How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)

Cat Wu describes AI-native product management as a race to shorten the distance between an idea and a user's hands. At Anthropic, feature timelines have moved from roughly six months to a month, a week, or sometimes a day. She says the PM role therefore emphasizes clear goals, fast experiments, research previews, and repeatable launch processes more than multi-quarter coordination. Product taste becomes more valuable as code gets cheaper: the scarce decision is what to build, for whom, and which experience is worth making.

Wu's partnership with Boris illustrates a less formal division of labor. Boris sets much of the product vision and technical direction, while Wu translates that vision into a near-term path and aligns marketing, sales, finance, capacity, and other functions so that a ready feature has no shipping blocker. Their responsibilities overlap heavily rather than following a rigid boundary. The team supports this speed with weekly metrics readouts and shared principles that explain key users, priorities, and trade-offs, allowing people to decide without waiting for a PM. PRDs remain useful for ambiguous features or infrastructure projects that need months, but they are not a default ceremony.

Anthropic's process, rather than access to a particularly powerful model, is Wu's main explanation for the shipping pace. The team tries to remove barriers and gives individuals permission to move an idea into the world in less than a week. Research previews lower the commitment required to learn from users, while a tight handoff between engineering, marketing, documentation, and developer relations lets a feature move from internal use to announcement quickly. She treats a source-code leak as a human process failure, not an individual moral failure, and says the response was to add safeguards.

Wu's account makes clear that the role is built around judgment rather than a fixed set of technical tasks. The interview identifies product taste as the scarce capability: deciding what to build as code becomes cheaper, while defining a clear user, problem, and use case narrows the possible solutions. Engineering knowledge helps a PM estimate trade-offs, but her broader argument is that the team needs people who can turn a vague goal into a fast, focused launch.