Anthropic's first technical PM reveals how Claude went from underdog to fastest-growing AI company, shipping frontier models while pioneering eval-driven development. Discover why safety and alignment make Claude better, not worse—and how to build products in an exponential AI era where emerging capabilities surprise even their creators. Evals are replacing PRDs as the core PM artifact at Anthropic; they define user feedback rigorously enough for researchers to act on, not just as documentation but as test-driven development for AI products.