Vibe-coded applications can become difficult to maintain when successive prompts change shared logic, introduce duplicate components or leave inconsistent assumptions about data and authentication. A feature may work in isolation but fail when another part of the application is updated.
Before more development, I would establish the last dependable version, identify the workflows that must remain intact and examine how changes affect the rest of the application. This helps distinguish an isolated bug from accumulated technical debt. A sensible stabilisation plan may include selective refactoring, dependency cleanup, version control, targeted tests and safer deployment procedures.
Read more about the warning signs in AI App Technical Debt. For code-level repair and refactoring, see AI Code Cleanup and Repair. Proper investigation is a paid diagnostic; implementation is separately scoped.