Coding agents

How to turn on Codex code review for a GitHub repo: automatic reviews, @codex trigger commands, and pinning review standards into AGENTS.md.

Clearing the wreckage sent me back to the first question: why does memory exist at all? An apple drawing, moments and principles, rules that changed nothing.

Monet began as shared memory. Then the state hypothesis from a thread with Ken, a new Monet built in two weeks, and the day nothing worked at the office.

Nine months on agent memory and almost nothing else, in three parts. Part 1: falling into vibe coding, building memory twice, and the night I decided to dig in.

One security PR passed through 209 review conversations on its way to merge. Ninety percent were accurate. Where the human’s job goes when accuracy is free.

Instead of hand-fixing an AI draft every time, I built a correction loop where one flag becomes a permanent rule. The skill structure, and a seed file.

One paragraph of Tolkien, a million tokens, two hours of autonomous execution, a 3D film. Not a demo review, but what changes about the unit of delegation.

I build an agent memory platform, and this week we reversed its direction. What store-everything memory breaks, and where state, facts, and principles belong.

A CLAUDE.md does not get better as it grows. Why piling up rules degrades your agent, what belongs in the file, and where the rest should live.

I built Monet to fix the agent that forgets yesterday’s conventions. How an agent that writes its own memory and reads back what mattered actually works.