How to know if your vibe-coded MVP needs a rewrite (or just a rescue)

Vibe coding is great for speed. Cursor, ChatGPT, and friends can get a founder from idea to clickable demo in a weekend. The hangover usually shows up later: flaky deploys, secret leaks, mystery bugs, and a codebase nobody wants to touch.

The question isn’t “was AI used?”—it’s whether you should rescue what you have or rewrite the risky parts. Here’s a practical way to decide.

Signs you can rescue (stabilize and harden)

Rescue is usually the better bet when:

• Core user flows work most of the time, even if edge cases fail
• You can still run the app locally and understand the main folders
• Auth, payments, or data models are roughly correct—just messy
• The product idea is validated; the engineering is what’s broken

In a rescue, the goal is not perfection. It’s a stable deploy path, fewer landmines, and a clear map of what to keep vs rewrite next.

Signs you should rewrite (selectively or fully)

Lean toward rewrite when:

• Nobody can explain how a critical path works—including the “author”
• Secrets, API keys, or env files were committed or exposed
• Every small change breaks three unrelated features
• There’s no test of any kind and behavior depends on prompt luck
• The stack choices fight your real constraints (hosting, team skills, cost)

Rewriting everything is rarely step one. Often you rewrite the hot path and rescue the rest.

A 60-minute triage checklist

1. Run it cold: fresh clone, install, env setup, start. Note every manual hack required.
2. Walk the money paths: signup, core action, payment or data write. Record failures.
3. Scan for secrets: keys in repo, client-exposed tokens, open admin routes.
4. Check deploy reality: is there CI/CD, or only “works on my laptop”?
5. Ask the ownership test: could a new engineer change this safely in a week?

What a good rescue engagement looks like

At wohnmohr, a Rescue Sprint typically produces:

• A risk map (security, reliability, maintainability)
• Stabilized staging/production path
• Hardening on the worst failure points
• A keep / rewrite / defer plan for the next 2–4 weeks

If you’re unsure where you stand, start with a free AI / vibe-code health check. You’ll get an honest read before you spend another sprint guessing.

Bottom line

Demos prove demand. Production proves craft. If your vibe-coded MVP is mostly sound but fragile, rescue it. If it’s a black box with real risk, rewrite the dangerous parts—on purpose, with a plan.