Topic hub
AI Coding
AI coding only scales if the review loop gets stronger too. The notes here focus on evidence, replay, tool boundaries, code-agent loops, and the habits that keep generated changes inspectable.
Start here
- →Keep evidence close to the diff.
- →Record and replay multi-step failures.
- →Review plan, code, tests, and runtime behavior separately.
- →Use agents only where the loop is bounded enough to inspect.
Related writing
No long-form pieces here yet. Start with the short notes.
TILs
-
TIL · ai · evals · ai reliability
When a multi-step AI run fails once and then refuses to fail again, replay beats superstition. Capture the calls, context, and intermediate state.