continuous-learning-v2

You can capture Claude Code session observations automatically by enabling the Stop hook, which runs a background observation system to gather knowledge with minimal manual input and ensures continuous growth across sessions.

Atomic instincts with confidence scoring are score-weighted behavioral patterns converted from session observations to guide actions. They enable faster learning and automation by applying score-weighted decision-making across multiple projects.

You promote project-scoped instincts to global patterns using built-in commands that evolve and promote successful instincts. This enables durable learning across multiple repositories by isolating instincts per project and providing a path to global promotion.

Continuous learning works across multiple projects by storing per-project instincts under a local directory and allowing successful patterns to be promoted globally. This isolation and promotion pipeline enables durable learning across multiple repositories.

Yes, you can export and import instincts between different development environments using the system's built-in commands. These commands expose functionality to evolve, export, import, and promote instincts across your various project workflows.