Two Days to Choose a Color

Twice a year, in a European capital whose name is not disclosed, a handful of people meet behind closed doors. Two days of presentations and debates. When they leave, they have chosen the following year’s color. Last December, the verdict for 2026 came in: Cloud Dancer, reference 11-4201. White.

Nobody elected them. No consumer voted. And this year, white appears on Motorola smartphones, Post-it notes, Command adhesives, Joybird furniture and Play-Doh. The vice president of the institute organizing all this explains that the color “reflects what people are looking for.” Perhaps. What is certain is that product directors across a dozen industries opened a file in January and wrote the word white in it.

The objection comes immediately, and it is a good one: if it is arbitrary, it shouldn’t work. A choice without an objective basis should collapse on contact with reality. People should shrug and buy the color they prefer.

Exactly the opposite happens. And I think this is one of the most misunderstood aspects of the economy that is coming.

Let’s start with wine, because that is where the demonstration is cleanest. In 2001, in Bordeaux, Frédéric Brochet had fifty-seven oenology students taste the same Bordeaux one week apart. The first time, the bottle bore a table-wine label. The second, a grand cru classé label. The liquid was identical. Fifty-two of the fifty-seven rated the grand cru higher.

The most interesting part isn’t the score. It is the vocabulary. Lexical analysis of the comments shows that the tasters looked for flaws in the first glass and qualities in the second. They didn’t simply judge the same wine differently. They perceived it differently.

Seven years later, a team from Stanford and Caltech repeated something similar, but this time inside a scanner. The same wine, two stated prices: ten dollars and ninety. Subjects said they preferred the ninety-dollar one. That was predictable. What was less predictable was that the medial orbitofrontal cortex, the region encoding experienced pleasure, became more active¹. They weren’t reporting pleasure out of politeness. It was visible in the imaging.

A label doesn’t change the wine. It changes the pleasure. And the pleasure is perfectly real.

That answers a question everyone asks the wrong way. People ask why someone pays three hundred dollars for a T-shirt that costs two dollars to make, assuming they are being taken for a ride. But they aren’t necessarily being fooled. They are buying a difference in pleasure that really exists, that is measurable, and whose cause simply isn’t in the cotton.

Bourdieu wrote a thousand pages about this in 1979, interviewing twelve hundred French people about their music, their furniture and their food. His conclusion fits into one sentence: taste classifies, and it classifies the classifier. Dominant classes don’t merely have preferences, they turn those preferences into the standard against which everything else is measured, then present that standard as natural. He called this the naturalization of the arbitrary.

Bourdieu discovered the mechanism and spent the rest of his life outraged by it. One industry read the same thing without the outrage and turned it into a business model: fashion.

Hermès deliberately limits production, maintains waiting lists lasting years, refuses to meet all demand, and delivers an operating margin no conventional manufacturer reaches. A Birkin 25 went from $9,400 in 2016 to $13,500 in 2026, and resells for two to three times the store price.

None of those numbers can be explained by the leather.

Now the real question: why is this mechanism going to spread beyond fashion into everything?

The 2026 Stanford HAI report says something that should occupy more people than it does: the leading models have become “almost indistinguishable” from one another. In March, six labs were neck and neck in the Elo rankings. Benchmarks designed to last years saturate within months. The cost of obtaining a given performance level falls by a factor of five to ten each year. Competitive pressure, they write, has shifted toward cost and reliability.

Translation: what can be measured has stopped deciding the winner.

And this movement isn’t just about AI. It concerns everything AI will touch, which is almost everything. When objectively measurable quality becomes abundant, uniform and nearly free, it gradually leaves the playing field. It doesn’t disappear: it becomes a prerequisite, which amounts to the same thing from the standpoint of differentiation.

That leaves one thing to fight over: the je-ne-sais-quoi.

Look at what is already happening. In May, Palantir, a data analytics company serving the military and intelligence services, released a $239 work jacket, bright blue or black, made in Montana. Sold out in five hours. Not a corporate polo, a French worker’s jacket, carrying all that European working-class history, sold by a company supplying ICE and the Israeli military. The same company had already sold a T-shirt that read “Dominate.”

They understood before the others that the jacket is part of the product.

Look at Pop Mart. Thirty-seven billion yuan in revenue in 2025, roughly $5.4 billion, up 185%. Gross margin: 72%. The product is a plastic figurine with teeth, no use, no function, sold in an opaque box so you don’t know which one you are buying. A third of revenue comes from a single character.

No data could have predicted Labubu. And Pop Mart itself doesn’t know how to make the next one, a fact the stock market reminded it of by wiping 30% off the shares on earnings day.

Silicon Valley sensed something and started talking about taste last February. Paul Graham, Greg Brockman, half the founders in San Francisco: within a week, everyone agreed that taste would be the last defensible skill. Kyle Chayka was right to mock it: within months, the word had become the “disruption” of the twenties.

They saw it correctly and got it entirely backward.

In their mouths, taste is a rampart. Something you cultivate to protect yourself, a moat you dig around yourself while the machine swallows the rest. It is the posture of people retreating.

Fashion has never used taste as a rampart. It uses it as a force for projection. You don’t shelter a taste, you spread it, establish it, make it so obvious that people who don’t share it feel not different but behind.

That is why we must stop treating taste’s arbitrariness as a weakness. It is precisely because nothing in reality grounds the choice of white for 2026 that whoever manages to impose it achieves something.

An objective fact imposes itself. It asks nobody’s permission and confers no particular power on whoever states it.

Taste doesn’t. Taste establishes itself only through an act.

The expression je-ne-sais-quoi was coined in seventeenth-century France to designate precisely what seduces and escapes reason. Jankélévitch devoted three volumes to it, defining it as what “protests and murmurs again within us against the success of reductionist enterprises.” He was talking about philosophical systems.

The sentence applies just as well to dashboards.

People will say algorithms can predict trends with 90% accuracy, and that is true. They can measure a taste once it exists. They can amplify it, accelerate it, flatten it.

But they cannot decree it.

And there, I think, lies the important distinction: predicting and instituting are two different acts.

Predicting means discovering a preference that already exists. Instituting means creating value where none yet existed. It means looking at a space in which several futures are possible and deciding it will be this one.

The committee will meet again this spring, somewhere in a European capital. For two days, they will discuss, hesitate and argue. Then they will choose 2027.

It will be arbitrary, indefensible, impossible to justify completely through data.

And the year after, you may have it on your phone.