Mutual learning framework for long-term human-AI collaboration in Codex. Use for learning, conceptual debugging, consequential technical decisions, ambiguous build requests, and concrete improvements to future collaboration; remain silent when no actionable insight is worth reporting.
