AI can generate interfaces, brands, campaigns, and entire products in seconds. The barriers to making things have collapsed. What remains is the question David Hume posed in 1757 and Pierre Bourdieu complicated in 1979: how do you know if it is any good? This piece argues that the answer is taste, not as aesthetic preference, but as accumulated, contextual judgement. The kind built through exposure, comparison, and honest reckoning with failure.
The argument turns on Graham Wallas's 1926 four-stage model of creative work: Preparation, Incubation, Illumination, Verification. The critical stage is incubation, the slow, largely unconscious process where raw input becomes genuine discernment. AI collapses this stage almost entirely. When Suno CEO Mikey Shulman claimed in 2024 that most people find making music 'not really enjoyable,' he revealed a foundational assumption in the industry: that creative difficulty is a bug to be fixed. The article treats this as the clearest possible statement of what goes wrong when tools are built to skip incubation entirely.
The cerulean monologue from The Devil Wears Prada is used not as decoration but as a precise analogy. AI can generate a thousand cerulean variations. It cannot tell you why cerulean matters here, now, and for this audience. That is the full case worth reading: not just the conclusion that taste matters, but the specific philosophical lineage, the Wallas model applied to AI workflows, and the Suno case study that shows what it costs when an industry mistakes the removal of difficulty for democratisation.
[READ ORIGINAL →]