Six ways a vibe-coded app quietly breaks

AI can build you a working app in a weekend. That is real, and it is genuinely useful. The trouble is that "it works in the demo" and "it holds up in production" are two very different claims, and the gap between them is where vibe-coded apps quietly fall apart.

We see the same six patterns over and over when we audit something that was built fast with AI.

The six ways they break

1. Nobody knows how it works. The app runs, but no human can explain why. The moment something needs changing, you are reverse-engineering your own product.
2. It breaks when you touch it. Change one thing and three unrelated things stop working, because nothing was built to be changed safely.
3. It falls over under real load. Fine for one user in a demo, not fine when real traffic, edge cases, and bad input all show up at once.
4. The doors are wide open. Exposed keys, no real auth, data anyone can reach. AI will happily ship the happy path and skip the locks.
5. It goes blind when it goes down. No logging, no monitoring, no alerts. You find out it broke when a customer tells you, which is the worst possible time.
6. It is held together with duct tape. Hardcoded values, no tests, copy-pasted fixes. It survives right up until the first thing it was not expecting.

None of this means do not build with AI. We use it on everything. It means the build is the easy part, and the unglamorous rest, the part that makes an app safe to depend on, is still real engineering.

That gap is exactly what we check for. Every one of our builds is something a real business actually runs on, which only works because we treat that unglamorous part as the job. If you have an app that works but you are not sure you would bet the business on it, the free Health Check is a good place to start.