Update with project explanation, 100 tests (model 1.0 + 2 full adaptive + 100 loop 100% success), rich VLM traces

Update dataset card for real-browser 3B LAM (no mocks), live visual from Playwright, iterative enhancements.

Update dataset card: note iterative training data generation via build_iterative_data.py and companion model now supports closed-loop iterative inference.

Upload README.md with huggingface_hub

Upload full.jsonl with huggingface_hub

Upload test.jsonl with huggingface_hub

Upload train.jsonl with huggingface_hub

initial commit