# How do I test an AI feature so it doesn't break?

## The short answer

- Tibo: Write behavioural test scenarios outside the codebase as a holdout set for AI agents

## Playbooks

### [Write behavioural test scenarios outside the codebase as a holdout set for AI agents](https://vibedraft.app/builder-brains/tibo/write-behavioural-test-scenarios-outside)

Tibo · Building · X posts

Because an agent can 'teach to the test' if it sees unit tests, store end-to-end user-behaviour specs separately and only use them to judge whether the software actually works for a real user.

**When it applies:** AI-native development at level 3+ (agent writes most code)

Source: [Tibo: AI made 99.9% of devs slower (X long post)](https://x.com/tibo_maker/status/2026276723872936386), 24 February 2026 on x.com

## Founders who talk about it

- [Tibo](https://vibedraft.app/builder-brains/tibo)

## Related topics

- [How do I set up agents that do real work?](https://vibedraft.app/builder-brains/topics/agent-workflows)
- [What can I automate with AI as a solo founder?](https://vibedraft.app/builder-brains/topics/ai-automation)
- [How do I keep AI costs from eating my margin?](https://vibedraft.app/builder-brains/topics/ai-costs)
- [Will AI make my startup pointless?](https://vibedraft.app/builder-brains/topics/ai-platform-risk)
- [How do I build an AI product that isn't just a wrapper?](https://vibedraft.app/builder-brains/topics/ai-products)
- [How do I ship faster with Claude Code and Cursor?](https://vibedraft.app/builder-brains/topics/coding-agents)

Updated 2026-11-07. [Web version](https://vibedraft.app/builder-brains/topics/evals)

Advice cards are rewritten summaries with a link to the original. They are not the founder's words unless shown in quotation marks.
