Lenny's Podcast: How Anthropic Builds AI Before the Future Arrives

AI products don’t emerge from models alone—they appear when product thinking, team culture, and timing all line up. In this summarized episode of Lenny's Podcast, host Lenny Rachitsky talks with Dianne Penn of Anthropic about how Claude evolved from coding autocomplete to long-form code generation, and how that insight shaped Opus 3 and Claude Code. The original episode is condensed from a full-length conversation into a quick listen that saves time while preserving the key ideas. Learn why Anthropic says “evals are the new PRDs,” how they turn vague AI failures into actionable testing, and why teams must “sweat the tokens as much as the pixels.” Penn also explains Labs, zero-to-one experimentation, red teaming, AI alignment, and why product leaders need to stay hands-on with AI models and customers. Listen now to get the key ideas in minutes.