Taste was never the moat

Every third post in your feed is explaining that taste is the last scarce resource. The machines can build anything now, the pitch goes, so the one edge left is your exquisite and brilliant eye. It is a flattering thing to be told, which is most of why it sells, and it is epically wrong.

Everyone alive is sure theirs is good. I have never met the person who confessed to bad taste, and I am not convinced they exist.

You type a sentence and out comes a film, a brand, a song in the style of someone who spent ten years learning to sound like that. Everyone has the wand. And everyone is sure their taste is excellent, which is the problem, because if everyone has the wand and everyone can point it, someone explain the feed. Explain the baby swaddled in spaghetti. Explain the man celebrating his one-thousand-eight-hundredth birthday. Explain why the internet is filling with slop while every single person holding the wand is certain their eye is impeccable.

A French sociologist named Pierre Bourdieu spent the 1960s working out what taste actually is. It is a uniform. What you find beautiful, what makes you wince, the wine you would never serve, all of it maps onto where you sit in the pecking order. Taste sorts people, and it sorts you while you are busy being sure you are the one doing the sorting. His line, which I have never shaken off: taste classifies, and it classifies the classifier. Everyone has it, because everyone has somewhere to broadcast from. It is as common as an opinion, and worth about as much.

But common is the least of it. The real problem is that taste is a verdict that never gets graded. Watch what happens when the tasteful thing flops. The product you knew was good. The essay only a philistine could fail to love. It dies. Do you decide your taste was off? You do not. I do not. I have shipped the beautiful thing, watched it sink, and blamed the launch window, the season, the audience that wasn't ready. I have blamed Mercury retrograde. The verdict walks away from every flop without a scratch, because I never once let a result touch it. Untested, taste never gets better. You stay exactly as sure as the day you formed it.

Taste is a verdict on the thing in front of you. Is it good. Discernment is a bet on what that thing does once it is loose in a system full of other people, other incentives, and a clock. Take a chair. On the showroom floor your taste can be perfect: the grain, the proportion, the way the light sits on it. But the one quality that matters is the one the showroom cannot show you, which is what the chair does to your lower back at hour six of back-to-back meetings. That quality is not in the chair. It only shows up once you use it. Taste judges the chair. Discernment judges hour six.

Discernment gets graded late. Sometimes decades late. In 2006 a designer named Aza Raskin built infinite scroll, the thing your thumb does where the feed never ends. He built it to be helpful. Smoother reading, fewer clicks. The taste was immaculate. Then it got loose inside companies paid by the minute of your attention, and the helpful little feature became the thing that eats your evening. Twenty years on he sits in courtrooms, most recently one in New Mexico, testifying about what a feed with no bottom does to a child. His taste was never the problem. The system he couldn't see past was.

A system pays out on the incentive you set, not the goal you meant. There is a story, probably made up, about the British in colonial Delhi: worried about cobras, they paid a bounty for dead ones, so people bred cobras to cash in. The bounty ended, the worthless snakes were let loose, and the city had more cobras than when it started. Probably didn't happen. The mechanism does, every time. Pay for dead cobras, get cobra farms. Pay for engagement, get rage. Taste asks whether your goal is good. Discernment asks what the incentive actually makes people do, which is the colder question, and the only one the system answers.

Snapchat invented the disappearing Story. Instagram copied it to the pixel and didn't bother to change the name, and inside five months the copy had a hundred and fifty million people a day, level with the thing it copied. Instagram's own boss said of Snapchat, with a straight face, they deserve all the credit. Taste rates your move on a still board. The board is never still. The discerning question is the one a chess player asks and a flatterer never does: what does the other side play next?

Discernment is not a gift you are born with, whatever the people who sell it that way would like you to think. It is a skill. It runs on rules, and rules can be taught. Start with the gut, since everyone is so attached to theirs. Two researchers from opposite corners, Daniel Kahneman on how intuition fails and Gary Klein on how it works, argued for years and ended up agreeing. The gut can be trusted under two conditions, and only two. The thing you are judging has to be regular enough to hold real patterns. And you have to have practised in it long enough, with fast and honest feedback, to learn them. Chess has both, so a grandmaster's gut can be trusted. The stock market has neither, so a pundit's gut can't be, however sure he sounds.

And ordinary people can build it. Between 2011 and 2015 the American intelligence community ran a forecasting tournament, and a psychologist named Philip Tetlock put volunteers off the street, people with day jobs and no clearance, up against professional analysts who had clearances and classified cables. The amateurs won. They beat the professionals by something close to thirty percent. Their whole edge was a handful of habits anyone can copy: they broke the big question into small ones, put a number on every guess, kept score, and changed their minds the moment the score told them to. It is a practice. Practices have steps.

Run the system forward. Before you ship, write down what the thing does once it is loose: the second move, the incentive you are about to create, what your rival does in reply. Tell the AI to attack the plan and play the rival and beat you. It will. It will also write you a beautiful disaster for a plan that is perfectly fine, because a story costs it nothing, so which disaster is real is still your call to make.

Write the prediction down, and make it something reality can break. Not "this'll do well." This hits a number by a date, and if it does something else, I was wrong, said out loud while your pride is still in the room to argue. The AI drafts the prediction and the kill-switch in seconds. It cannot decide what you are willing to be publicly wrong about. That is the bet, and the bet is the whole point.

Get the feedback fast, and clean, because that is the condition everything else starves without. Shorten the loop until it stings. Ship the smallest real version to real people and watch what the system does, not what the room says. The AI will build the test and read ten thousand lines of the result. The one thing it won't tell you is that you're measuring the wrong thing.

Keep a log, and review it on a calendar, not when the mood takes you, because the mood only takes you after a win. Every real bet: what you predicted, why, and what actually happened. Let the AI keep the ledger and dig up the bets you quietly buried. It cannot supply the nerve to look.

And then the one that breaks the pattern. The bet you only get to place once, where there is no small version and no second try and the two conditions for a trustworthy gut are just not there. The instinct is to trust your gut harder. That is exactly the instinct to ignore. When a decision can't be graded, you don't make it braver, you make it reversible. Buy the option, not the obligation. Keep the bet small enough that being wrong won't end you. Ask the AI which of your doors swing both ways and which lock behind you, then decide how much of your future to stake on a verdict that, this once, will never be graded at all.

All of it runs through the AI, and the AI does one reliable thing. It amplifies whatever you walked in with. Bring discernment and it becomes the most powerful tool you have ever held, running the simulations, drafting the predictions, keeping the log, finding the exits, faster than you could alone. Bring taste, just the warm certainty that your eye is good, and it does that too, at scale. It will help you make more slop. Prettier slop. Slop at a higher resolution, delivered to the same nowhere, faster, and better lit. Same tool. The whole difference is what you walked in with.

So the question was never the one you have been answering. You have been answering do I have good taste, and your answer is yes, and you are right, and it changes nothing, because everyone's answer is yes. That unanimous, ungraded yes is most of why the internet looks the way it does. The real question is smaller, and worse. Which of your decisions this year ever actually got graded? Not admired. Not approved in a meeting. Not proven right by a number you picked yourself. Graded: by the system, against the clock, by people who had a move of their own. If you can't name three, you don't have a taste problem. You have a feedback problem. And that was the only part that was ever yours to build.