The Taste Gap

Building software used to be the bottleneck. If your startup wasn't growing, the default excuse was reasonable: we haven't built enough yet, the product isn't ready, we need three more months of engineering. The difficulty of building provided cover for not having figured out whether anyone wanted what you were building.

AI removed that cover. When a solo founder can ship a working product in a weekend, "we're still building" stops being a credible explanation for why no one's using it. The thing is built, it's live, and nobody cares.

That's when you're face-to-face with the actual hard part – the part that was always hard, but was easy to avoid when building took long enough to delay the reckoning. Knowing what's worth building in the first place.

I've started calling this the taste gap. Not taste in the design sense – not typography choices or color palettes. Taste as in the skill of recognizing a real problem, understanding what a solution actually needs to do, and having the discipline to cut everything else. The judgment to build what people want, not just build something.

That skill was always the differentiator. The founders I've watched find PMF weren't the best engineers. They were the ones who understood their users deeply enough to make the right calls about what to build and – just as importantly – what not to. Engineering speed was a constraint that slowed everyone down roughly equally. What separated outcomes was judgment.

Now that the speed constraint is gone, judgment is all that's left. And the gap between founders who have it and founders who don't is becoming painfully obvious. I see it in the wave of AI-built products hitting the market – technically functional, shipped fast, solving problems nobody actually has. Or solving real problems in ways that miss the point. The demo looks great. The product-market fit isn't there.

The way I think about it: speed is a multiplier. If you multiply zero judgment by ten, you still get zero – just faster. Multiply good judgment by ten and you get a decade of iteration compressed into months. Same tool, opposite outcomes.

The uncomfortable truth is that this skill can't be shortcut. You don't prompt your way to understanding what people need. You develop it by talking to users, getting it wrong, watching people struggle with your product, and doing that enough times to build genuine intuition. AI can accelerate every part of building a startup except the part that matters most: learning what's worth building.

And that gap is only getting wider as the tools get faster.