Notes on building software with coding agents. Mostly about the boring parts that decide whether any of it survives contact with a real codebase: fail-closed gates, test harnesses, review loops, and the failure modes that only show up at scale.

No posts yet.