
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Scott Alexander splits “taste” into at least eight different things — he argues people lazily mash together sensory delight, novelty, pattern language, context, literal understanding, fashion, ideology, and transformation, then pretend they’re discussing one clean concept.
His core move is radical decoupling: judge beauty like medicine, not like branding — using a restaurant-critic parable and invoking RCT logic, placebo effects, and predictive coding, he says aesthetics should try much harder to isolate the thing itself from reputation, provenance, and social context.
The Michelangelo-vs.-boomer-from-Ohio thought experiment is the essay’s pressure test — if a sculpture genuinely fills you with awe before you learn it was allegedly mass-produced in 1995, Scott asks why that feeling should become “wrong” once provenance changes, unless you care more about status than art.
He says much modern criticism protects a narrow equilibrium instead of escaping it — citing Freddy Deboer and Erik Hoel, he argues contemporary literary norms reward defensive, low-attack-surface fiction and then use that very conformity to make anything more ambitious feel cringe or unserious.
Frank Lantz’s strongest reply gets taken seriously, then partially rejected — Scott agrees art can be a recursive historical conversation involving Warhol, Duchamp, Jeff Koons, and alignment-adjacent questions, but says that project should not get to crowd out plain old making beautiful things.
His blunt bottom line: novelty without beauty has become a dead-end game — once you’re on the 500th “shark in formaldehyde” or the joke-version “dolphin pancreas baby,” the conversation isn’t deepening; it’s just repeating the same transgression with worse material.
Scott opens by revisiting responses from Ozy, Frank Lantz, and others, saying they helped him realize why arguments about art go nowhere: people use “taste” to mean a pile of different phenomena. He lays out eight of them — from raw sensory delight and innovation to ideological messaging and transformative power — then says his “bold stand” is simply that conflating all these together makes everyone think less clearly.
Young Scott imagined restaurant critics as quasi-scientists doing blind, repeated tastings, controlling for reputation, plating, and ambience the way medicine controls for bias. Instead he found essays about lighting, chef backstory, and Sardinia, which to him felt like surrendering to branding and placebo. The point isn’t that ambience is fake; it’s that if nobody isolates the steak itself, criticism can only reinforce conventional wisdom.
He pushes the argument with a museum thought experiment: a Renaissance-style sculpture awes you, then a placard says it was made in 1995 by a boomer from Ohio for rich dentists’ McMansions. If your awe instantly curdles into embarrassment, Scott thinks that reveals an uncomfortable dependence on provenance rather than the actual experience of beauty. He compares this to doctors choosing Ailify versus Thorazine based on a “cool story” instead of efficacy — absurd in medicine, but common in art.
Scott gets personal here, saying he has no visual-art taste but some poems really do “transform” him, especially G.K. Chesterton. If a hundred lost Chesterton poems turned up, it would be “Christmas, Hanukkah, and my birthday all at once”; if they were later exposed as forgeries yet remained just as good, he says the honest response would be an aesthetic crisis, not indifference. His acid test is simple: if you still prefer the original only because he was first, maybe you don’t like poetry so much as affiliation with famous poets.
Pulling in Freddy Deboer and Erik Hoel, Scott says contemporary literary fiction has become cramped, polished, minimal, autofictional, socially cautious, and obsessed with shrinking its “attack surface.” He mocks the idea that today’s narrow style rules are eternal truths of prose, contrasting them with Homer and the whole premodern tradition. His Lincoln paraphrase lands hard: critics first create the conditions that make ambitious work feel bizarre, then excuse the bizarre feeling by saying the work is too self-consciously strange.
Scott gives Lantz the strongest steelman: art is not merely a 600x400 JPEG but a situated intervention in history, discourse, galleries, markets, and audience expectations. That frame makes sense of Warhol, Duchamp, Koons, and the recursive game between genuine expression, social ritual, philosophy, and spectacle. Scott clearly respects the ambition here, especially when Lantz links the restlessness of art to rationalist concerns like formal systems, collective traps, value drift, and alignment.
Lantz’s example, Paul Klee’s Angelus Novus, lets Scott showcase Walter Benjamin’s famous commentary about the “angel of history” blown helplessly into the future by the storm called progress. Scott loves Benjamin’s prose and admits it nearly converts him — until you actually look up the painting and feel, in his words, like you’ve been pranked. That becomes his key complaint: maybe context-rich commentary is valuable, but if the object itself doesn’t deliver, then letting criticism borrow grandeur from the conversation feels parasitic.
The ending is Scott at full exasperated comic force. Even granting that art is historically informed commentary on art, he says too much modern work says nothing new — just the 500th version of “might this transgressive ugly thing in some sense be art?” with sharks in formaldehyde, concrete cubes, and the deliberately grotesque “dolphin pancreas baby.” His final standard is brutally simple: yes, Gaudí was breaking tradition too — but “he was good at it and you are bad,” and if you can’t innovate beautifully, better to stay within the old forms until a real genius arrives.
Share
Keep Reading
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.

Playbook
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.

Playbook
OpenAI shipped /goal for the Codex CLI. It turns a prompt into a persisted, self-continuing contract.