An AI feature is a capability added to a product that works without it. An AI-first product is one where, take the AI away, there is no product at all. Knowing which you are building changes the scope, the funding, the risk, and the moat.
An AI product MVP has a different job from a normal MVP. A normal MVP tests whether people want the thing. An AI MVP also has to test whether the AI can actually do the thing reliably — and those are two separate risks.
For an AI product, the evaluation suite is not QA you add at the end. It is the product specification, the regression safety net, and the only honest measure of whether the thing works. For an AI-first product, the evals come first.