What lives in your head doesn't flow across AI sessions, and doesn't always make it to your team. How features fit together in your system, the patterns you follow in an area, the reasoning behind decisions — they're not always captured by CLAUDE.md and .cursorrules. Coding with agents made me more productive, but made me repeat myself. So I built aide-memory — an auto-captured, auto-recalled, path-scoped memory tool for AI coding agents and teams. Why rules files aren't enough on their own CLAUDE.md and .cursorrules help with part of this. But they have real limits: Corrections don't make it back. You correct the agent in conversation. The lesson works for that session. Unless someone manually edits the rules file afterwards, it doesn't persist. Area-specific knowledge isn't always documented. How a feature ties into the larger system, the architectural decisions in a specific module, the reasons certain things are built the way they are.…