window.CRUX_PROJECTS = [ { name: 'Complide CLI & Claude plugin', summary: 'A local-first action gate with setup, review, drift, learning, and receipt commands for coding agents.', status: 'public', category: 'integration', maturity: 'Working OSS', fit: 'Crux substrate', source: 'main · packages/cli, hooks, agents', evidence: '145 CLI tests; golden gate suite; offline gate path.', next: 'Refresh the README lead and release an RC after repository hygiene checks.', href: 'https://github.com/aayush-complide/cli', }, { name: 'Complide rules engine', summary: 'The tested policy, goals, question-bank, and elicitation controller underneath the CLI and MCP surfaces.', status: 'public', category: 'integration', maturity: 'Working OSS', fit: 'Core substrate', source: 'main · packages/engine', evidence: '697 unit tests and a guarded public export surface.', next: 'Add a package README and document the question-policy API separately from compliance rules.', href: 'https://github.com/aayush-complide/cli/tree/main/packages/engine', }, { name: 'Complide MCP server', summary: 'A local stdio bridge that exposes agent gating and status to Cursor, Windsurf, and other MCP hosts.', status: 'public', category: 'integration', maturity: 'Working OSS', fit: 'Core distribution', source: 'main · packages/mcp', evidence: '10 tests; depends only on the engine.', next: 'Merge cross-client fixtures, add a package README, and publish a tagged preview.', href: 'https://github.com/aayush-complide/cli/tree/main/packages/mcp', }, { name: 'Agents, skills & commands', summary: 'Reusable reviewer, think, bootstrap, drift, setup, and control protocols for host-native agent workflows.', status: 'public', category: 'integration', maturity: 'Working OSS', fit: 'Crux V1', source: 'main · agents, skills, commands', evidence: 'Frontmatter CI, UAT, and runnable plugin examples.', next: 'Lead with the think skill as the reference question-layer integration.', href: 'https://github.com/aayush-complide/cli/tree/main/skills', }, { name: 'Question-quality evaluation', summary: 'A dataset and rubric for measuring whether a question surfaces the decision-changing gap rather than sounding helpful.', status: 'public', category: 'evaluation', maturity: 'Working OSS', fit: 'Crux core', source: 'main · eval/question-quality-*', evidence: 'Synthetic dataset, scored report, and reviewer-context comparison.', next: 'Add a dataset card, JSON schema check, and citation file.', href: 'https://github.com/aayush-complide/cli/tree/main/eval', }, { name: 'Session trajectory & drift', summary: 'Local transcript ingestion, event correlation, and goal-drift reporting for replayable agent sessions.', status: 'public', category: 'evaluation', maturity: 'Working OSS', fit: 'Learning loop', source: 'main · packages/trajectory, cli/drift', evidence: '13 drift command tests and local-only storage.', next: 'Add direct trajectory-package tests and a privacy-focused README.', href: 'https://github.com/aayush-complide/cli/tree/main/packages/trajectory', }, { name: 'Crux public website', summary: 'The audience-specific public surface for builders, decision work, research, projects, and private-preview access.', status: 'public', category: 'product', maturity: 'Beta', fit: 'Crux core', source: 'current · site', evidence: 'Dependency-free Pages build, route checks, responsive manual test.', next: 'Connect the allowlist endpoint and point getthecrux.com at Pages.', href: '../index.html', }, { name: 'Complide standup & domain reviews', summary: 'Multi-persona decision reviews for product, UX, operations, growth, and cross-functional standups.', status: 'next', category: 'product', maturity: 'Working OSS', fit: 'Decision-work surface', source: 'main · packages/standup', evidence: '32 tests and no runtime dependencies.', next: 'Add package metadata and a README, then publish as the first standalone package.', href: 'https://github.com/aayush-complide/cli/tree/main/packages/standup', }, { name: 'Signed decision receipts', summary: 'Ed25519-signed, hash-chained local records proving how an agent action was classified and reviewed.', status: 'next', category: 'integration', maturity: 'Working OSS', fit: 'Decision record', source: 'main · packages/cli/src/receipts.ts', evidence: '12 receipt tests; full command text stays local.', next: 'Publish a focused threat model and key-lifecycle guide.', href: 'https://github.com/aayush-complide/cli/blob/main/packages/cli/src/receipts.ts', }, { name: 'ComplyBench fixtures & eval package', summary: 'Terraform, Kubernetes, and policy scenarios with deterministic checks for consequential infrastructure actions.', status: 'next', category: 'evaluation', maturity: 'Working OSS', fit: 'Adjacent benchmark', source: 'main · data/eval-dataset, packages/evals', evidence: '35 eval tests across ten task families.', next: 'Confirm redistribution rights, add third-party notices, and split a clean dataset release.', href: 'https://github.com/aayush-complide/cli/tree/main/data/eval-dataset', }, { name: 'Agent trajectory harness', summary: 'A headless harness for replaying reviewer and gate sessions against known expectations.', status: 'next', category: 'evaluation', maturity: 'PoC', fit: 'Crux evaluation', source: 'test/agent-trajectory-harness', evidence: 'Documented test runs and reusable fixtures.', next: 'Extract synthetic fixtures and add a deterministic CI entry point.', href: null, }, { name: 'Cross-client compatibility layer', summary: 'Cursor-safe fixtures, host matrices, and compatibility tests for the plugin and MCP surfaces.', status: 'next', category: 'integration', maturity: 'Release candidate', fit: 'Core distribution', source: 'feat/phase3a-oss-launch-compat', evidence: 'A broad compatibility test set on the release-prep branch.', next: 'Rebase the branch, merge the fixtures, and publish the host decision matrix.', href: null, }, { name: 'vLLM GPU AWS infrastructure', summary: 'Cost-controlled EC2 provisioning, idle shutdown, and serving scripts for open-model experiments.', status: 'next', category: 'infrastructure', maturity: 'Working infra', fit: 'Research enabler', source: 'infra/vllm-gpu-aws · research/vllm', evidence: 'Provisioning scripts and system services.', next: 'Tighten IAM and ingress defaults, then publish a cost and teardown guide.', href: null, }, { name: 'Crux question layer', summary: 'Host-native hooks and skills that compare user intent with an agent plan and ask only when the answer changes the work.', status: 'preview', category: 'product', maturity: 'Private preview', fit: 'Crux core', source: 'current prototypes · think and reviewer', evidence: 'Working scaffold plus question-quality evaluations; no field traction claim.', next: 'Run three consented PM or solution-architect workflows and publish aggregate results.', href: '../join/index.html', }, { name: 'Crux Integration SDK', summary: 'A planned typed protocol for questions, answers, decision receipts, and host adapters.', status: 'preview', category: 'integration', maturity: 'Design', fit: 'Future core', source: 'current roadmap', evidence: 'Interface boundary defined; package not yet released.', next: 'Extract the minimum protocol from two real host integrations.', href: null, }, { name: 'Complide elicitation research program', summary: 'A synthesis and preregistered experiment program for ask-when-needed behavior across harness, model, and benchmark axes.', status: 'research', category: 'research', maturity: 'Research', fit: 'Crux roadmap', source: 'research · SYNTHESIS, VALIDITY, PREREGISTRATION', evidence: 'Preregistration, falsification rules, and multiple harness conditions.', next: 'Publish a redacted synthesis and protocols after IP and provenance review.', href: null, }, { name: 'Global Workspace / JLens replication', summary: 'Open-model readout, steering, swap, and ablation experiments for signals related to uncertainty and control.', status: 'research', category: 'research', maturity: 'Research', fit: 'Mechanistic roadmap', source: 'global-workspace-research lineage', evidence: 'Pinned result artifacts, papers, and aggregate figures.', next: 'Release protocols and selected aggregates only after publication and patent review.', href: null, }, { name: 'When-to-ask frontier paper', summary: 'A dose-response and steering study that separates ask behavior from end-to-end task success.', status: 'research', category: 'research', maturity: 'Paper draft', fit: 'Mechanistic roadmap', source: 'global-workspace v2 paper lineage', evidence: 'Paper, figures, calibration scripts, and explicit negative results.', next: 'Complete disclosure review and publish the adverse and null results with the positive ones.', href: null, }, { name: 'DFC crosswalk & benchmark', summary: 'A replication of data-flow-control results plus experiments joining semantic sensing to deterministic enforcement.', status: 'research', category: 'research', maturity: 'PoC', fit: 'Research lineage', source: 'cursor/dfc-crosswalk-integration', evidence: 'Benchmark report, charts, package tests, and policy-language comparisons.', next: 'Carve out the neutral crosswalk and replication from patent-adjacent control work.', href: null, }, { name: 'Training & ablation design', summary: 'Pre-, mid-, and post-training hypotheses for improving elicitation propensity without rewarding indiscriminate asking.', status: 'research', category: 'research', maturity: 'Research design', fit: 'Crux roadmap', source: 'research · TRAINING_DESIGN, conditions', evidence: 'Condition files, planned interventions, and stop criteria.', next: 'Run the smallest discriminating experiment before releasing a training recipe.', href: null, }, { name: 'Negotiation elicitation simulator', summary: 'An offline environment that retargets the question controller from risk states to fit and preference discovery.', status: 'research', category: 'research', maturity: 'PoC', fit: 'Adjacent ask-policy lab', source: 'agent-negotiation-personalization', evidence: 'Deterministic simulation, tests, and a small demo UI.', next: 'Publish as an elicitation-economics lab after removing product-specific coupling.', href: null, }, { name: 'Small-LM personalization stack', summary: 'Probe, steering, and context experiments for per-user behavior control without full fine-tuning.', status: 'research', category: 'research', maturity: 'Research', fit: 'Adjacent roadmap', source: 'small-lm-personalization-stack', evidence: 'Benchmark figures and AxBench-oriented scripts.', next: 'Separate the reproducible research package and define a privacy-safe data model.', href: null, }, { name: 'LatentLens reproduction', summary: 'A reproducibility package for third-party multimodal interpretability claims.', status: 'research', category: 'research', maturity: 'Reproduction', fit: 'Research craft', source: 'latentlens-paper-repro', evidence: 'Logbook, smoke tests, and generated figures.', next: 'Release as a standalone reproduction with clear third-party attribution.', href: null, }, { name: 'Agent safety evaluation harnesses', summary: 'Adapters and scripts for injection, harmful-action, and scheming-oriented external benchmarks.', status: 'research', category: 'evaluation', maturity: 'Research tools', fit: 'Safety adjacent', source: 'research-safety-eval', evidence: 'Script suite and common research principles.', next: 'Publish methodology after checking benchmark licenses and service terms.', href: null, }, { name: 'Agent manipulation experiments', summary: 'Probe-and-steer studies testing whether manipulation susceptibility is readable and causally reducible.', status: 'research', category: 'research', maturity: 'Research', fit: 'Safety adjacent', source: 'ai-agent-market-research · agent-manipulation', evidence: 'Internal result set and demo application.', next: 'Separate the methods from competitive research and publish only audited aggregates.', href: null, }, { name: 'Complide Workshop', summary: 'A forward-deployed enterprise POC joining Socratic intake, architecture, manifests, controls, FinOps, and exports.', status: 'archive', category: 'product', maturity: 'Historical PoC', fit: 'Pattern library', source: 'workshop branch stack · apps/web', evidence: 'Demo deck, many package tests, and a complete facilitated workflow.', next: 'Reuse the PM and solution-architect interaction patterns; do not ship the full console as Crux V1.', href: null, }, { name: 'CUJL, manifest & evidence pack', summary: 'A declarative journey language and versioned manifest/export system developed inside the workshop lineage.', status: 'archive', category: 'product', maturity: 'Historical PoC', fit: 'Reusable patterns', source: 'workshop/DFC lineage · packages/cuj, manifest, exports', evidence: 'Specification, parsers, export artifacts, and integration tests.', next: 'Extract only the typed decision and receipt primitives that real Crux integrations need.', href: null, }, { name: 'Crux PWA & native iOS plan', summary: 'A prior decision-feed PWA and native-app migration plan, now frozen in favor of integration-first V1.', status: 'archive', category: 'product', maturity: 'Frozen prototype', fit: 'Historical surface', source: 'crux-pwa-native-ios-plan', evidence: 'PWA prototype, navigation spec, and native design notes.', next: 'Retain as design reference; do not market it as an active first-party product.', href: null, }, { name: 'FounderAGI / IncOS', summary: 'An earlier research-to-revenue and operating-console thesis retained as provenance, not current company direction.', status: 'archive', category: 'product', maturity: 'Retired thesis', fit: 'Historical', source: 'founder-agi-incos strategy lineage', evidence: 'Strategy decks, experiment portfolios, and market studies.', next: 'Archive privately and reuse only verifier-first research principles.', href: null, }, { name: 'Complide enterprise scroll site', summary: 'A cinematic compliance-first landing experiment superseded by the current Crux audience site.', status: 'archive', category: 'product', maturity: 'Superseded prototype', fit: 'Historical marketing', source: 'complide-enterprise-landing', evidence: 'Static scroll engine, generated media, and landing copy.', next: 'Retain the motion system only if it can be reused without stale compliance claims.', href: null, }, { name: 'Compliance market & FEW dossiers', summary: 'Deep market research across regulated software, food, energy, water, finance, and healthcare.', status: 'archive', category: 'research', maturity: 'Internal research', fit: 'Retired wedge', source: 'compliance-research-dossier and DFC lineages', evidence: 'Structured datasets, market maps, and framework crosswalks.', next: 'Keep internal unless a neutral dataset has complete source and redistribution review.', href: null, }, ];