Building with “AI” for the last three years feels like the Miller’s Planet scene in Interstellar. On that planet, one hour equals seven years on Earth. Three hours cost 23 years.

That is what AI feels like. It’s a paradigm shift of a different level. There is no steady state with continuous momentum and acceleration.

Three years into building with AI, I’ve watched the market get flooded with AI slop that oerpromise and underdeliver. As a result, ROA, return on attention, keeps climbing and the output alone no longer impresses.

People are measuring signal density with the foundational product questions that matter more than ever:

What problem are we solving?
Is it real?
Is it painful?
Is it worth solving now?

If that alignment is fuzzy, velocity does not help. It just amplifies irrelevance. Rapid testing, feedback, iteration are now table stakes. What remains rare is discernment and judgment that often feels like a taste-market fit. PMF is a function of that. The product may be measurably utilitarian, but does it feel right? Is it coherent? With AI, words are cheap and attention is on the decline. One could infer that the more you have to explain your product in words on any given screen, the more likely it is that the alignment is fuzzy. What are your thoughts?

# words in a given experience are inversely correlated with taste-market fit. Anyone can generate outputs, but can they create with taste, discern what feels exactly right for a tiny niche?