Claude Code is the fastest coder I've ever worked with. It can scaffold a feature, write tests, and open a PR in minutes. But I kept running into the same problem: the code worked, and then it didn't.
A race condition in a status transition. A hard-coded string that should have been a constant. A transaction that rolled back an audit record it was supposed to keep. Tests that asserted true instead of asserting the right value.
The fixes were always fast too. But each fix came with a side quest: the incident, the regression, the "why didn't we catch this?" retro. The velocity was high. The net velocity, after accounting for the bugs, wasn't.
And even with a decent CLAUDE.md, I was still babysitting every session. Please don't forget TDD this time. Hey, you forgot to check Bug Bot. Can you actually run the tests before opening the PR? Each prompt felt like a conversation with someone extremely talented who also had the short-term memory of a goldfish. The problem wasn't Claude's ability. It was that good habits written down somewhere in a markdown file don't automatically become practiced habits. I was the process. Which meant the process was inconsistent, forgetful, and increasingly annoyed at itself.
