When the machine produces execution for free, and infinitely, value moves to the one thing it can't do: knowing what's worth making. Taste stops being a nice-to-have. It becomes the job.
Execution just became free
For the whole history of making things, execution was the bottleneck. Producing a working screen, a decent illustration, a functional feature took skill, skill took years, and years cost money. The economy of creative and technical work was priced on that scarcity. We paid for the making.
That pricing just collapsed. Code, layouts, images, copy: a model now produces competent versions of each in seconds, at near-zero marginal cost, in unlimited quantity. Whatever you think of the quality, the economic fact is settled. Execution went from scarce to abundant.
And abundance does a predictable thing to value: it pushes it up the stack, toward whatever is still scarce. So the useful question, for anyone who builds, is what's still scarce.
The value moves to taste
Watch what the machine produces when nobody exercises judgment: the average. Fluent, plausible, infinite average. I've written about why this happens mechanically with AI-generated apps. The model can execute anything. It can't want anything. It has no opinion about what should exist.
Now look at what's left standing when execution costs nothing. Knowing what's worth making, and what isn't, which is harder. Recognizing, among a hundred generated options, the one that's right, and being able to say why. Holding an intention steady across a thousand micro-decisions so the result feels like one mind made it.
There's an old word for that cluster: taste.
We used to treat taste as the decoration on top of competence, the extra that distinguished the very good from the great. That hierarchy has inverted. When anyone can generate the executions, what differentiates products is no longer who could build it but who could choose.
I'm far from alone in seeing this. Investors now call taste the new moat. And the sharpest version of the argument I've read is Ali Albakri's essay on taste in the AI era: AI, he writes, "democratizes execution without democratizing judgment". Everyone gains the ability to produce. Almost nobody gains the discernment to choose. It's the right diagnosis, and it usually stops there, at the flattering conclusion that the tasteful few will inherit the earth. I want to push on the half nobody addresses: can judgment be democratized too?
Taste sounds like an innate gift, which would mean no. It isn't one. Taste is trained judgment: exposure, repetition, attention, thousands of small verdicts with feedback. Which means it can be developed. And, less obviously, it can be engineered.
A design system is an act of taste
The other day, on the French podcast Continue tu m'intéresses, hosted by Patrick Baud (an excellent show, by the way, for those who understand French), I heard the philosopher Charles Robin draw a line that stuck with me: an AI that produces thought bothers him less than an AI that produces art, because art expresses an affect, something of the heart, while thought is judged only on the pertinence of what it states.
I fully agree with him. And his line helped me place my own trade: we don't make art, an app is judged on its pertinence, the head side. But our work is about bringing apps as close to that border as they can get. Our design system is the means we found for it: we put into it our taste, our affects, what the years have taught us, in a form that thousands of app creators can receive, and that a machine can apply. The goal was never to flatten that taste by writing it down. The goal is to share it.
Taste held in one person's head is a bottleneck. It dies at the edge of that person's availability. But taste can be written down as rules. This is how type scales. This is how spacing breathes. These are the four things a color is allowed to mean. This never goes with that. Do it rigorously enough and you get a system that applies judgment automatically, at scale, on behalf of people who don't have it, and against machines that don't either.
That's what a design system is: taste, formalized until it becomes reproducible. Ours fits in three layers and eighteen typographic levels. But the engineering is the smaller half of the story. The bigger half is that every rule in those layers is a crystallized aesthetic verdict. Someone decided what "right" looks like. The architecture just makes that decision tireless.
Seen this way, the division of labor in the machine age gets clear. The machine executes, infinitely. The system holds the judgment, structurally. The human does the one thing left that neither can do: deciding what the judgment should be, and what deserves to exist at all. What makes your product distinct sits upstream of all execution. My colleague Jérôme has written on exactly that.
The augmented curator
So the anxious question, does the machine replace the designer, the developer, the maker, gets an answer I find genuinely optimistic.
The maker's role shifts from executor to curator. The person with the intention, directing infinite cheap execution, selecting, rejecting, refining. Less time transcribing, more time deciding. That's not a demotion. Deciding was always the noble part of the work. Execution was just where the hours went.
For fifteen years I've built tools whose premise is that execution shouldn't be the barrier between someone's intention and a real product. AI radicalizes that premise beyond anything we imagined. It also sharpens the split this essay has been circling: a system can democratize the floor of judgment, so that nothing ships broken or incoherent. The ceiling, knowing what's worth making, no system can hand you. The people who thrive next won't be the ones who execute fastest, the machine won that contest. They'll be the ones who build that ceiling for themselves. It can be learned, and the time to start is now, while everyone else is still racing the machine at its own game.
Mathieu Poli — Head of Frontend Engineering @ GoodBarber. I teach and write about frontend engineering, product design, and AI — and everything that happens when the three meet. · X: @hellomathieup
Top comments (0)