In practice
Its purpose is to answer a question you can't answer another way. “Minimum” and “viable” pull against each other: too small and the test proves nothing; too polished and it stops being cheap.
An MVP isn't simply version 1 of the full product. It also isn't a prototype, which usually never reaches real customers.
Why it matters for product builders
The real cost is the time between asking the question and getting evidence.
What changes in AI-first products
MVPs got smaller and faster. The new risk is overbuilding because it's easy, then forgetting what the test was meant to learn.