How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
Cat Wu — Head of Product for Claude Code + cowork at Anthropic — gives the inside view of how Anthropic's product team ships at unprecedented pace. Feature timelines collapsed from 6 months → 1 month → sometimes 1 day. Most external PM candidates she interviews are 'approaching it incorrectly.' The right framing: ruthlessly remove process, brand everything as Research Preview, ship to dogfood by week's end. The scarce hire isn't more PMs OR more engineers — it's people with product taste.
Key points
- Org structure with Boris (tech lead, sets 3-6 month vision = 'AGI-pilled' product). Cat owns the path-from-now-to-there + cross-functional alignment + unblocking shipping. ~80% mind-meld; 20% split by who cares more.
- Timeline collapse: feature cycles from 6 months → 1 month → sometimes a single week or day. Implication for PMs — much less time on multi-quarter roadmap alignment, much more time on 'fastest path from idea to user's hands.'
- Most external PM candidates approach the role incorrectly. They optimise for coordination across long horizons; the actual job is shortening the idea-to-launch loop and defining 'must-work-out-of-the-box' tasks for the product.
- Three speed levers Anthropic uses: (1) **Set clear goals.** LLMs are so general that ambiguity in 'who, what, why' is the biggest cost. Cat's example: 'professional developers, permission-prompt fatigue, get to zero permission prompts safely.' (2) **Repeatable shipping process.** Almost everything ships as Research Preview — explicit framing 'might not be supported forever.' Lowers shipping commitment. (3) **Cross-functional framework so engineers don't need PM permission.** Evergreen launch room → docs + PMM + DevRel turn around an announcement next-day.
- PRDs are not dead — used selectively for ambiguous features and infrastructure-heavy projects. Otherwise replaced by weekly metrics readouts + a published 'team principles' doc that lets anyone make decisions independently.
- On Anthropic's pace: Mythos has accelerated things 'a little' but not most of it. The bulk is process discipline + the expectation that every person on the team can ship in <1 week.
- Claude code source-code leak: 'human error.' A human working with Claude to write a PR slipped through two layers of human review. Anthropic hardened processes; the engineer responsible is still there. 'The most important thing is to learn from it.'
- On OpenClaude subscription restriction: Anthropic prioritised first-party products + API. Compute is too scarce to subsidise third-party usage patterns. 'Businesses are trying to be profitable here.'
- Anthropic PM org: ~30-40 PMs across Research, CDP, Claude Code/cowork, Enterprise, Growth. Research PM team feeds customer signal into research priorities + shepherds model launches.
- Cat's PM-future answer (third in this issue's thread): 'all the roles are merging.' Engineers do PM work, designers ship code, PMs do engineering. Claude Code team's specific bias: hire **engineers with great product taste** rather than more PMs. 'Product taste is still rare and we'll hire anyone who has it.'
- AGI-pilled framing: 'It's very easy to build the product for the super-AGI strong model. The hard thing is figuring out, for the current model, how to elicit maximum capability.' Calibration between visionary and present is the actual core skill.
Notable quotes
The timelines for a lot of our product features have gone down from six months to one month and sometimes to one day.
I've never seen anything like the pace folks at Anthropic are shipping at.
We want to remove every single barrier to shipping things. We want every person on the team to feel empowered to take their idea to in-the-world in less than a week.
It is very hard to be the right amount of AGI-pilled.
Engineers with great product taste — many can go end-to-end from a Twitter user-feedback to a shipped product by the end of the week with almost no PM involvement.
Themes
- How Anthropic ships in a day what other companies ship in a quarter
- AGI-pilled calibration as the core PM skill
- Engineers with product taste over more PMs
- Inside the OpenClaude subscription decision
Mentioned
People
Ideas
- AGI-pilled vs present-model-pilled calibration
- Timeline collapse 6mo → 1mo → 1day
- Research Preview as default branding
- Evergreen launch room
- Engineers with product taste as the scarce hire
- Eliminate process, not add it
- Team principles doc as decision-enabling artefact
- Roles merging (eng/PM/design)
- Source-code-leak as human-process failure
- Subscription compute as a finite resource