Thoughts on agent memory, context engineering, and building tools that remember.
Coding agents are powerful but forgetful. They don't reliably remember why things were built a certain way or apply corrections across sessions. Here's what's missing and how to think about fixing it.
