Dianne Penn, Anthropic’s first technical PM and current head of product for the AI research and labs teams, discusses the company's journey from underdog to leader in AI, the importance of first-principles thinking and hands-on experimentation, and how product management is evolving. She emphasizes that success in AI requires sweating tokens as much as pixels, using evals as the new PRDs, and maintaining a culture of adaptability and joy in discovery.
Summarized by Podsumo
Anthropic's culture of bottom-up experimentation and mission-alignment was key to its rapid rise from a startup with 5 engineers to a leading AI company.
Product managers must now write evals (test cases for model behavior) instead of traditional PRDs to drive user value in an AI context.
Token maxing—spending heavily on tokens to explore capabilities—is a way to live in the future and shape product strategy.
Successful PMs combine deep hands-on use of AI with first-principles thinking, and must stay close to details like transcripts and user feedback.
AI's 'jagged edge' means capabilities emerge unpredictably, requiring ongoing discovery and adaptation rather than rigid roadmaps.
"If you are willing to spend $100,000 a year right now on tokens, you are living the way somebody in 2028 is gonna live."
— Dianne Penn
"Evals are the new PRDs."
— Dianne Penn
"In order to be good managers of teams and PMs working with this technology, you have to be really hands-on yourself and have shipped with it."
— Dianne Penn