Claude is in My Commit History

My recent commits have a new co-author:

Co-Authored-By: Claude Opus 4.6 (1M context) <[e-mail]>

It started as accurate bookkeeping. Now it's just how I work.

What I'm actually using it for

I build a lot of tooling — test runners, CI pipelines, API integrations. Claude doesn't help me write README files or summarise meetings. It's in the code, catching the things I miss when I'm moving fast.

A few recent examples:

JWT expiry logic. I was generating tokens for a test runner. Claude flagged that my expiry config would silently break long test runs — the kind of failure that only shows up at 2am. Fixed before it shipped.

Firebase key resolution. I wanted the tool to find credentials automatically. Claude suggested the fallback chain: check the repo root first, then ~/.ssh/. Obvious in hindsight. I'd have hardcoded the path.

API endpoint shape differences. Caught a mismatch between dev and staging environments before it became a test failure.

What the pattern looks like

I describe the problem. Claude asks questions — usually good ones. We work it out together. Then I write the code, or we write it together if it's straightforward.

The "AI writes your code" framing is wrong. It's closer to: pair programming with someone who's read everything, never has meetings, and doesn't care about tabs vs spaces.

Why I credit it in commits

Because it's accurate. If a human pair programmer caught the JWT bug, I'd mention them. Same logic applies. It also makes the history honest when I'm reviewing it six months later.

---

The commit history doesn't lie. If Claude is in it, it earned the attribution.