reproducibility", "tooling"]}}
+{"id": "train_03869", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "ci_integration"]}}
+{"id": "train_03870", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03871", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_03872", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_03873", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03874", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "tooling", "security_gates"]}}
+{"id": "train_03875", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_03876", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_03877", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03878", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "security_gates"]}}
+{"id": "train_03879", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_03880", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "security_gates"]}}
+{"id": "train_03881", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03882", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tooling", "ci_integration"]}}
+{"id": "train_03883", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03884", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03885", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_03886", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03887", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_03888", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03889", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03890", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03891", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_03892", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_03893", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03894", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03895", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_03896", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03897", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03898", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03899", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_03900", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03901", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03902", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_03903", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "documentation", "evaluation_metrics"]}}
+{"id": "train_03904", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "tests_are_truth"]}}
+{"id": "train_03905", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "tooling"]}}
+{"id": "train_03906", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_03907", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03908", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03909", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03910", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "ci_integration"]}}
+{"id": "train_03911", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_03912", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_03913", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03914", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_03915", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_03916", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03917", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_03918", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03919", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03920", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03921", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03922", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_03923", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_03924", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_03925", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03926", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_03927", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03928", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_03929", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_03930", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_03931", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_03932", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_03933", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_03934", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_03935", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03936", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_03937", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_03938", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_03939", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_03940", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "security_gates"]}}
+{"id": "train_03941", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "reproducibility", "tests_are_truth"]}}
+{"id": "train_03942", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03943", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_03944", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03945", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "documentation"]}}
+{"id": "train_03946", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_03947", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03948", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_03949", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_03950", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_03951", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03952", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_03953", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03954", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "evaluation_metrics"]}}
+{"id": "train_03955", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03956", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "governance", "tests_are_truth"]}}
+{"id": "train_03957", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_03958", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "governance"]}}
+{"id": "train_03959", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_03960", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_03961", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "governance", "security_gates"]}}
+{"id": "train_03962", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_03963", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_03964", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03965", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03966", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_03967", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_03968", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_03969", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_03970", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_03971", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03972", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_03973", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_03974", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_03975", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_03976", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_03977", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_03978", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_03979", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03980", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "security_gates", "ci_integration"]}}
+{"id": "train_03981", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03982", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_03983", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_03984", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_03985", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03986", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_03987", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_03988", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_03989", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "reproducibility", "security_gates"]}}
+{"id": "train_03990", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_03991", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_03992", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_03993", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "reproducibility"]}}
+{"id": "train_03994", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_03995", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_03996", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_03997", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_03998", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_03999", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04000", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04001", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04002", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04003", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "documentation", "governance"]}}
+{"id": "train_04004", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_04005", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_04006", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "ci_integration", "security_gates"]}}
+{"id": "train_04007", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_04008", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_04009", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04010", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "security_gates", "governance"]}}
+{"id": "train_04011", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_04012", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04013", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "governance"]}}
+{"id": "train_04014", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "tooling"]}}
+{"id": "train_04015", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04016", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04017", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "security_gates"]}}
+{"id": "train_04018", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_04019", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04020", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04021", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_04022", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_04023", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04024", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_04025", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04026", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_04027", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04028", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04029", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04030", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04031", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_04032", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04033", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04034", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04035", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04036", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04037", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04038", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "documentation", "ci_integration"]}}
+{"id": "train_04039", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04040", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04041", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tooling", "tests_are_truth"]}}
+{"id": "train_04042", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_04043", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_04044", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04045", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "documentation", "tooling"]}}
+{"id": "train_04046", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_04047", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04048", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_04049", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04050", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_04051", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04052", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04053", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_04054", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "governance", "tooling"]}}
+{"id": "train_04055", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "security_gates"]}}
+{"id": "train_04056", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_04057", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04058", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_04059", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_04060", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "documentation"]}}
+{"id": "train_04061", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04062", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "reproducibility", "security_gates"]}}
+{"id": "train_04063", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04064", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "ci_integration", "reproducibility"]}}
+{"id": "train_04065", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04066", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04067", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04068", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_04069", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04070", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04071", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_04072", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04073", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04074", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04075", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04076", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04077", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04078", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "ci_integration"]}}
+{"id": "train_04079", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04080", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "tooling"]}}
+{"id": "train_04081", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04082", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_04083", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04084", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_04085", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_04086", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_04087", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04088", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tooling", "evaluation_metrics"]}}
+{"id": "train_04089", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04090", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04091", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04092", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04093", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04094", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_04095", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04096", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04097", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_04098", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_04099", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_04100", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04101", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04102", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04103", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04104", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04105", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04106", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_04107", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04108", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "documentation"]}}
+{"id": "train_04109", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "evaluation_metrics", "documentation"]}}
+{"id": "train_04110", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_04111", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_04112", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04113", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04114", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_04115", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04116", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04117", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04118", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04119", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04120", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_04121", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04122", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_04123", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "governance", "tests_are_truth"]}}
+{"id": "train_04124", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04125", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04126", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "documentation"]}}
+{"id": "train_04127", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04128", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04129", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "ci_integration", "documentation"]}}
+{"id": "train_04130", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_04131", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_04132", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_04133", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04134", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04135", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_04136", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04137", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_04138", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04139", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04140", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04141", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_04142", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04143", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "tooling"]}}
+{"id": "train_04144", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04145", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04146", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "reproducibility"]}}
+{"id": "train_04147", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04148", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04149", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_04150", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04151", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04152", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04153", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04154", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04155", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "tests_are_truth"]}}
+{"id": "train_04156", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_04157", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "ci_integration"]}}
+{"id": "train_04158", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04159", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04160", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04161", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04162", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "documentation"]}}
+{"id": "train_04163", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_04164", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "documentation", "ci_integration"]}}
+{"id": "train_04165", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04166", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04167", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04168", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "documentation", "governance"]}}
+{"id": "train_04169", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04170", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_04171", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04172", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "tooling"]}}
+{"id": "train_04173", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04174", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_04175", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "tooling"]}}
+{"id": "train_04176", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_04177", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04178", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_04179", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04180", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04181", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04182", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04183", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04184", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_04185", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "tooling"]}}
+{"id": "train_04186", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04187", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "security_gates", "reproducibility"]}}
+{"id": "train_04188", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04189", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04190", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04191", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04192", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04193", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "security_gates"]}}
+{"id": "train_04194", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_04195", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_04196", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_04197", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04198", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04199", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "governance"]}}
+{"id": "train_04200", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "security_gates", "tooling"]}}
+{"id": "train_04201", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04202", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04203", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04204", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_04205", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "governance", "tests_are_truth"]}}
+{"id": "train_04206", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04207", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04208", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04209", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04210", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "security_gates"]}}
+{"id": "train_04211", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04212", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_04213", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_04214", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "security_gates"]}}
+{"id": "train_04215", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04216", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_04217", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04218", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_04219", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04220", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04221", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_04222", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04223", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_04224", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "reproducibility", "security_gates"]}}
+{"id": "train_04225", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_04226", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04227", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04228", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04229", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04230", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04231", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_04232", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "governance"]}}
+{"id": "train_04233", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04234", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_04235", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04236", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04237", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04238", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04239", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_04240", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04241", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04242", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04243", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04244", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "governance", "reproducibility"]}}
+{"id": "train_04245", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_04246", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_04247", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_04248", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04249", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "ci_integration"]}}
+{"id": "train_04250", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_04251", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04252", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "reproducibility"]}}
+{"id": "train_04253", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "ci_integration", "governance"]}}
+{"id": "train_04254", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04255", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04256", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04257", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_04258", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "reproducibility"]}}
+{"id": "train_04259", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04260", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_04261", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "governance", "evaluation_metrics"]}}
+{"id": "train_04262", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "security_gates"]}}
+{"id": "train_04263", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04264", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04265", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04266", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04267", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "ci_integration", "tooling"]}}
+{"id": "train_04268", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_04269", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04270", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04271", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_04272", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_04273", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04274", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "governance"]}}
+{"id": "train_04275", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "governance"]}}
+{"id": "train_04276", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "governance"]}}
+{"id": "train_04277", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_04278", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "reproducibility", "security_gates"]}}
+{"id": "train_04279", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tooling", "reproducibility"]}}
+{"id": "train_04280", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_04281", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04282", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_04283", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_04284", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04285", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_04286", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "ci_integration", "documentation"]}}
+{"id": "train_04287", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "reproducibility", "documentation"]}}
+{"id": "train_04288", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04289", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04290", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_04291", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "evaluation_metrics"]}}
+{"id": "train_04292", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04293", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_04294", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04295", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04296", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04297", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_04298", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04299", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04300", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04301", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04302", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04303", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "evaluation_metrics"]}}
+{"id": "train_04304", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04305", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "security_gates"]}}
+{"id": "train_04306", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tooling", "documentation"]}}
+{"id": "train_04307", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04308", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_04309", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04310", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "evaluation_metrics"]}}
+{"id": "train_04311", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_04312", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "tooling"]}}
+{"id": "train_04313", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04314", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_04315", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_04316", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04317", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_04318", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04319", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "evaluation_metrics"]}}
+{"id": "train_04320", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04321", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_04322", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04323", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "reproducibility"]}}
+{"id": "train_04324", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_04325", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_04326", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04327", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04328", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "governance"]}}
+{"id": "train_04329", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "governance"]}}
+{"id": "train_04330", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04331", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "reproducibility"]}}
+{"id": "train_04332", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "governance"]}}
+{"id": "train_04333", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04334", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04335", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "reproducibility"]}}
+{"id": "train_04336", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "governance"]}}
+{"id": "train_04337", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_04338", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04339", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04340", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04341", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04342", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04343", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04344", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04345", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_04346", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_04347", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "security_gates", "reproducibility"]}}
+{"id": "train_04348", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_04349", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04350", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04351", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "tooling", "governance"]}}
+{"id": "train_04352", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_04353", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04354", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_04355", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tooling", "evaluation_metrics"]}}
+{"id": "train_04356", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04357", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04358", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_04359", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_04360", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04361", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04362", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04363", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "tests_are_truth"]}}
+{"id": "train_04364", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_04365", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04366", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_04367", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04368", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "ci_integration", "security_gates"]}}
+{"id": "train_04369", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_04370", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_04371", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "governance"]}}
+{"id": "train_04372", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_04373", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04374", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "reproducibility", "security_gates"]}}
+{"id": "train_04375", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04376", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04377", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "security_gates", "documentation"]}}
+{"id": "train_04378", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04379", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_04380", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04381", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "reproducibility", "governance"]}}
+{"id": "train_04382", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04383", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04384", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04385", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04386", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04387", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04388", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04389", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "governance"]}}
+{"id": "train_04390", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04391", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "tests_are_truth", "governance"]}}
+{"id": "train_04392", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04393", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04394", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04395", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "tooling", "documentation"]}}
+{"id": "train_04396", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tooling"]}}
+{"id": "train_04397", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "governance"]}}
+{"id": "train_04398", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04399", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04400", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_04401", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "governance", "security_gates"]}}
+{"id": "train_04402", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04403", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "ci_integration"]}}
+{"id": "train_04404", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04405", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_04406", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04407", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04408", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "security_gates"]}}
+{"id": "train_04409", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "governance", "tests_are_truth"]}}
+{"id": "train_04410", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04411", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04412", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "reproducibility", "documentation"]}}
+{"id": "train_04413", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "documentation"]}}
+{"id": "train_04414", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_04415", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "governance"]}}
+{"id": "train_04416", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "governance", "tooling"]}}
+{"id": "train_04417", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04418", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04419", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04420", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_04421", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tooling", "evaluation_metrics"]}}
+{"id": "train_04422", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_04423", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_04424", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04425", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tooling"]}}
+{"id": "train_04426", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04427", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_04428", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04429", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04430", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04431", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_04432", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_04433", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04434", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04435", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "tooling", "governance"]}}
+{"id": "train_04436", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04437", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04438", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_04439", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_04440", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_04441", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_04442", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_04443", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04444", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_04445", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_04446", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04447", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_04448", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04449", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04450", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_04451", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04452", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_04453", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "governance", "evaluation_metrics"]}}
+{"id": "train_04454", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04455", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04456", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04457", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "security_gates"]}}
+{"id": "train_04458", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04459", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04460", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_04461", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04462", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04463", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04464", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04465", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04466", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04467", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04468", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "ci_integration"]}}
+{"id": "train_04469", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "governance", "ci_integration"]}}
+{"id": "train_04470", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04471", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04472", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_04473", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "governance"]}}
+{"id": "train_04474", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04475", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "governance"]}}
+{"id": "train_04476", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04477", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "tooling"]}}
+{"id": "train_04478", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04479", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04480", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04481", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04482", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "governance"]}}
+{"id": "train_04483", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_04484", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04485", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04486", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tooling", "ci_integration"]}}
+{"id": "train_04487", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04488", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_04489", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "ci_integration", "documentation"]}}
+{"id": "train_04490", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04491", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04492", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04493", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04494", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04495", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_04496", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04497", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "reproducibility"]}}
+{"id": "train_04498", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "security_gates", "tooling"]}}
+{"id": "train_04499", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "reproducibility"]}}
+{"id": "train_04500", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04501", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "documentation"]}}
+{"id": "train_04502", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_04503", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "documentation", "tests_are_truth"]}}
+{"id": "train_04504", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "reproducibility", "tooling"]}}
+{"id": "train_04505", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04506", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_04507", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04508", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04509", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_04510", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04511", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04512", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_04513", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_04514", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "tooling"]}}
+{"id": "train_04515", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "ci_integration", "documentation"]}}
+{"id": "train_04516", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04517", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04518", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04519", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_04520", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_04521", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04522", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04523", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04524", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_04525", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04526", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04527", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "documentation", "evaluation_metrics"]}}
+{"id": "train_04528", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04529", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04530", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04531", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04532", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04533", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "documentation", "evaluation_metrics"]}}
+{"id": "train_04534", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04535", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04536", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_04537", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "tooling"]}}
+{"id": "train_04538", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04539", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "tooling"]}}
+{"id": "train_04540", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "governance", "tooling"]}}
+{"id": "train_04541", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_04542", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "governance"]}}
+{"id": "train_04543", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04544", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_04545", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_04546", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04547", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04548", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "ci_integration", "governance"]}}
+{"id": "train_04549", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "governance", "tests_are_truth"]}}
+{"id": "train_04550", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04551", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04552", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04553", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "tooling"]}}
+{"id": "train_04554", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "ci_integration", "reproducibility"]}}
+{"id": "train_04555", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04556", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "ci_integration", "reproducibility"]}}
+{"id": "train_04557", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04558", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04559", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04560", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04561", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04562", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_04563", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_04564", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04565", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04566", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04567", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "governance"]}}
+{"id": "train_04568", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_04569", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04570", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04571", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_04572", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04573", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "documentation", "ci_integration"]}}
+{"id": "train_04574", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04575", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "security_gates"]}}
+{"id": "train_04576", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "tooling", "tests_are_truth"]}}
+{"id": "train_04577", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04578", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_04579", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04580", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04581", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "security_gates"]}}
+{"id": "train_04582", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_04583", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04584", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04585", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04586", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04587", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04588", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "tooling", "tests_are_truth"]}}
+{"id": "train_04589", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04590", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "governance"]}}
+{"id": "train_04591", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "security_gates"]}}
+{"id": "train_04592", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04593", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_04594", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "security_gates", "tests_are_truth"]}}
+{"id": "train_04595", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_04596", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04597", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04598", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04599", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "tooling"]}}
+{"id": "train_04600", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04601", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04602", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "security_gates", "tooling"]}}
+{"id": "train_04603", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "documentation", "repo_scale_reasoning"]}}
+{"id": "train_04604", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["ci_integration", "tests_are_truth", "tooling"]}}
+{"id": "train_04605", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04606", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_04607", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04608", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "documentation", "reproducibility"]}}
+{"id": "train_04609", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "reproducibility"]}}
+{"id": "train_04610", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tooling"]}}
+{"id": "train_04611", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_04612", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "governance", "tests_are_truth"]}}
+{"id": "train_04613", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04614", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "tests_are_truth", "documentation"]}}
+{"id": "train_04615", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04616", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04617", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_04618", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "documentation", "security_gates"]}}
+{"id": "train_04619", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_04620", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04621", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "documentation"]}}
+{"id": "train_04622", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04623", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04624", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "tooling", "documentation"]}}
+{"id": "train_04625", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04626", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "governance", "evaluation_metrics"]}}
+{"id": "train_04627", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04628", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "reproducibility", "ci_integration"]}}
+{"id": "train_04629", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "ci_integration"]}}
+{"id": "train_04630", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "ci_integration"]}}
+{"id": "train_04631", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tooling", "evaluation_metrics"]}}
+{"id": "train_04632", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04633", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_04634", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "documentation"]}}
+{"id": "train_04635", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "documentation", "governance"]}}
+{"id": "train_04636", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04637", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04638", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_04639", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_04640", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_04641", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_04642", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "documentation"]}}
+{"id": "train_04643", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tests_are_truth", "tooling"]}}
+{"id": "train_04644", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "governance", "evaluation_metrics"]}}
+{"id": "train_04645", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04646", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04647", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04648", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04649", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "security_gates"]}}
+{"id": "train_04650", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04651", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04652", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "documentation", "reproducibility"]}}
+{"id": "train_04653", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "governance"]}}
+{"id": "train_04654", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "evaluation_metrics", "tooling"]}}
+{"id": "train_04655", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_04656", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04657", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04658", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["tests_are_truth", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04659", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04660", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_04661", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04662", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "evaluation_metrics", "governance"]}}
+{"id": "train_04663", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04664", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04665", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "tooling", "ci_integration"]}}
+{"id": "train_04666", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "tooling", "security_gates"]}}
+{"id": "train_04667", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04668", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "repo_scale_reasoning", "documentation"]}}
+{"id": "train_04669", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04670", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_04671", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_04672", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04673", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_04674", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04675", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "governance", "ci_integration"]}}
+{"id": "train_04676", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04677", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tests_are_truth", "documentation"]}}
+{"id": "train_04678", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_04679", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04680", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04681", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "tooling", "reproducibility"]}}
+{"id": "train_04682", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_04683", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_04684", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04685", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_04686", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_04687", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04688", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04689", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04690", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04691", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04692", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "documentation", "security_gates"]}}
+{"id": "train_04693", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "governance"]}}
+{"id": "train_04694", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_04695", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "ci_integration"]}}
+{"id": "train_04696", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04697", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "security_gates", "tests_are_truth"]}}
+{"id": "train_04698", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "governance", "security_gates"]}}
+{"id": "train_04699", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "reproducibility", "security_gates"]}}
+{"id": "train_04700", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "tests_are_truth"]}}
+{"id": "train_04701", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04702", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04703", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04704", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "documentation"]}}
+{"id": "train_04705", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04706", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04707", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "security_gates"]}}
+{"id": "train_04708", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04709", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04710", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04711", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_04712", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_04713", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04714", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04715", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["documentation", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04716", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04717", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_04718", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tooling", "governance", "documentation"]}}
+{"id": "train_04719", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04720", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_04721", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04722", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04723", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "security_gates", "ci_integration"]}}
+{"id": "train_04724", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_04725", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "governance"]}}
+{"id": "train_04726", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tooling", "reproducibility"]}}
+{"id": "train_04727", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "tests_are_truth", "governance"]}}
+{"id": "train_04728", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04729", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "tooling", "security_gates"]}}
+{"id": "train_04730", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04731", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04732", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_04733", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04734", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04735", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04736", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "ci_integration", "reproducibility"]}}
+{"id": "train_04737", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04738", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04739", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04740", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04741", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "governance", "documentation"]}}
+{"id": "train_04742", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "security_gates", "evaluation_metrics"]}}
+{"id": "train_04743", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04744", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "governance", "security_gates"]}}
+{"id": "train_04745", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04746", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04747", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "tooling"]}}
+{"id": "train_04748", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04749", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "governance", "evaluation_metrics"]}}
+{"id": "train_04750", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_04751", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04752", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "evaluation_metrics", "tooling"]}}
+{"id": "train_04753", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "governance"]}}
+{"id": "train_04754", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "governance", "ci_integration"]}}
+{"id": "train_04755", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "ci_integration", "governance"]}}
+{"id": "train_04756", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_04757", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04758", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04759", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "reproducibility", "cost_latency_tradeoffs"]}}
+{"id": "train_04760", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "security_gates", "documentation"]}}
+{"id": "train_04761", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04762", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04763", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "documentation", "reproducibility"]}}
+{"id": "train_04764", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "security_gates", "tests_are_truth"]}}
+{"id": "train_04765", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04766", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "documentation", "ci_integration"]}}
+{"id": "train_04767", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "tooling", "tests_are_truth"]}}
+{"id": "train_04768", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "security_gates", "governance"]}}
+{"id": "train_04769", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04770", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_04771", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "security_gates"]}}
+{"id": "train_04772", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04773", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04774", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_04775", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "cost_latency_tradeoffs", "repo_scale_reasoning"]}}
+{"id": "train_04776", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04777", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "tooling", "documentation"]}}
+{"id": "train_04778", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "tooling", "reproducibility"]}}
+{"id": "train_04779", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04780", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04781", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "security_gates", "tooling"]}}
+{"id": "train_04782", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04783", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04784", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "tests_are_truth", "tooling"]}}
+{"id": "train_04785", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_04786", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "governance", "security_gates"]}}
+{"id": "train_04787", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_04788", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04789", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04790", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["security_gates", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04791", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04792", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "tests_are_truth", "documentation"]}}
+{"id": "train_04793", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tooling", "security_gates"]}}
+{"id": "train_04794", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04795", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "ci_integration", "cost_latency_tradeoffs"]}}
+{"id": "train_04796", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04797", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04798", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_04799", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04800", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04801", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04802", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "ci_integration", "documentation"]}}
+{"id": "train_04803", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04804", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["repo_scale_reasoning", "security_gates", "tests_are_truth"]}}
+{"id": "train_04805", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04806", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04807", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04808", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04809", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04810", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_04811", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "governance", "evaluation_metrics"]}}
+{"id": "train_04812", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04813", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04814", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tooling"]}}
+{"id": "train_04815", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["governance", "reproducibility", "ci_integration"]}}
+{"id": "train_04816", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "tooling", "governance"]}}
+{"id": "train_04817", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "evaluation_metrics", "tooling"]}}
+{"id": "train_04818", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "governance", "repo_scale_reasoning"]}}
+{"id": "train_04819", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: Go\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_04820", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04821", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04822", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_04823", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "documentation", "governance"]}}
+{"id": "train_04824", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04825", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "governance", "evaluation_metrics"]}}
+{"id": "train_04826", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_04827", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_04828", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04829", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04830", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04831", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04832", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "ci_integration"]}}
+{"id": "train_04833", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "documentation", "tests_are_truth"]}}
+{"id": "train_04834", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04835", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "reproducibility", "documentation"]}}
+{"id": "train_04836", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04837", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04838", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "documentation", "security_gates"]}}
+{"id": "train_04839", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04840", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["repo_scale_reasoning", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04841", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "ci_integration"]}}
+{"id": "train_04842", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "documentation", "security_gates"]}}
+{"id": "train_04843", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_04844", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_04845", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["reproducibility", "tooling", "tests_are_truth"]}}
+{"id": "train_04846", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04847", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "documentation", "ci_integration"]}}
+{"id": "train_04848", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04849", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04850", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "security_gates"]}}
+{"id": "train_04851", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["repo_scale_reasoning", "documentation", "security_gates"]}}
+{"id": "train_04852", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04853", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "documentation", "reproducibility"]}}
+{"id": "train_04854", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: expert\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "documentation"]}}
+{"id": "train_04855", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04856", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "tooling", "security_gates"]}}
+{"id": "train_04857", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04858", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "reproducibility", "repo_scale_reasoning"]}}
+{"id": "train_04859", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["cost_latency_tradeoffs", "governance", "reproducibility"]}}
+{"id": "train_04860", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04861", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_04862", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "governance"]}}
+{"id": "train_04863", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "documentation", "reproducibility"]}}
+{"id": "train_04864", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: JavaScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_04865", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04866", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["evaluation_metrics", "documentation", "ci_integration"]}}
+{"id": "train_04867", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_04868", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04869", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["governance", "evaluation_metrics", "tooling"]}}
+{"id": "train_04870", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04871", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "C#", "developer_needs": ["cost_latency_tradeoffs", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04872", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04873", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "documentation", "governance"]}}
+{"id": "train_04874", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "tooling", "reproducibility"]}}
+{"id": "train_04875", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "evaluation_metrics", "governance"]}}
+{"id": "train_04876", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "documentation", "tooling"]}}
+{"id": "train_04877", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "governance", "evaluation_metrics"]}}
+{"id": "train_04878", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "expert", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04879", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "governance"]}}
+{"id": "train_04880", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tests_are_truth", "security_gates"]}}
+{"id": "train_04881", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: Rust\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04882", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: expert\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04883", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04884", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "expert", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "tests_are_truth", "repo_scale_reasoning"]}}
+{"id": "train_04885", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "evaluation_metrics", "documentation"]}}
+{"id": "train_04886", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "ci_integration", "governance"]}}
+{"id": "train_04887", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "ci_integration"]}}
+{"id": "train_04888", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "governance", "reproducibility"]}}
+{"id": "train_04889", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04890", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "governance", "tooling"]}}
+{"id": "train_04891", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04892", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "ci_integration"]}}
+{"id": "train_04893", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "governance", "security_gates"]}}
+{"id": "train_04894", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "evaluation_metrics", "reproducibility"]}}
+{"id": "train_04895", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04896", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04897", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "repo_scale_reasoning", "tests_are_truth"]}}
+{"id": "train_04898", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04899", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "documentation"]}}
+{"id": "train_04900", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "security_gates", "repo_scale_reasoning"]}}
+{"id": "train_04901", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "expert", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: expert\nTarget language: Python\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "governance", "tests_are_truth"]}}
+{"id": "train_04902", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "documentation"]}}
+{"id": "train_04903", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "reproducibility", "documentation"]}}
+{"id": "train_04904", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["evaluation_metrics", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04905", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "cost_latency_tradeoffs"]}}
+{"id": "train_04906", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["evaluation_metrics", "security_gates", "governance"]}}
+{"id": "train_04907", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["documentation", "ci_integration", "evaluation_metrics"]}}
+{"id": "train_04908", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["security_gates", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04909", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "documentation", "tests_are_truth"]}}
+{"id": "train_04910", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: expert\nTarget language: Go\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Go", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04911", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04912", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "reproducibility", "tests_are_truth"]}}
+{"id": "train_04913", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "ci_integration", "documentation"]}}
+{"id": "train_04914", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["cost_latency_tradeoffs", "tests_are_truth", "documentation"]}}
+{"id": "train_04915", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04916", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "agent_loop", "difficulty": "advanced", "instruction": "Task: agent_loop\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: advanced\nTarget language: JavaScript\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["reproducibility", "evaluation_metrics", "tooling"]}}
+{"id": "train_04917", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "repo_scale_reasoning", "cost_latency_tradeoffs"]}}
+{"id": "train_04918", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tests_are_truth", "ci_integration", "reproducibility"]}}
+{"id": "train_04919", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04920", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "ci_integration", "tooling"]}}
+{"id": "train_04921", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["tooling", "ci_integration", "reproducibility"]}}
+{"id": "train_04922", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04923", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "SQL", "developer_needs": ["ci_integration", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04924", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Go\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "documentation", "tests_are_truth"]}}
+{"id": "train_04925", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: advanced\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "tests_are_truth", "documentation"]}}
+{"id": "train_04926", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["documentation", "governance", "ci_integration"]}}
+{"id": "train_04927", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["evaluation_metrics", "tooling", "security_gates"]}}
+{"id": "train_04928", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "documentation", "tests_are_truth"]}}
+{"id": "train_04929", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "security_gates", "tests_are_truth"]}}
+{"id": "train_04930", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Bash", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_04931", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "ci_integration", "tooling"]}}
+{"id": "train_04932", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["ci_integration", "security_gates", "reproducibility"]}}
+{"id": "train_04933", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["documentation", "tooling", "tests_are_truth"]}}
+{"id": "train_04934", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tests_are_truth", "security_gates", "ci_integration"]}}
+{"id": "train_04935", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["repo_scale_reasoning", "reproducibility", "governance"]}}
+{"id": "train_04936", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: Python\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "tests_are_truth", "governance"]}}
+{"id": "train_04937", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Python", "developer_needs": ["ci_integration", "reproducibility", "evaluation_metrics"]}}
+{"id": "train_04938", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: expert\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "reproducibility", "governance"]}}
+{"id": "train_04939", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "eval", "difficulty": "expert", "instruction": "Task: eval\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "tests_are_truth"]}}
+{"id": "train_04940", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: intermediate\nTarget language: Java\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_04941", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04942", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["documentation", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04943", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "tooling", "evaluation_metrics"]}}
+{"id": "train_04944", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "tooling", "repo_scale_reasoning"]}}
+{"id": "train_04945", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04946", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "advanced", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReference scaffold:\n```python\ndef agent_loop(plan, edit, test, issue, max_iters=3):\n    history = []\n    p = plan(issue)\n    for _ in range(max_iters):\n        patch = edit(issue, p)\n        ok, report = test(patch)\n        history.append({\"plan\": p, \"passed\": ok, \"report\": report[:200]})\n        if ok:\n            return patch, history\n        p = p + \" | refine from failures\"\n    return patch, history\n\n```\nOperational notes: sandbox, pinned deps, human gate.", "metadata": {"target_language": "Python", "developer_needs": ["reproducibility", "tooling", "cost_latency_tradeoffs"]}}
+{"id": "train_04947", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Bash\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Bash", "developer_needs": ["tests_are_truth", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04948", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Rust\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "tooling"]}}
+{"id": "train_04949", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Secure code generation and policy gates\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["tooling", "evaluation_metrics", "security_gates"]}}
+{"id": "train_04950", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "compare", "difficulty": "advanced", "instruction": "Task: compare\nTopic: Governance, provenance, and licensing for code data\nDifficulty: advanced\nTarget language: TypeScript\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tooling", "governance"]}}
+{"id": "train_04951", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: JavaScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["security_gates", "evaluation_metrics", "documentation"]}}
+{"id": "train_04952", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: intermediate\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04953", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["governance", "cost_latency_tradeoffs", "evaluation_metrics"]}}
+{"id": "train_04954", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "expert", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: expert\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["ci_integration", "cost_latency_tradeoffs", "tooling"]}}
+{"id": "train_04955", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "data_pipeline", "difficulty": "intermediate", "instruction": "Task: data_pipeline\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: Go\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Go", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "governance"]}}
+{"id": "train_04956", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "review", "difficulty": "advanced", "instruction": "Task: review\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: advanced\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["evaluation_metrics", "tests_are_truth", "tooling"]}}
+{"id": "train_04957", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "design", "difficulty": "advanced", "instruction": "Task: design\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04958", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "design", "difficulty": "foundation", "instruction": "Task: design\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: Go\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Go", "developer_needs": ["ci_integration", "repo_scale_reasoning", "reproducibility"]}}
+{"id": "train_04959", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "eval", "difficulty": "advanced", "instruction": "Task: eval\nTopic: Secure code generation and policy gates\nDifficulty: advanced\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "tooling", "ci_integration"]}}
+{"id": "train_04960", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: expert\nTarget language: Bash\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["governance", "tests_are_truth", "reproducibility"]}}
+{"id": "train_04961", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "design", "difficulty": "expert", "instruction": "Task: design\nTopic: SWE-bench style real-repo evaluation\nDifficulty: expert\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["cost_latency_tradeoffs", "ci_integration", "repo_scale_reasoning"]}}
+{"id": "train_04962", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "tests_are_truth", "cost_latency_tradeoffs"]}}
+{"id": "train_04963", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "reproducibility", "governance"]}}
+{"id": "train_04964", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "ci_integration", "reproducibility"]}}
+{"id": "train_04965", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: C#\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["evaluation_metrics", "security_gates", "documentation"]}}
+{"id": "train_04966", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: intermediate\nTarget language: Bash\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["repo_scale_reasoning", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04967", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "expert", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: expert\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Python", "developer_needs": ["evaluation_metrics", "cost_latency_tradeoffs", "tests_are_truth"]}}
+{"id": "train_04968", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "security_gates", "tooling"]}}
+{"id": "train_04969", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Rust\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["governance", "cost_latency_tradeoffs", "ci_integration"]}}
+{"id": "train_04970", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "agent_loop", "difficulty": "intermediate", "instruction": "Task: agent_loop\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: SQL\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "tests_are_truth"]}}
+{"id": "train_04971", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "ci_integration", "tooling"]}}
+{"id": "train_04972", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "expert", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: expert\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["documentation", "security_gates", "reproducibility"]}}
+{"id": "train_04973", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: SWE-bench style real-repo evaluation\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "reproducibility", "tooling"]}}
+{"id": "train_04974", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "tooling", "governance"]}}
+{"id": "train_04975", "created": "2026-01-01", "topic": "Tool calling, sandboxes, and CI integration", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Tool calling, sandboxes, and CI integration\nDifficulty: foundation\nTarget language: Java\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "reproducibility", "governance"]}}
+{"id": "train_04976", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: SWE-bench style real-repo evaluation\nDifficulty: advanced\nTarget language: Python\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Python", "developer_needs": ["security_gates", "cost_latency_tradeoffs", "reproducibility"]}}
+{"id": "train_04977", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "review", "difficulty": "foundation", "instruction": "Task: review\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["reproducibility", "tooling", "security_gates"]}}
+{"id": "train_04978", "created": "2026-01-01", "topic": "SWE-bench style real-repo evaluation", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: SWE-bench style real-repo evaluation\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["governance", "ci_integration", "security_gates"]}}
+{"id": "train_04979", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Java\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["reproducibility", "repo_scale_reasoning", "governance"]}}
+{"id": "train_04980", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "compare", "difficulty": "intermediate", "instruction": "Task: compare\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: SQL\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "SQL", "developer_needs": ["reproducibility", "governance", "documentation"]}}
+{"id": "train_04981", "created": "2026-01-01", "topic": "Reasoning-first coding models and tunable deliberation", "task_type": "compare", "difficulty": "foundation", "instruction": "Task: compare\nTopic: Reasoning-first coding models and tunable deliberation\nDifficulty: foundation\nTarget language: Bash\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nCompare: capability, cost, latency, reliability, governance\n", "metadata": {"target_language": "Bash", "developer_needs": ["reproducibility", "evaluation_metrics", "ci_integration"]}}
+{"id": "train_04982", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "eval", "difficulty": "foundation", "instruction": "Task: eval\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Rust", "developer_needs": ["documentation", "repo_scale_reasoning", "tooling"]}}
+{"id": "train_04983", "created": "2026-01-01", "topic": "Model merging, distillation, and continued pretraining", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Model merging, distillation, and continued pretraining\nDifficulty: foundation\nTarget language: Rust\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["security_gates", "documentation", "cost_latency_tradeoffs"]}}
+{"id": "train_04984", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Governance, provenance, and licensing for code data\nDifficulty: foundation\nTarget language: C#\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["governance", "security_gates", "cost_latency_tradeoffs"]}}
+{"id": "train_04985", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "review", "difficulty": "intermediate", "instruction": "Task: review\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nReview: correctness, security, performance, governance\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "evaluation_metrics", "repo_scale_reasoning"]}}
+{"id": "train_04986", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: intermediate\nTarget language: C#\nContext: Design a data pipeline for continued pretraining with auditability.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "C#", "developer_needs": ["reproducibility", "documentation", "evaluation_metrics"]}}
+{"id": "train_04987", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "foundation", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: foundation\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Bash", "developer_needs": ["tooling", "governance", "cost_latency_tradeoffs"]}}
+{"id": "train_04988", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: intermediate\nTarget language: SQL\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "SQL", "developer_needs": ["cost_latency_tradeoffs", "security_gates", "reproducibility"]}}
+{"id": "train_04989", "created": "2026-01-01", "topic": "Extended context and repo-scale understanding", "task_type": "data_pipeline", "difficulty": "advanced", "instruction": "Task: data_pipeline\nTopic: Extended context and repo-scale understanding\nDifficulty: advanced\nTarget language: Java\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Java", "developer_needs": ["tests_are_truth", "repo_scale_reasoning", "evaluation_metrics"]}}
+{"id": "train_04990", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Rust", "developer_needs": ["tooling", "evaluation_metrics", "tests_are_truth"]}}
+{"id": "train_04991", "created": "2026-01-01", "topic": "Governance, provenance, and licensing for code data", "task_type": "code", "difficulty": "intermediate", "instruction": "Task: code\nTopic: Governance, provenance, and licensing for code data\nDifficulty: intermediate\nTarget language: C#\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["tooling", "security_gates", "governance"]}}
+{"id": "train_04992", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "code", "difficulty": "foundation", "instruction": "Task: code\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: foundation\nTarget language: TypeScript\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["security_gates", "repo_scale_reasoning", "ci_integration"]}}
+{"id": "train_04993", "created": "2026-01-01", "topic": "Multimodal dev workflows (docs, diagrams, traces)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Multimodal dev workflows (docs, diagrams, traces)\nDifficulty: advanced\nTarget language: Java\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "Java", "developer_needs": ["repo_scale_reasoning", "tests_are_truth", "tooling"]}}
+{"id": "train_04994", "created": "2026-01-01", "topic": "Secure code generation and policy gates", "task_type": "design", "difficulty": "intermediate", "instruction": "Task: design\nTopic: Secure code generation and policy gates\nDifficulty: intermediate\nTarget language: C#\nContext: Fix a failing issue with tests as the oracle and produce a safe patch.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "C#", "developer_needs": ["ci_integration", "tests_are_truth", "evaluation_metrics"]}}
+{"id": "train_04995", "created": "2026-01-01", "topic": "Mixture-of-Experts (MoE) for code", "task_type": "explain", "difficulty": "intermediate", "instruction": "Task: explain\nTopic: Mixture-of-Experts (MoE) for code\nDifficulty: intermediate\nTarget language: TypeScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "TypeScript", "developer_needs": ["documentation", "ci_integration", "security_gates"]}}
+{"id": "train_04996", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "data_pipeline", "difficulty": "foundation", "instruction": "Task: data_pipeline\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: Rust\nContext: Create an eval harness that reflects real developer workflows.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nPipeline:\n1) Ingest\n2) Normalize\n3) Filter\n4) Dedupe\n5) Quality score\n6) Sample\n7) Audit\n", "metadata": {"target_language": "Rust", "developer_needs": ["tests_are_truth", "evaluation_metrics", "governance"]}}
+{"id": "train_04997", "created": "2026-01-01", "topic": "Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)", "task_type": "explain", "difficulty": "advanced", "instruction": "Task: explain\nTopic: Agentic coding systems (plan\u2192edit\u2192test\u2192reflect)\nDifficulty: advanced\nTarget language: JavaScript\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nDesign guidance with risks, metrics, acceptance criteria\n", "metadata": {"target_language": "JavaScript", "developer_needs": ["cost_latency_tradeoffs", "governance", "tooling"]}}
+{"id": "train_04998", "created": "2026-01-01", "topic": "Code-specialized model families and sizing tradeoffs", "task_type": "eval", "difficulty": "intermediate", "instruction": "Task: eval\nTopic: Code-specialized model families and sizing tradeoffs\nDifficulty: intermediate\nTarget language: Bash\nContext: Integrate an LLM agent into CI for a large monorepo.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nEval:\n- Tasks: real issues\n- Metrics: pass@k, time-to-green\n- Gates: lint/security\n", "metadata": {"target_language": "Bash", "developer_needs": ["evaluation_metrics", "documentation", "tooling"]}}
+{"id": "train_04999", "created": "2026-01-01", "topic": "Dataset curation pipelines (filter, dedupe, quality)", "task_type": "agent_loop", "difficulty": "foundation", "instruction": "Task: agent_loop\nTopic: Dataset curation pipelines (filter, dedupe, quality)\nDifficulty: foundation\nTarget language: SQL\nContext: Evaluate two coding models for internal rollout under strict governance.\nDeliver production-grade guidance or artifacts.", "input": "", "output": "Key facts:\n- Modern AI coding emphasizes correctness via tests, agentic loops, and real-repo evaluation.\n- Reasoning-first and MoE approaches improve capability-per-compute when paired with tools.\n\nAgent loop: Plan \u2192 Edit (diff) \u2192 Test \u2192 Reflect \u2192 Human gate\n", "metadata": {"target_language": "SQL", "developer_needs": ["security_gates", "tests_are_truth", "reproducibility"]}}