continuous-learning-v2

To turn Gemini Code sessions into reusable knowledge, enable observer hooks to capture tool usage and automatically extract atomic instincts with confidence scoring.

Atomic instincts are captured behavioral patterns from coding sessions. They use confidence scoring to track reliability before evolving into commands, agents, or skills.

To prevent cross-project contamination, use project-scoped instincts that isolate learning data per project, ensuring knowledge from one workspace does not pollute another.

Yes, you can promote project-scoped instincts to a global scope. This allows generalized knowledge to be shared across projects once confidence thresholds are met.

Hooks observe sessions by capturing tool usage data during Gemini Code interactions. This data is then processed to extract and score atomic instincts automatically.

You should evolve captured instincts into skills and commands when their confidence scores indicate high reliability, allowing the system to promote them into actionable components.