Cat Wu, Head of Product for Claude Code at Anthropic, has watched her team's shipping cadence compress from months to weeks to days. She interviews hundreds of PMs trying to break into AI and has a clear view of who survives the transition and who doesn't. The answer is not who knows the most about models. It's who defaults to action.
The original piece is worth reading for three specific arguments. First, Wu's case for building products that don't fully work yet, positioning them to capitalize the moment the next model closes the capability gap. Second, her identification of an underrated PM skill: prompting the model to introspect on its own errors, which surfaces failure modes faster than any manual review. Third, her claim that Claude's personality is not a feature, it is a core driver of product success, and that Anthropic's mission alignment removes the internal friction that kills velocity at larger organizations.
The through-line is a single operating principle Wu calls 'just do things.' For PMs accustomed to consensus cycles and requirements documents, this is a harder shift than learning to write prompts. The full conversation runs on YouTube, Spotify, and Apple Podcasts. If you manage product at a company still measuring release cycles in quarters, this is the competitive gap you are currently losing.
[READ ORIGINAL →]