Silicon Valley's New Obsession With 'Taste'

Silicon Valley’s New Keyword: “Taste”

I’ve been seeing the word “taste” pop up constantly in Silicon Valley lately. It’s a subject I covered in the book I wrote last year and in a blog post from two years ago. But as similar takes kept piling up, my reaction wasn’t “see, I was right all along” — it was “wait, are we even using this word the same way?” Rather than repeat an argument that’s already widely known, I found myself more curious about what needs explaining next.

In a February 2026 post on X, Paul Graham wrote that as AI makes it possible for anyone to build almost anything, the taste to choose what’s worth building will matter more than ever. Greg Brockman and Koen Bok, co-founder of Framer, have made similar arguments about the growing importance of taste. The claim, in short: the easier the tools for making things become, the more choice matters.

Kyle Chayka, the New Yorker’s tech-culture columnist, has pushed back on this trend. He coined the term “taste-washing,” which describes how AI companies dress up automation in imagery of human sensibility and creativity to make the technology feel friendlier and more palatable.

The same word — “taste” — is being stretched to cover an aesthetic sense for beauty, the judgment to pick a good product, and the instinct to spot what will sell. That’s the distinction I want to start by pulling apart.

The Easier It Gets to Build, the More Choosing What to Build Matters

AI tools cut down the effort needed to draft text and images, and to try out simple apps.

Coding tools like Claude Code let you build a prototype even with little programming experience, and large language models (LLMs)¹ help you write or revise sentences. Even so, you still need the ability to run a product reliably and to review the quality of the output. It’s not fair to say technical execution has been entirely solved.

As the options multiply, there’s more to decide: what to make, and how good is good enough.

In his 2002 essay “Taste for Makers,” Paul Graham emphasized both a rigorous standard for recognizing good work and the ability to actually produce it. He moved across mathematics, painting, architecture, and programming to describe what good design has in common. Reading it again in 2026, what still matters is that he weighs the ability to judge alongside the ability to build.

Chaika cites an example in which Marc Andreessen suggested that even in the AI era, there’s still a role for venture capitalists who pick promising investments. In this context, taste comes close to the judgment of choosing what the market will value.

I think the problem with this kind of discussion is that it skips over what “good” actually means. Looking good, being easy to use, and being profitable sometimes overlap, but they’re different standards.

What Voltaire and Bourdieu Meant by Taste, Versus Silicon Valley

Placing this against other perspectives on taste makes the difference clearer.

The Voltaire that Csikszentmihalyi quotes emphasizes not just recognizing beauty but feeling it and being moved by it. Here, taste holds both emotion and aesthetic judgment together.

Pierre Bourdieu² analyzed, in his 1979 book Distinction, how taste relates to education, class, and cultural capital³. His point was that social background shapes not only what we like but also what we judge to be refined. This shouldn’t be read as claiming that a person’s every choice is determined solely by the class they were born into.

Seen through this lens, defining taste merely as “the ability to pick products that will sell in the market” is too narrow. There are works people love even when they make no money, and there are experiences where encountering something unfamiliar actually changes one’s taste.

That said, it’s also not fair to read Graham’s original essay purely as a claim about profitability. He talks about simplicity, designing to solve the right problem, and the process of iterating and refining. What I want to criticize isn’t his whole argument, but the way it gets summarized as “now all you need is taste” — a phrase that quietly erases execution and learning.

Why Chayka Criticized AI Companies’ Marketing

Chayka pointed out that AI companies’ promotions emphasize the image of a person making things by hand and thinking for themselves.

The examples he cites are Anthropic’s 2025 Manhattan pop-up café with hats bearing the word “thinking,” and OpenAI’s 2026 Super Bowl ad. The ad shows hands gripping bicycle handlebars or writing in a notebook. His point is that while introducing automation tools, these companies show scenes of people directly experiencing and creating things.

Chayka also connects this to past consumer cultures that expressed individuality through craft beer or independent music. His interpretation is that this is a recurring process where a distinctive lifestyle becomes marketing material for corporations. This should be taken as a critique that reads advertising and cultural phenomena, not as research proving the intent of every AI company.

I think this critique lets us separate a tool’s function from the self-image the tool promises. Helping you write quickly is different from helping you judge what good writing is. Even if the former function is useful, the latter ability doesn’t automatically follow.

For someone worried that AI will change the meaning of work or creation, saying “all you need is taste” isn’t a sufficient answer. It would help more to explain which tasks can be delegated and which judgments you must make yourself.

What Does the NYT’s Writing Preference Experiment Really Tell Us

On March 9, 2026, The New York Times published a quiz asking readers to compare short pieces of writing by humans and AI and pick which one they liked better. In a tally Kevin Roose shared the next day, roughly 86,000 people had taken part, and 54% of all choices went to the AI-written piece. He initially framed this as a share of participants, then corrected it to a share of total votes.

This result only tells us which of the given short passages people preferred. Since the respondents were self-selected quiz-takers, it’s hard to treat their choices as representative of all readers, and the quiz didn’t evaluate an entire novel’s structure or originality either.

What strikes me is that choosing which sentence reads better in the moment is a different exercise from judging the worth of a whole work. The fact that people liked a short excerpt doesn’t settle whether the process behind it—or the creator who made it—needs to be replaced.

Nor can we conclude that readers who picked the AI passage have worse taste. The quiz includes no comparison showing how the same person’s judgment has changed because of the digital environment. If anything, I see this as a case that forces us to ask what exactly we’re evaluating when we call writing “good.”

Oswarld’s Lens

What I want to hear more of in this discussion is “conviction” — the attitude of acting on your own judgment and taking responsibility for the outcome.

Advertising professional Artem Voronov made this same point in a February 2026 piece. He wrote that in the commentary he read from the 2024-2025 Cannes Lions Grand Prix jury chairs, what got emphasized wasn’t “taste” but bold ideas and executions that took real risks. This isn’t a systematic tally of all jury commentary — it’s an observation based on the cases he happened to read.

Apple’s “1984” ad can also be seen as a case of exactly this kind of judgment and execution. According to the recollections of Andy Hertzfeld, one of the Macintosh’s developers at the time, when outside board members pushed back hard against the ad, Apple told its agency to resell the airtime it had already bought. The agency, Chiat/Day, only managed to sell off 30 seconds of it, so Apple decided to run the ad in the remaining 60 seconds. The ad went on to get replayed repeatedly in the news and drew enormous attention.

In my own experience, too, the ability to choose what to make has mattered a great deal. But in business, “this is just my taste” isn’t a sufficient explanation on its own. You need to form a hypothesis about which customers will want it and why, explain your reasoning, and then look at the results of executing on it. I’d rather understand conviction not as ignoring dissenting opinions outright, but as the willingness to make the call once you have grounds you can point to and verify.

This same standard needs to apply when we use AI feedback. Anthropic’s 2023 research found a tendency toward “sycophancy” — agreeing with users’ views — across the five AI assistants it examined. In some cases, the assistants favored answers that pleased the user over answers that were factually accurate. The mere fact that an AI agreed with my thinking doesn’t mean that thinking has been verified as correct.

Closing

I agree that as AI takes over the drudgery of drafting, choice and judgment matter more than they used to. But if we don’t explain what good taste actually means, “good taste” easily becomes just a way of praising people whose taste happens to match our own.

So when I talk about business and products, I want to attach concrete reasoning to any claim of “good taste.” I should be able to say who it’s good for, what specifically makes it good, and what evidence would change my judgment.

In writing or art, beyond immediate preference, we can look at the creator’s experience, the context of the whole work, and its power to make us think long after we’ve encountered it. Depending on which standard you apply, the same piece can be judged very differently.

I’ll keep writing about taste going forward, but I don’t want that word to be where the explanation stops. I want to keep talking about the whole process—explaining why you chose something, actually doing it, and revising your thinking based on the results.

References & Further Reading

• Kyle Chayka, Why Tech Bros Are Now Obsessed with Taste, The New Yorker, March 18, 2026.
• Paul Graham, Taste for Makers, February 2002.
• Kevin Roose·Stuart A. Thompson, Who’s a Better Writer: A.I. or Humans?, The New York Times, March 9, 2026. Roose’s March 10 tally and correction.
• Pierre Bourdieu, Distinction, original French edition 1979. A book analyzing the relationship between taste and social background.
• Kyle Chayka, Filterworld, original English edition 2024.
• Artem, You’re Wrong About Taste, February 19, 2026.
• Andy Hertzfeld, 1984, Folklore.org. A recollection by a Macintosh development team member of the ad’s airing.
• Anthropic, Towards Understanding Sycophancy in Language Models, October 23, 2023.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. LLM: a large language model trained on massive amounts of text to generate sentences. It’s the underlying technology behind services like ChatGPT and Claude. ↩

2. Pierre Bourdieu (1930–2002): a French sociologist who studied the relationships among education, class, and cultural practice. ↩

3. Cultural capital: a concept describing resources—knowledge and attitudes acquired through education, cultural experiences, credentials, and the like—that can influence social status and opportunity. It’s also used to examine the conditions under which taste is formed. ↩