What We Actually Mean by Taste
Deconstructing the black box of professional taste.
Ask a senior person why AI won’t replace them and you get the same two words every time. Judgment. Taste.
They’re said the way people used to say grace. Not as arguments. As wards. The words land, everyone nods, and the conversation moves on. Nobody asks what’s actually inside them.
I want to open the box.
Taste is three things pretending to be one
When someone says they have taste, they mean one of three things. They don’t know they mean three, because from inside, it all feels like the same thing. The three got bundled somewhere around year ten of their career and have been moving together ever since.
The first is pattern recognition. The real kind. You’ve seen this shape of problem enough times that you know which variables matter and which are noise. A good editor reads a paragraph and knows it’s broken before knowing why. A good investor reads a deck and knows the founder is lying to themselves before they can name the tell. That isn’t magic. It’s compression, earned the slow way, by actually doing the thing for years.
The second is preference. Opinions you formed early, defended often, and eventually stopped distinguishing from facts. You like clean interfaces. You hate meetings that should’ve been emails. You think serifs look serious and sans-serifs look cheap. None of these are wrong. None of them are right either. They’re yours. You’ve carried them around long enough that they feel like truth.
The third is the one nobody admits to. Reputation protection. The part of you that knows what happens when you back something publicly and it flops. The cost is asymmetric and you know it cold. Missing an opportunity is invisible. Endorsing a dud is not. So a quiet part of your nervous system has learned to lean toward no, and dress the lean in the language of standards.
All three feel like taste. All three answer to the same name.
When the world was slow, this worked
For a long time you didn’t have to separate them. Pattern recognition, preference, and reputation protection mostly voted together. The world moved slowly enough that what you’d seen before kept happening, what you preferred was still defensible, and the thing that would damage you if it failed was usually also the thing that shouldn’t ship.
Three votes, same answer, decision made. Call it judgment. Charge for it.
The whole apparatus of seniority assumed those three would keep agreeing. That’s why we paid for grey hair. Not because older people are smarter. Because in a slow world, the bundle is reliable.
In a fast world, the bundle splits
What’s happening now is that the three are voting differently for the first time in most senior careers, and nobody has the words for it.
Pattern recognition says: this is actually new. I haven’t seen this shape. The variables I know how to weigh aren’t the ones that matter here.
Preference says: but we don’t do it this way. We’ve never done it this way. It feels wrong.
Reputation says: if I back this and it fails, I’m the person who fell for the AI hype. If I block it and we miss, no one will remember. Block it.
The bundle splits. And because the person carrying it can’t tell which vote is talking, the loudest one wins. The loudest one is almost always reputation, because it has the most at stake. So the senior person says “this doesn’t meet our standards” or “it isn’t quite there yet” or “I just don’t think it’s right.” They mean it. They feel it. They’re not lying.
They’re also not telling the truth, because they can’t see which of the three is moving their mouth.
What we’ve been calling taste, in a lot of these moments, is the third vote wearing the first two’s clothes.
The young person’s accidental advantage
The thing a 25-year-old has isn’t better taste. It’s that the bundle hasn’t formed yet. Their pattern recognition is thin. Their preferences are still negotiable. Their reputation is small enough that protecting it doesn’t dominate every call. The three votes are three separate things in their head, and they can hear which one is talking.
This isn’t a virtue. It’s a phase. Most of them will lose it. Somewhere between year five and year fifteen, the environment will teach them that endorsing a failure costs more than blocking a winner. They’ll learn to dress that lesson in the language of standards. They’ll start saying the word taste.
What they have right now, almost by accident, is the ability to say “I don’t know, let’s try it” without three parts of themselves vetoing the sentence on the way out.
That’s the thing AI is exposing. Not that taste is fake. Real pattern recognition is still real and still valuable. But that most of what gets called taste in senior rooms is two parts preference and reputation, dressed as one part pattern recognition, charged at the rate of all three.
So what
If you’re senior, stop trusting the bundle. When the no starts rising, ask which vote it is. Is this a pattern you’ve actually seen, or a preference you’ve defended so long it feels like one? Is this a real flaw, or a fear about what backing it would cost you? You won’t always know. Sometimes the honest answer is “I can’t tell.” That’s more useful than “it doesn’t have the polish yet,” because at least it’s true.
The senior operators who’ll matter in the next decade are the ones who can pull the three apart. Who can say: my pattern recognition says go, my preference says no, my reputation says block, and I’m going to act on the first one. That’s hard. Nothing in your career prepared you for it. The whole point of seniority was that you didn’t have to.
If you’re young, your asset isn’t your taste. You don’t have any yet. The asset is that you can still tell the difference between what you actually think, what you prefer, and what would cost you to be wrong about. Most people lose that. The environment trains it out of them, gently, over a decade, and at the end of the decade everyone calls the result wisdom.
Find a place that rewards keeping the three separate.
Most won’t.
Pick carefully.