I built a proentropic memory layer for AI coding agents — every mistake makes the system stronger

Every AI coding agent has the same problem: it makes a mistake, you correct it, and next session it...