Guides
What to check, by the tool you built with.
Each guide covers what the tool generates, where those apps usually break, how to check it yourself, what fixing it costs and who does it.
Guides by platform
Lovable
Lovable app rescue: what breaks and how to check it
Most Lovable apps that fail in production fail on the Supabase side: tables without access rules, keys in the wrong place, and no backups.SupabaseSupabase RLS audit: how to check an AI-built app's database rules
Supabase is safe when Row Level Security is on and its policies are right.BoltBolt app rescue: moving a Bolt.new project to production
Bolt apps usually need three things before real users: code that runs outside Bolt, secrets moved to the server, and database rules that keep each user's data private.v0v0 app rescue: Next.js apps generated with v0
v0 apps usually break behind the interface: server actions and API routes that never check who the user is, and secrets exposed through NEXT_PUBLIC_ variables.ReplitReplit app rescue: Replit Agent projects in production
Replit apps most often fail on separation: one database for testing and real users, an AI agent with access to production, and no tested way to restore data.CursorCursor and Windsurf apps: when AI-written code needs an engineer
Cursor-built apps most often break where nobody on the team understands the code: endpoints without access checks, no tests to catch what the next change breaks, and packages nobody vetted.Claude CodeClaude Code projects: taking over a codebase an agent wrote
Codebases written with Claude Code usually run fine and fail on what nobody reviewed: secrets committed to Git, commands run against the live database, and large changes merged without tests.When something has already gone wrong
Urgent