01
AI Code Debugging
The app half-works. You've described the bug to the AI six times, and each fix moves the problem somewhere else. At some point the loop stops being cheaper than reading the code.
What this usually looks like
- A bug that reappears in a different form every time it's "fixed".
- Errors that only happen in production, or only for some users.
- State that goes out of sync and nobody can say why.
- A stack trace pointing at a library, not at your code.
- Something that broke after a change nobody thinks was related.
What I actually do
- 01
Reproduce it
A bug that can't be reproduced can't be confirmed fixed. I start by getting the failure to happen on demand — which on its own often reveals the cause.
- 02
Trace it to the source
I follow the actual data path rather than guessing from the symptom. Generated code tends to fail where two locally-correct pieces meet, and that seam is usually some distance from where the error surfaces.
- 03
Fix the cause, not the symptom
A patch that suppresses the error leaves you with the same bug and less visibility. I fix the thing underneath, and tell you if that means a larger change.
- 04
Verify and explain
I confirm the original reproduction now passes, check the obvious neighbours for the same mistake, and write up what was wrong in language you can act on.
How long it takes
Often within a day, sometimes two. It depends on the requirements — an intermittent bug, or one that only appears in production, takes longer to reproduce than it does to fix. I’ll give you a real estimate once I’ve seen it, not before.
Questions about this
Can you work on code generated by ChatGPT?
Yes, and it’s a large part of what I do. Generated code tends to be locally sensible but globally inconsistent, which is exactly the kind of problem that’s hard to prompt your way out of. I read it the same way I’d read any unfamiliar codebase.
Can you fix production bugs?
Yes. If something is broken for real users right now, say so when you get in touch and I’ll treat it accordingly. I’ll want access to logs and the deployment, not just the code.
What do you need from me?
Less than you’d think. A repository or a deployed URL, some idea of what’s broken, and access to whatever is failing — logs, the hosting dashboard, an error message. If you don’t have all of that, send what you do have and I’ll tell you what else I need.
Your AI-built app doesn’t need another prompt.
Sometimes it needs an engineer.
Send the messy version. That’s usually the useful version.