---
title: "Comment by Aamer Mehaisi on \"I Made Every AI Builder Hallucinate With 6 Words\" | Hashnode"
description: "Excellent article. The 6-word test reveals something fundamental about AI builders - tools optimize for wow moments, not for correctness. The pattern connects t"
url: "https://hashnode.com/posts/i-made-every-ai-builder-hallucinate-with-6-words/69c6acf6ae4d2d0c2f449d13/comment/69c6b5754d820c96db453200"
---

[Discussion](https://hashnode.com/posts/i-made-every-ai-builder-hallucinate-with-6-words/69c6acf6ae4d2d0c2f449d13)

Comment [I Made Every AI Builder Hallucinate With 6 Words](https://hashnode.com/posts/i-made-every-ai-builder-hallucinate-with-6-words/69c6acf6ae4d2d0c2f449d13)

[Aamer Mehaisi](https://hashnode.com/@mehaisi)

[Mar 27](https://hashnode.com/posts/i-made-every-ai-builder-hallucinate-with-6-words/69c6acf6ae4d2d0c2f449d13/comment/69c6b5754d820c96db453200)

Excellent article. The 6-word test reveals something fundamental about AI builders - tools optimize for wow moments, not for correctness.

The pattern connects to a deeper issue: AI builders conflate inference with intent. When asked for navbar, hero, skills, they infer a complete portfolio from training patterns. But inference without explicit consent is hallucination dressed as helpfulness.

What Drafte is doing - decisions before code - maps to what works in agent systems: explicit contracts over implicit expectations. The AI should surface decision points, not guess what you want.

The contact section hallucination shows why schemas beat prose - a schema would say navbar required, hero required, skills required, everything else forbidden. The AI cannot add what the schema disallows.

This philosophy - contract first, execution second - is what production agent systems need.