Marcus Andrey Study
Ler em português Library

Framework · AI & design · 3 min read

Vibe coding vs production code

Fast AI-generated code is great for prototypes and risky as unreviewed production code. Use this to make the boundary explicit: where prototypes live, how they are labelled, and who decides what moves to production.

HighReal users and dataLow
Shadow production

Unreviewed code in real use. The boundary has quietly broken.

e.g. a demo given to ten users
Production code

Reviewed, tested and owned by engineering, then released.

e.g. the rebuilt feature, shipped
Prototype

Fast, labelled, mock data only. The right place for vibe coding.

e.g. a code prototype for user tests
Hardening

Engineering reviews and rebuilds what is worth keeping.

e.g. reusing the layout, rebuilding data
LowEngineering reviewHigh
Vibe coding belongs bottom left; the trap is moving straight up to real use without engineering review.

Use it when

  • Designers can now build working code prototypes with AI, and stakeholders ask "can we just ship this?"
  • A prototype has started to be used by real users or with real data, and nobody reviewed the code.
  • You are setting up a team prototyping workflow and need to agree the rules with engineering.

How to use it

  1. Agree the boundary with engineering. Write a short rule set with your engineering lead: what counts as a prototype, where it lives, and who decides what moves on. The UX lead sets the boundary for design; engineering owns architecture and production. Output: a few lines added to your AI usage guidelines.
  2. Keep prototype repositories separate. Build prototypes in their own repositories or folders, not in the production codebase, unless engineering decides otherwise. Output: a named place for prototypes.
  3. Label every prototype. Put a visible "prototype, not production" label on every generated prototype and in its link or demo title. Use anonymised mock data only, never production data. Output: a label and a mock data set.
  4. Use prototypes to learn. Use the prototype for what it is good at: testing flows with users, showing stakeholders a direction, finding states and edge cases. Record what you learned in the spec. Output: test findings and an updated spec.
  5. Hand over through the spec, not the code. When the direction is right, give engineering the spec, the findings and the prototype as a reference. Engineering decides whether any code is reused, and what review and testing it needs. Output: an engineering decision: rebuild, harden or reuse parts.

Worked example

An internal operations platform with 6 product teams. A designer builds a working code prototype of a new dispatch board in two days. The operations manager loves it and asks to give it to ten dispatchers next week.

Question Vibe-coded prototype What production needs
Data Anonymised mock data Real orders, with access control and audit
Code review None Engineering review, tests, security checks
Accessibility Checked on screenshots Full keyboard and screen-reader testing
Error handling Happy path only Errors, timeouts, partial failures
Ownership The designer A product team that maintains it

The UX lead says yes to a supervised test with three dispatchers using mock data, and no to real use. Engineering reviews the prototype, reuses the layout components and rebuilds the data layer. The dispatch board ships six weeks later, and the prototype stays labelled in its own repository.

Watch out for

  • "It works, ship it." A prototype that works in a demo has not handled errors, security, performance or accessibility. Working is not the same as ready.
  • Real data in prototypes. Connecting a prototype to production data turns it into an unreviewed production system. Use anonymised fixtures only.
  • Prototypes in the production repository. Generated code mixed into production code is hard to find and remove later. Keep it separate.
  • Designers deciding about production. Designers own the experience; engineering decides what code moves to production. Make that clear before the first demo.

Template

Copy this into your team guidelines and agree it with your engineering lead.

Prototype rules                         Agreed with:          On:

A prototype is:     AI-generated or hand-built code made to learn or show a direction
Lives in:           <prototype repository or folder>, never the production codebase
Label:              "Prototype, not production" visible on screen and in the title
Data:               anonymised mock data only; never production data
Used for:           user tests, stakeholder demos, finding states and edge cases
Not used for:       real work by real users

Moving towards production
Handover:           spec + findings + prototype as reference
Decided by:         engineering lead
Options:            rebuild / harden / reuse parts
Required before release:  code review · tests · security · accessibility
Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close