// track · 9 lessons
When it breaks, triage before you rebuild
The rewrite feels faster and almost never is. Diagnose the one line first.
- 01When it breaks, triage before you rebuildThe rewrite feels faster and almost never is. Diagnose the one line before you throw away the whole thing.
- 02The rewrite reflex, and why regeneration feels like progressWhen AI code breaks, the urge is to regenerate the whole thing. It feels like progress and usually isn't. Diagnose before you delete.
- 03Diagnose the one line before you throw away the whole thingMost breaks are one wrong line in a system that's otherwise fine. Find that line first. The rewrite that skips diagnosis almost never pays off.
- 04Surface fix versus semantic fixMaking the symptom disappear is not the same as fixing the cause. Ask what made the bug possible, or you'll patch the same failure forever.
- 05The rescue-spec, or specifying a system before you touch itBefore fixing a broken system you didn't fully build, write down what it should do. The rescue-spec turns a mystery into something you can verify.
- 06Reading AI-built code you didn't writeOrchestrating AI means inheriting code you didn't type. Learn to read for intent and invariants, not line by line, or the drift owns you.
- 07When rebuilding actually is the right callSometimes the rewrite is correct. The test is simple: rebuild on a diagnosis, never on a mood. If you can't say what's wrong, you're not ready.
- 08The glitch class, or when the tool fails mid-generationSome failures aren't your code, they're the tool glitching mid-output. Learn to spot this class and switch to delegation instead of fighting it.
- 09Containment, or halting a broken agent before it spreadsWhen an agent starts going wrong, first move is stop, not steer. Halt it, assess, then decide. A running broken agent digs the hole deeper.