Where this was said
Taste Labs: Can You Train Taste Into AI?
At 1:14:10 · chapter starts 1:09:35
Andrew's second confusing AI company is Taste Labs, which has positioned itself as ending AI design slop by curating high-quality training data to give AI models genuine aesthetic taste. Andrew's critique: you can't train taste without just making the curated aesthetic the new slop. David agrees the design pendulum metaphor is apt — once everything looks uniformly good, 'good' becomes bland, and the cycle resets toward intentional ugliness as the new hip. [1] — Andrew Manganelli "Taste Labs raised money on the premise of giving AI models 'taste' to stop AI design from looking generic. Andrew and David were unconvince…" 1:09:35 The hosts also riff on the Chobani Serif font as a case study: a beautiful typeface originally commissioned by a yogurt brand that has now been adopted by so many brands it risks becoming cliché. Anthropic's soft-serif, newsprint-influenced logo gets name-dropped as another example of the warm-AI-branding trend that risks becoming the next design slop. Andrew's closing verdict: 'We're not ending AI slop. We're just making different slop with a little garnish on top — and then the garnish gets wilted.'
Taste Labs raised money on the premise of giving AI models 'taste' to stop AI design from looking generic. Andrew and David were unconvinced: curating better training data just makes that style the next AI slop. David added that design trends are a pendulum — the better AI design gets uniformly, the faster it starts to look dated.
