Almost certainly, your app doesn't need rebuilding. What it needs is for someone to work out which specific things are broken — usually a handful — and fix those. Rewriting is the most expensive option available and it is very rarely the one the code actually calls for.
When an AI-built app stops feeling trustworthy, it's rarely one fault. It's typically a real bug, plus something environmental that only shows up in production, plus a structural issue that makes both harder to find.
They get experienced as one big problem because they arrived together. Separating them is most of the work, and it's the part that makes the rest tractable — each of the three usually has a different, much smaller fix.
Rewriting feels decisive. Here's what each option actually costs, honestly.
What works, what half-works, and what only appears to work. This takes a day or two and changes the shape of everything after it.
The things stopping you shipping get fixed first. Everything else goes on a list with a priority, not into the current work.
Small, reviewable commits. If something goes wrong you know exactly which change did it — which is not true of a rewrite.
Your repository, your accounts, your deployment. Nothing depends on me being available next month.
Usually not. Rewriting is the most expensive option and it’s rarely the one the code actually calls for. I’d rather find the specific things that are broken and fix those.
Yes. I start by understanding the system before changing anything — dependencies, data model, auth, deployment, and the parts that are load-bearing. Then I’ll tell you what I found before I start editing.
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.
The service
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.
A practical method for debugging AI-generated code when prompting has stopped helping: reproduce it, read the seams, check the assumptions, and fix the cause.
Read
8 min read
Nine differences between your machine and production, in the order they break: env vars, build output, ports, database, CORS, paths, time, memory and secrets.
Read
9 min read
Sometimes it needs an engineer.
Send the messy version. That’s usually the useful version.