You Can’t Prompt Your Way to Taste

A fundraising announcement video came across my feed last week for a startup called Taste Labs. Its mission? To end AI slop, build the data and infrastructure layer to give AI models and agents taste.

This comes on the heels of the “taste” trend plaguing Silicon Valley. As Kyle Chyka noted in a recent New Yorker column, “[taste] has become as much of a Silicon Valley cliché as ‘disruption’ was in the twenty-tens.” 

Tech bros view taste as inherently profitable. With enough data, enough training, and enough fine-tuning on the right inputs, you can produce output that doesn't just perform well but actually embodies it.

What Silicon Valley is getting wrong about trying to commoditize taste reveals something it has never fully understood about how culture works, how authority is assigned, and what media has always actually been for.

Before media was an industry, it was an editor's judgment made public.

The great magazines of the twentieth century were not primarily vehicles for content. They were instruments of cultural orientation. Vogue didn't just tell you what to wear. It told you who was worth being, and by extension, who you might become if you paid close enough attention. 

Condé Nast built an empire not on information but on aspiration and selectivity: the premise that being chosen, curated, and anointed by an editorial gaze carried meaning that mass production could never replicate. Scarcity was the point. To appear in those pages, or to be the kind of person who read them, said something real about where you stood in the cultural hierarchy. The goal was never to build a big circulation, but the right one. 

This was taste as identity infrastructure, a framework for self-construction. Editors weren’t just curators who covered culture, but cultural authorities who convened it. They exercised judgment about what mattered, and their judgment compounded over time into something readers trusted more than their own instincts. 

Yet, the power was never just in what was published; it was in what wasn't. Editorial restraint was the signal that separated taste from trend. Trend is about reading what's already moving, whereas taste is about knowing what deserves to move before it does, and being willing to wait. And to shape taste, one had to embody it.

The fracture started when the question shifted from "Is this worth your attention?" to "How much attention can this generate?" Traffic, reach, impressions, and shares are not taste metrics, but attention metrics. And when you optimize for attention at scale, you don't sharpen judgment. You train yourself to recognize patterns in what has already worked, and you reproduce those patterns until they stop working and you need new ones.

AI didn't create the problem. AI revealed it.

In a 2025 MIT Media Lab survey, ChatGPT users showed the weakest neural engagement and the lowest sense of ownership over their work. The data captured the consequence of what happens when you outsource the judgment layer of creative work: you get output that looks right but doesn't carry the cognitive fingerprint of having actually decided anything. Trustworthiness, it turns out, is not a function of polish. It is a function of the accumulated decisions a specific person, with a specific history, made in a specific sequence.

That is what taste actually is. It is not an aesthetic preference. It is a record of what you have chosen to pay attention to, for long enough, with enough genuine curiosity, to develop a filter. Taste is the byproduct of sustained exposure processed through a particular self. It requires time, friction, and most importantly, an identity stable enough to have preferences in the first place.

In a world of scarcity, we treasure tools. In a world of abundance, we treasure taste. It's the judgment about what the tool should make. Taste is that judgment operating at the level of form, not just function. Which is precisely why you cannot train it into a model.

What Taste Labs—and every “taste layer” startup is building is not taste. It is pattern-matched to things that previously required taste to make. There is a difference. Taste produces the original. The taste layer produces a highly legible approximation of the original. At the level of a single artifact, the gap can be hard to see. But, over time, throughout a body of work, it becomes the only thing you see.

This is where the media history becomes directly relevant to the creators building right now.

The magazines were aspirational because you could sense, across every issue, that someone had judgment. Their taste was the coherence principle that made the whole thing feel like it came from somewhere, not from everywhere. That singularity of perspective was what readers were actually paying for when they subscribed. They were purchasing access to a consistent way of seeing.

This is the structure underneath every independent media brand that earns genuine authority. The publication is not the product. The judgment that produced it is. And judgment of that kind is assembled slowly, through the accumulation of real experiences processed through a real self. It cannot be bought or prompted—it can only be developed.

Taste is a capacity built through sustained, friction-rich attention — curiosity, exposure, and self-knowledge compounding into a filter that no optimized input can replicate. Lived experience is the data. The filter it produces is editorial intelligence. The choices you make, the curation you exercise, the things you refuse: all of it is downstream of an embodied question you keep asking until you have an answer.

The cultural tension underneath all of this is between two deeply incompatible theories of value.

The first theory, which Silicon Valley has always operated from, holds that value is generable. Given the right inputs and the right system, you can produce the right outputs at scale. This theory has been enormously productive for a certain class of problems. It has never been true for the problem of meaning.

The second theory, which the great editors knew and which the best independent media builders are rediscovering, holds that value is evidenced. You cannot generate authority. You can only accumulate it by making enough good decisions in public, over enough time, that people who are paying attention begin to trust your judgment before you've explained your reasoning.

This is the reason the magazine editor was a cultural celebrity in a way the algorithm has never been, despite the algorithm's vastly greater reach and processing power. The editor was a human who had preferences. They made calls that were wrong sometimes and right often enough to earn the benefit of the doubt. They were accountable to a point of view in a way that made agreeing with them feel like belonging to something.

That is what is actually being eroded by the commoditization of taste. Not aesthetics or polish, but the sense that the thing you're reading was made by someone who cared enough about what they believe to be wrong about it sometimes. Genuine taste always carries the risk of being wrong. Pattern-matched approximation of taste, curated by AI, cannot be wrong in the same way, because it was never actually committed to anything in the first place.

The question worth asking right now is whether new independent media can do what the old magazines did: not just distribute taste, but cultivate it. Whether a newsletter, a podcast, or a carefully built social presence can function the way a great editor once did. Whether small, high-conviction, deliberately restricted publications can create the same orientation infrastructure that the print era did at an industrial scale.

The publications gaining the most trust are not the ones with the widest reach. They are the ones where you can feel a specific judgment operating across every surface. 

You can’t publish your way into taste or build an editorial identity by deciding what your categories are and filling them efficiently. The editors who shaped culture throughout the twentieth century weren’t systematizing workflows; they were expressing a worldview built over decades of genuine attention, genuine curiosity, and genuine refusals in their actual lives.

The question for creators now is not, “How do I systematize taste?” It’s, “What have I paid enough attention to, long enough, to actually know something?” And then, “How do I build a publication that makes that knowing visible?”

The creators who understand this have the most important competitive advantage available right now — a self that has been paying attention long enough to have something to say that couldn't have come from anywhere else, and the editorial infrastructure to make that self legible, consistently, over time.

That is what the magazines were. That is what the best independent media will be. The window between the collapse of institutional taste and its commoditization is exactly where editorial intelligence is scarcest and most valuable.

The question is whether you're building toward it, or just publishing while waiting to see.

What’s trending across the interweb

Media news, launches, and insights

What I’m consuming this week