Generative Vision Interview Questions, issue 22, Jun 30, 2026

The DreamBooth Trap

Generative AI Engineer interview at Midjourney, and the interviewer asks:

You DreamBooth a model on 5 photos of a client’s product. It generates that product perfectly, but now everything else it draws looks broken. Why, and how do you fix it without re-collecting data?

Don’t say: Just train on more images of the product

Why naive few-shot personalization quietly steamrolls your model’s prior knowledge, and how to use a frozen base model as its own supervisor to stop catastrophic forgetting.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.