On Wednesday, a software engineer writing under the name NotAShelf published a 15-minute read titled “Taste Is All That’s Left.” By Thursday morning it sat at #1 on Hacker News with nearly 500 upvotes and over 350 comments, most of them variations on “yes, exactly.” The essay’s argument is elegant: now that AI can generate passable code, passable prose, and passable design at near-zero cost, the only thing that distinguishes good work from the flood of “good enough” is taste — the slow, hard-won judgment to know what deserves to exist.
It is a beautiful essay. It is also, in ways its audience seems not to have noticed, an argument for inherited advantage dressed in the language of craft.
The Curriculum Nobody Can Afford
The essay’s most quoted line is also its most revealing: “The friction was not an obstacle to developing taste. The friction was the curriculum.” The idea is that the difficulty of making things — the years of producing bad work, the slow accumulation of judgment through failure — was not a bug but a feature. It filtered out the people who didn’t care enough to suffer through it.
This is true in the way that “the best way to learn French is to spend a year in Paris” is true. It is correct and it is useless to anyone who cannot afford the plane ticket. The “curriculum” of friction requires time, and time requires money, or a safety net, or both. The person who can spend five years making bad software before producing something good is, almost by definition, someone who does not need to ship quickly to pay rent. The essay’s ideal craftsman is a trust-fund kid of the spirit — someone whose material circumstances allow them the luxury of saying “no, again” while the market moves on.
This is not a criticism of the author, who is making a philosophical argument, not a policy proposal. But it is a criticism of the reception. The Hacker News crowd nodding along to “friction was the curriculum” is, disproportionately, a crowd that had the curriculum subsidized — by parents, by a first job that tolerated mediocrity, by a tech economy that for two decades paid junior developers handsomely to learn on the job. That economy is contracting. The curriculum is getting more expensive just as the essay declares it essential.
The Taste That Actually Wins
But there is a deeper problem with the essay’s argument, and it has nothing to do with privilege. It has to do with speed.
The taste the essay celebrates is slow. It “says ‘no, again.’ It sends the plausible thing back because plausible is not the same as right.” This is the taste of the craftsman, the connoisseur, the person who refines a single object over many iterations until it is exactly right. It is the taste of the novelist who revises a manuscript for two years, the designer who moves a button four pixels to the left after the seventh round of user testing.
That kind of taste is not the solution to an AI-saturated world. It is the casualty.
When a machine can generate a thousand plausible variations in the time it takes a human to evaluate three, the scarce resource is not refinement. It is decisiveness. The person who wins is not the one who can tell you, after careful study, why variation #847 is superior to variation #846. The person who wins is the one who can look at a thousand outputs and pick the right one before the standup ends — and who feels no grief about the 999 that die in the process.
This is a different kind of taste entirely. It is not cultivated through slow, humiliating failure. It is cultivated through volume, through pattern-matching, through the willingness to make fast calls and live with them. It is the taste of the editor, not the author. The curator, not the creator. And it is not romantic.
The Editor, Not the Craftsman
I asked a product director at a mid-sized SaaS company — someone who manages a team that now generates roughly 40% of its production code with AI assistance — what she looks for in new hires. Her answer, sent over Slack while her team was in the middle of a sprint review, was blunt: “I don’t need someone who can tell me why the button should be 4px to the left after three rounds of revision. I need someone who can look at fifty button variations and pick the right one before I finish my coffee. The first person is a craftsman. The second person is useful.”
She is not alone. Across the industry, the premium is shifting from the ability to make to the ability to choose quickly. And choosing quickly is not the same skill as choosing well over a long horizon. It is a skill that rewards confidence, a high tolerance for imperfection, and a certain unsentimentality about the work. You cannot mourn the 999 variations you killed. You cannot even look at them.
The essay’s author is right about one thing: when making becomes free, choosing becomes the craft. But he is wrong about what kind of choosing will be rewarded. The market does not want a sommelier who can spend twenty minutes contemplating the nose. It wants someone who can pick the right bottle off a shelf of ten thousand in under ten seconds, and who is right often enough that the mistakes don’t matter.
That person will not have developed their taste through years of slow, painful iteration. They will have developed it through exposure to staggering volume, through the repeated exercise of snap judgment, through a kind of mental muscle memory that has nothing to do with the romantic ideal of the craftsman suffering for his art. And they will be faster than the person who still thinks taste means saying “no, again.”
The essay went viral because it told a flattering story to people who fear being replaced. You will not be replaced, it said, because you have taste, and taste cannot be automated. This is comforting. It is also, I suspect, wrong. Taste can be automated — not the slow, curatorial taste of the connoisseur, but the fast, commercial taste of the editor who makes good-enough decisions at machine speed. And when that happens, the craftsman waiting to revise his work one more time will not even see what passed him by.