Taste is a moat: why the best builders can't fully explain their decisions

Ask a great builder why they cut a feature everyone loved and you’ll often get a shrug. Not evasion — a genuine blank. They knew it was wrong before they could tell you why. That gap between the certainty and the explanation is the most interesting thing in product work, and I think it’s also the most defensible.

Here’s how I’ve come to understand it. Taste is compressed judgment. Somewhere along the way, a person absorbs thousands of small examples — products that felt right, products that felt off, moments where a decision aged well or curdled — and their brain quietly builds a model out of the pattern. The model runs faster than language. When they look at a new thing, it returns a verdict before the reasoning has finished loading. So the shrug isn’t a lack of thought. It’s thought that finished so fast the words never caught up.

Which means “why did you cut that?” is a slightly unfair question. You’re asking someone to decompress, in real time, a judgment that was valuable precisely because it stayed compressed. The honest answer is usually reconstructed after the fact — a story assembled to explain a call that was already made. That’s not lying. It’s the difference between knowing how to ride a bike and writing the physics of balance. Both are real. Only one is fast enough to be useful in the moment.

This is why taste beats process in the messy early stretch of a project. Process is compressed judgment too — but it’s someone else’s, frozen into steps, optimized for situations that repeat. A checklist is a brilliant thing when the ground is stable and the failure modes are known. Early 0→1 work is the opposite: you don’t yet know what matters, the feedback is thin and noisy, and half the important variables haven’t shown up. In that fog, a rulebook gives you false confidence and the wrong resolution. Taste gives you a fast, adaptive guess that updates on contact. It’s worse than process when the problem is legible, and much better when it isn’t. Most of the interesting problems aren’t legible yet.

And this is where the moat lives. A competitor can read your strategy doc, copy your roadmap, hire away your process. What they can’t do is copy the tacit model in your best people’s heads, because it was never written down — it can’t be written down without losing most of its value. You can hand someone your conclusions. You can’t hand them the ten thousand examples that let the conclusions feel obvious. That’s the whole reason taste is defensible: the thing that makes it powerful is the same thing that makes it uncopyable. If you could fully document it, it wouldn’t be much of an edge.

I want to be careful here, because this is exactly the point where the idea tips into mysticism, and I don’t buy the mystical version. Taste isn’t a gift you’re born with or a vibe you either have or don’t. It’s a learned compression, which means it can be trained — and the failure mode of “trust your gut” is that untrained guts are just confident and wrong. The romance is the enemy. The interesting question isn’t whether taste is real. It’s how you build it on purpose.

The mechanism, as far as I can tell, is volume plus friction plus feedback. Volume: you have to see an enormous number of examples, good and bad, close up — not admire finished products but sit inside the decisions that made them. Friction: you have to actually commit to a verdict before you know the outcome. Judgment you never risk anything on doesn’t sharpen. And feedback: you need to find out whether you were right, and sit with the times you weren’t, long enough for the model to update. A thousand calls with no consequences teaches you nothing; a hundred calls you cared about and got graded on teaches you a lot.

There’s a small practice buried in that. When your gut fires, try to write down the reason before you act — not to slow the gut down, but to catch it in the act. Over time you’ll notice which reconstructions were real signal and which were flattering noise. That’s how the tacit becomes slightly less tacit without pretending it’s a formula.

So maybe the shrug is the point. The builder who can explain every decision cleanly is probably working in a domain that’s already been solved, where the answers fit in a doc. The one who can’t quite explain the good call is often standing somewhere newer, running a model no one else has built yet. That inarticulate confidence isn’t a bug in their thinking. Handled honestly, it might be the most valuable thing they own.