Keeping AGENTS.md and CLAUDE.md small
How I keep agent instruction files short, readable, and focused by moving changing project knowledge into Memory.
Use cases
Where durable context helps AI coding agents.
Practical notes on keeping product intent, repo conventions, and sharp edges available across agent sessions.
Talking to ChatGPT about a project without handing it the repo
How I use project memory to discuss architecture and planning without pasting the whole repository into a chat.
Watching multiple agent-built projects with Memory
How I use the Memory viewer to keep track of several agent-built projects without reading every chat transcript.