Taste cannot be prompted
You can generate a hundred product directions in a second. Volume was never the problem - knowing which direction carries real weight always was.
The abundance problem
A generation ago the constraint was output. Ideas took time to develop, decks took days to build, user research took weeks to conduct. Scarcity forced a kind of accidental rigor - you couldn't pursue every direction, so you had to choose.
That constraint is gone. A founder today can generate fifty product concepts before breakfast, complete with positioning statements, feature lists, and competitive moats. The volume is genuine. The problem is that volume and quality have always been different things, and removing the barrier to one does nothing for the other.
What taste is
Taste isn't aesthetic preference. In a product context, taste is the accumulated ability to feel where a decision is structurally wrong before you can prove it - to sense that a particular mechanic will collapse under real human behavior even when it looks sound in a brief. It comes from close, repeated contact with the moment a product meets a person. It cannot be inherited from a dataset, because datasets record what happened, not why it felt the way it did.
The substitution error
The current substitution - volume of output for depth of judgment - produces a particular failure mode: products that are architecturally sound and psychologically hollow. Every user flow is logical. The onboarding converts. The feature set is complete. And six months after launch, nobody cares about it. Not because it doesn't work, but because it never acquired gravity. It was built from pattern, not from understanding.
Where the gap opens
Between an automated product direction and a product that people return to, there is a gap filled entirely by human judgment. Someone has to decide which of the fifty directions carries a genuine insight about human behavior - an anxiety the user has but hasn't named, a trade-off they're making that they'd abandon if someone gave them a better option, a ritual that already exists in their life that the product can anchor to.
That judgment is not a bottleneck to be removed. It's the work.
What to do with it
Stop treating the generation of options as progress. It is preparation, at best. The work begins at the moment of choosing - and that choice deserves more time and structural rigor than any prompt will ever require.