I have been talking to people building apps with tools like Lovable, Bolt, Replit, and Cursor, and I keep seeing the same pattern.
They get surprisingly far on the prototype. Then things start breaking when real users, auth, payments, database changes, and edge cases show up.
The most common failure points I keep hearing about are:
auth and database issues,
AI losing track of prior decisions and rewriting working code,
weak error handling,
underestimated scope once real integrations are involved,
and security problems around permissions and user data.
I am curious how developers here think about this. When you inherit an AI-built project, what usually determines whether it is salvageable, needs hardening, or needs a rewrite?
I have been exploring a site called OpenBaton.com around this handoff problem, but I am mainly posting because I want to understand whether other people are seeing the same thing.
kellonedwards•1h ago
They get surprisingly far on the prototype. Then things start breaking when real users, auth, payments, database changes, and edge cases show up.
The most common failure points I keep hearing about are: auth and database issues, AI losing track of prior decisions and rewriting working code, weak error handling, underestimated scope once real integrations are involved, and security problems around permissions and user data.
I am curious how developers here think about this. When you inherit an AI-built project, what usually determines whether it is salvageable, needs hardening, or needs a rewrite?
I have been exploring a site called OpenBaton.com around this handoff problem, but I am mainly posting because I want to understand whether other people are seeing the same thing.