rescue
Your app runs. It will not survive real users.
Generated apps fail in the same few places, and which places depends on what generated them. Pick the tool you used.
- a prompt-to-app builder that generates a React front end on Supabase
Row-level security left open
Lovable gives you a working React front end and a Supabase project behind it. The front end is usually the part that is fine. What needs work is everything the database is trusting it to do.
What breaks in Lovable apps → - an in-browser builder that generates and runs a full-stack app in a sandbox
Secrets that were never server-only
Bolt builds and runs the whole app inside the browser, which is what makes it fast to try and awkward to ship. The code assumes an environment that only exists while the tab is open.
What breaks in Bolt apps → - a generator that turns prompts into React and shadcn/ui components
Mock data still standing in for a database
v0 is good at the part you can see. It produces clean React and sensible components, and it stops at the edge of the interface. What is missing is not broken code — it is the half of the application that was never generated.
What breaks in v0 apps → - an editor where an agent writes into your existing codebase
Three patterns for the same job
Cursor does not produce a recognisable house style, because it writes whatever the surrounding prompt suggested that day. The result is a codebase that works and that nobody can predict.
What breaks in Cursor apps →
Built with something else? The $99 audit reads any codebase and the finding list is the same shape whatever wrote it.
Start with the audit.
$99, back within 48 hours, every finding with a file and a line. The fee credits in full into the fix.
Written and asynchronous. Nobody here will ask you to book a call.