Why We Like Things

An Asterisk Magazine essay tracing what "taste" actually is, why Silicon Valley has made it the catch-all term for whatever survives AI automation, and how it is cultivated — through the author's own life, the history of 19th-century British design schools and the Bauhaus, and a survey of theory from Bourdieu and Girard to Kant and Arendt. The through-line: taste is not an innate, private gift but a trainable, essentially social faculty, and explaining our tastes to one another is the real skill.

The AI-Era Thesis: Taste Is What's Left

The essay's premise is that taste has become the tech industry's word for whatever automation hasn't yet absorbed. As generative AI shifts knowledge work from production to selection and curation, taste becomes the competitive edge:

"In the AI age," the Y Combinator cofounder Paul Graham tweeted in February, "taste will become even more important. When anyone can make anything, the big differentiator is what you choose to make."

"Taste," [Greg Brockman] wrote, "is a new core skill."

"I have no technical ability… The confidence I have in my taste." — Rick Rubin, explaining what he's paid for.

The author's sharpest move is to note that everyone who invokes taste "fail[s] to explain what it is" — and to hand the essay's tension to a skeptic. The literary scholar Leif Weatherby argues that LLMs already incorporate "judgments of taste" from literary-critical texts to Likert-scale quality ratings "in a direct and infrastructural way," so taste as "the last preserve of the human" is a misnomer:

The idea that "this is the last preserve of the human, the ability to make judgments… it's anxiety on the part of industry leaders…because they're deskilling, right? They're saying, If Codex writes the code, what's your job?"

This is the essay's load-bearing insight, and it cuts both ways: taste-as-human-remainder is partly a cope, an attempt to locate irreplaceability somewhere when the actual work is commoditizing.

A Short History: Taste as a National Project

To explain what taste is, the author reaches past Silicon Valley to the last time automation triggered the same anxiety. In 19th-century Britain, mass production produced "more goods" but "more mediocrity," and the luxury market was dominated by France. The statesman William Ewart argued the gap was "the reverence that prevailed for the Arts in all classes," and his 1835 motion led to publicly funded design schools and the V&A.

The point of the parallel is that taste was once understood as a public, institutional, national responsibility — cultivable "through educational and cultural institutions" — not a private, innate gift. The design schools embraced the machine; Ruskin and William Morris's Arts and Crafts movement resisted it. Both still treated taste as something a society builds, not something a founder is born with.

The Bauhaus: Taste as a Trainable, Domain-Independent Skill

The essay's most useful historical material is the Bauhaus. Johannes Itten's Vorkurs replaced medium-specific first-year training (figure drawing, copying ornamentation) with abstract concepts — line, form, color, contrast, material — plus breathing exercises for "a more instinctual, intuitive way of working." Josef Albers carried this to Black Mountain College, where he taught Cy Twombly, Ruth Asawa, and Robert Rauschenberg and explicitly rejected "that talent and an aptitude for art were inherent gifts and prerequisites."

The key insight, per the author: the foundation year treated aesthetic perception as a domain-independent skill, refined through study of existing works and through critique — "coming from the [other] students and then from the teacher" — which also "diminish[ed] the tendency to overestimate one's own work."

The Theory: Bourdieu, Girard, Arendt

Three accounts of taste compete in the essay:

• Bourdieu — taste as class strategy: fluency in elite codes accumulates cultural capital. The author finds this "oppressively jaded," and quotes Marco Roth's lament that sociology "easily obscures" the "private and idiosyncratic" reasons we like what we like.
• Girard — mimetic desire: we desire what Others desire; what looks like personal taste is "petty rivalry." The author notes, with an acid aside, that "those who cite Girard are often imitating the tastes of his former student, Peter Thiel."
• Arendt/Kant — the essay's choice. Kant's Critique of Judgment defines aesthetic judgment as "enlarged mentality": you interpret your private reaction alongside the imagined reactions of others, engaged in "an anticipated communication."

"Taste judgments do not compel" agreement… they are "persuasive; the judging person — as Kant says quite beautifully — can only 'woo the consent of everyone else.'"

"It is as though taste decides not only how the world is to look, but also who belongs together in it."

For Arendt, this makes judgment "one of the fundamental abilities of man as a political being" — taste is how we participate in a shared world, not merely how we curate a feed.

The Tyranny of Good Taste

The essay ends by insisting taste is never settled. Once a preference becomes hegemonic, "a backlash usually results"; universally held tastes are "boring." Sontag's camp ("the good taste of bad taste") and John Waters's "tyranny of good taste" are offered as permission slips, and Carl Wilson supplies the closing norm:

"The goal," as the critic Carl Wilson wrote, "is not that we all end up with the same taste."

The final claim: taste helps with minor decisions (what to listen to, "when to use an em dash… or, for the advanced aesthetes, an en") and major ones (what to fund, what to ship), but its higher purpose is that "our tastes are how we understand ourselves — and communicating them is how we understand each other."

Key Themes

• #concept Taste as the last preserve of the human — the AI-age claim that selection/curation is what automation can't take, and Weatherby's rejoinder that this is deskilling anxiety, since models already encode judgments of taste.
• #concept Enlarged mentality (Kant via Arendt) — taste as judging your private reaction alongside the imagined reactions of others; persuasive, not coercive; a political faculty, not a private preference.
• #concept Taste as trainable craft — the Bauhaus Vorkurs / Albers pedagogy: domain-independent aesthetic perception built from study of prior work plus peer critique. Directly parallel to Product-Minded Engineers in an AI-Native World.
• #pattern Pre-training / fine-tuning / RLHF as the structure of taste — the author's own Tumblr-scrolling adolescence is "a large dataset of images" refined by "feedback from peers and mentors," which raises the uncomfortable question: if that's all taste is, what distinguishes a human from a model?
• #person Rick Rubin, Josef Albers, Hannah Arendt, Paul Graham — the essay's recurring authorities: Rubin as taste-confidence guru, Albers as democratic pedagogue, Arendt as the theorist who makes taste political, Graham as the investor framing taste as the differentiator.

Critical Analysis

The essay's greatest strength is also its central evasion. The author's own autobiography — "pre-training, fine-tuning, and RLHF" — is a machine-learning account of taste formation, and the Bauhaus material lands on the same conclusion: taste is exposure plus critique, a trainable skill. That account has no principled place for human irreplaceability; a model with more images and more feedback should, by the essay's own logic, develop more taste. The Arendt answer — that taste is irreducibly relational, a "wooing of consent" — is what the author reaches for, but it's asserted, not derived. The essay wants to have both: taste as trainable and taste as the thing that makes us irreplaceable. The tension is real and unresolved, and it's the most honest thing about the piece.

The dismissal of Bourdieu and Girard is also too quick. "Oppressively jaded" is a feeling, not a refutation, and Arendt's "enlarged mentality" can be read as a friendlier name for exactly the social conditioning Bourdieu described — we form our judgments in anticipation of some audience, and which audience we anticipate is largely a matter of class position. The essay gestures at this with its own admission that "I'd be lying if my preferences weren't shaped by, and sometimes outsourced to, external curators," then moves on.

The historical parallel is the most transferable contribution, and the essay under-uses it. Nineteenth-century Britain answered automation anxiety by building institutions — design schools, museums, a national curriculum of taste. The essay never asks what the modern equivalent would be: if taste is genuinely the "new core skill," which institutions are training it now? The implicit answer is alarming — mostly nobody, and increasingly the models themselves.

Finally, the essay is self-consciously an example of its own thesis: it "woos consent" rather than proves, and its persuasiveness depends on whether you trust the author's taste. That's the point, and it's also the limit. As an argument it is circular; as a demonstration of taste-as-persuasion, it works.

This lands directly in conversation with Product-Minded Engineers in an AI-Native World (taste as trainable craft, and the open question of what's left for the human), Slop Score (a diagnostic for developing taste that treats style as measurable), The Education of the Broligarchy (Smith's diagnosis that tech elites never develop taste, and Girard's place in their canon), and The Usefulness of Useless Knowledge (Flexner's case that apparently "wasted" training is exactly what produces judgment — the June Huh "poetic detour" in miniature).

Sources: raw/why-we-like-things, Why We Like Things Last updated: 2026-08-25