Builders versus product managers in AI

Tony Fadell: How to build real taste (and why AI makes it matter more)

Topic 9 of 16

Published Jun 7, 2026 · Clip 48:40–58:02 (9:22)

Tony describes product management as the discipline that sits between all functions — marketing, sales, distribution, engineering, manufacturing, support — stitching them together, and argues AI won't wash those roles away. He cites the leaked Claude source code: despite claims that 90–100% of Anthropic's code is AI-written, real engineers found the main loop brittle and unreadable. AI-generated code without architects, optimizers, security reviews, and expert segmentation accumulates technical debt — short-term gain for very long-term loss — like fast fashion versus luxury craft, and a real company can't be throwaway. Use AI agents for prototypes and scoped sub-segments with locked-in architecture while humans stay in the loop. Because building is now easy, the well-thought-through stands out: he argues you couldn't vibe-code Flighty v1 ('luxury software'), though a v2 knockoff might be possible, and he evangelizes Flighty to everyone for its care and craft.