What if your AI coding agent could remember every bad idea it already tried? 👇 That’s basically the idea behind ergon-wrong-turn-logger, our favorite new GitHub project. Every time you or an AI takes a wrong turn, you log what happened, what broke, and what you should have done instead. Those failures then live in a database instead of disappearing when the task ends. So months later, you can ask: have we tried this before? And instead of starting from zero, you get the clues from the last failed attempt. As coding agents take on more work, teams will need a persistent memory layer for what failed as well as what worked. It serves as institutional memory for AI agents. Ultimately, we look forward to Claude developing generational trauma from every failed refactor. Link to the project here: