Taste is alpha, not a moat — AI absorbs and commoditizes individual judgment¶
Insight: According to Shrivu Shankar, taste functions as "alpha" (a temporary competitive edge) rather than a "moat" (a durable advantage). Individual judgment is "only valuable relative to what AI produces by default, and that default gets better on its own." By encoding taste into productive outputs, practitioners simultaneously train the systems that will eventually replicate that judgment without them.
Detail: Shankar maps taste degradation over time: 2024 — AI handled routine work, humans owned judgment; 2025 — frontier models matched human experts on roughly half of professional tasks; 2026 — AI-generated content dominates production (one-third of music uploads, majority developer adoption). Platforms like TikTok, Spotify, and YouTube aggregate millions of micro-signals to synthesize preference patterns at scale, making individual tastemakers unnecessary. The paradox: "your taste-informed outputs perform well, get clicked, shared, imitated, and that performance signal feeds back into the next generation of models."
Sources
- Shrivu Shankar — "Taste Is Not a Moat" (2026-02-16)