Taste is the capacity to distinguish adequate from excellent within a field of aesthetic objects. Bourdieu demonstrated that taste is socially constituted — built through exposure, practice, and calibration against a ground of contrast. The wine student who has tasted only excellent wine cannot identify excellence. The architecture student who has studied only beautiful buildings cannot perceive beauty. Taste requires the experience of gradation: mediocre, competent, good, extraordinary. AI compresses this spectrum. The floor of AI production is remarkably high — code works, prose is fluent, analysis is structured. The ceiling is characteristically modest. The range between worst and best narrows. The result: a production environment where adequacy is ambient, everything is pretty good, and the distance taste exists to perceive has become so narrow that perception itself becomes difficult.
The crisis is not a crisis of supply — there is no shortage of aesthetic objects to evaluate. It is a crisis of demand. Demand for taste presupposes a gap between what exists and what should exist. When the gap is wide, demand for taste is self-evident. Someone must separate wheat from chaff. When the gap narrows — when chaff has been algorithmically eliminated and what remains is undifferentiated competent wheat — demand for taste becomes obscure. Why discriminate when everything is acceptable? The answer is that the difference between adequate and excellent is qualitative, not quantitative. Adequate code solves the problem. Excellent code teaches the reader something about the nature of the problem. The qualitative difference is invisible to metrics measuring volume, speed, competence.
Taste develops through exposure to the full spectrum including failure. The developer who reads only AI-generated code — uniformly competent, narrowly ranged — develops different perceptive capacity than the developer who reads human code in all its variability: brilliant misfires, instructive failures, terrible choices that reveal what not to do. The second developer builds taste. The first builds tolerance for adequacy. And tolerance for adequacy, when adequate becomes the ambient standard, is not a vice but an adaptation — the subject learning to inhabit a production environment where excellence is structurally rare and adequacy is the stable norm.
The practical crisis is that judgment — Segal's identified scarcity in the AI age — is taste applied to decisions. Judgment distinguishes what should be built from what can be built. This distinction requires perceptive capacity to evaluate quality differences too fine for metrics to capture. The builder whose taste was developed in pre-AI environments possesses this capacity. The builder whose taste develops entirely within AI-compressed spectrum may not. The developmental gap is not visible in outputs — both builders produce adequate work. It is visible in trajectories: over years, one builder's judgment deepens while the other's remains at adequacy's ceiling, unable to perceive the difference between pretty good and genuinely good because her perceptual apparatus was never exposed to the contrast required to build that discrimination.
Ngai's framework suggests the prescription: deliberate exposure to failure. The architecture school that shows students failed buildings. The writing workshop that analyzes bad prose. The code review that examines catastrophic implementations. Each is aesthetic education through negative example — calibration against the low end of the spectrum. AI production cultures systematically eliminate this exposure. The failed attempt is not celebrated, not examined, not preserved for pedagogical use. It is corrected, deleted, disappeared. The smooth's ambient pressure converts failure from developmental resource into inefficiency to be eliminated. The result: a generation of practitioners whose taste develops in a compressed range, producing work that is reliably adequate and structurally incapable of excellence because excellence requires perceptive discrimination the compressed range cannot build.
Taste as philosophical category was established by Kant, refined by Bourdieu as social instrument. Ngai's contribution is recognizing that taste is also developmental capacity — and that the conditions under which it develops are historically specific, technologically mediated, and currently under threat. The AI moment is not the first compression of the aesthetic spectrum — photography compressed painting, recorded music compressed live performance, each time raising floors and narrowing ranges. But AI's compression is the first to operate in the domain of knowledge work itself, where taste is not ornament but the primary scarce human input.
Taste requires contrast. Perceptive discrimination develops through exposure to the full spectrum from failure to excellence.
AI compresses the spectrum. The floor rises (code works, prose is fluent), the ceiling holds, the range between them narrows dramatically.
Adequacy becomes ambient. When everything is pretty good, taste loses its field of contrast — discrimination becomes difficult.
The crisis is developmental. Not visible in current outputs but in future capacity — practitioners whose taste develops in compressed range cannot perceive quality differences the range cannot build.
Deliberate exposure to failure. Aesthetic resistance requires seeking negative examples, studying instructive errors, calibrating against the low end of the spectrum.