Built with
Every AI builder breaks in its own way
Lovable leaves Row Level Security off. Bolt puts API keys in the browser. v0 ships Server Actions anyone can call. Cursor drifts. Pick your tool for the specific list — and what we do about each item.
- Lovable
Lovable got you a working app. We get it to paying customers
Row Level Security is off, or lets everyone in · Edge functions that trust the caller · Secrets in the browser
Lovable to production - Bolt.new
Bolt built it in the browser. We make it survive outside one
API keys called from the browser · Functions anyone can call · Works in preview, dies on deploy
Bolt.new to production v0 drew the app. We wire it to something real
Server Actions with no auth · Mock data that never left · `"use client"` on everything
v0 by Vercel to production- Cursor
Cursor moves fast. We make sure the codebase still holds
Three architectures in one repo · Tests that can't fail · `any` as a type strategy
Cursor to production - Replit
Replit shipped your first version. We ship the one you can bill for
Routes with no gate · Relational data in a key-value store · Schema by `db push`
Replit to production - Claude Code
Claude Code wrote a lot of good code. Someone should read it
Interfaces for one implementation · Tests that assert the mock · Diffs nobody read
Claude Code to production
Also built with ChatGPT or Windsurf? Same process, same prices — the failure modes overlap with Cursor's, and the audit is tool-agnostic.