Key Moments
How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
Lenny's PodcastKey Moments
Anthropic ships AI features in days or hours, not months, by aggressively removing barriers and embracing research previews, challenging traditional PM roles.
Key Insights
Anthropic's shipping cadence has decreased from months to weeks to as little as one day by removing all barriers to shipping.
Product Managers (PMs) in AI need to prioritize speed and iterating quickly, launching features weekly, as code becomes cheaper to write.
Cloud Code utilizes a 'research preview' model to de-risk shipping early-stage ideas within a week or two.
Anthropic actively uses internal models like Mythos to improve shipping rates, though process and team expectations are primary drivers of speed.
The 'applied AI' team, acting as technical go-to-market, heavily utilizes both Cloud Code and Co-work for prototyping and customer engagement.
PMs are shifting to become orchestrators of AI capabilities, focusing on defining what to build, guiding models, and patching weaknesses.
Shipping at unprecedented speed
Anthropic has drastically accelerated its product release cycles, moving from shipping features over months to weeks and even down to a single day. This rapid cadence is achieved by systematically removing every barrier to shipping. The emphasis has shifted from multi-quarter roadmap alignment with partner teams to a focus on getting concepts into users' hands quickly, often within a week or even a day, through initiatives like 'concept corners' for rapid idea testing.
Redefining the Product Manager role in AI
The traditional Product Manager role is evolving rapidly in the AI space. With code becoming much cheaper and faster to produce, the critical skill is now 'product taste' – deciding *what* to build and *how*. PMs must excel at defining clear goals for their products, such as identifying 'professional developers' as key users and focusing on specific problems like 'permission prompt fatigue' to achieve 'zero permission prompts safely.' This clarity helps narrow down the vast possibilities offered by AI models and guides development effectively.
Iterative shipping strategies for rapid feedback
To maintain speed and gather feedback, Cloud Code employs a 'research preview' strategy for nearly all its features. This approach clearly labels early-stage products as experimental, reducing the commitment required for shipping. This allows for rapid iteration – getting a feature out in a week or two – and then incorporating user feedback. This also necessitates a tight, pre-defined process between engineering, marketing, and documentation teams to ensure smooth handoffs and timely product announcements, with dedicated roles like PMM and DevRel jumping in immediately when a feature dogfooded internally is ready.
The evolving definition of Product Manager skills
In the realm of AI product development, Product Managers (PMs) are increasingly expected to possess a blend of technical understanding and strong product intuition. The ability to define product direction a month or more out is challenging due to the rapid evolution of AI models and user behavior. The critical skill lies in guiding current model capabilities to elicit maximum effectiveness, helping users navigate the 'golden path' by leveraging model strengths and mitigating weaknesses. This requires a deep understanding of model limitations and an ability to adapt forecasts as AI capabilities accelerate.
Leveraging internal models and robust processes
While Anthropic benefits from access to frontier models like Mythos, the accelerated shipping pace is primarily attributed to internal processes and team expectations. The company prioritizes removing barriers and empowering every team member to take an idea from conception to release. This is reinforced by rigorous weekly metrics readouts that ensure the entire team understands business goals and performance, alongside a set of team principles that align everyone on key users, problems, and acceptable trade-offs, enabling autonomous decision-making.
Navigating the chaos with a mission-driven focus
Anthropic's exceptional growth and rapid development are underpinned by a strong, unifying mission: bringing safe AGI to humanity. This shared goal acts as a compass, enabling swift and unified decision-making by prioritizing the mission over individual product line goals. This focus means teams are willing to make sacrifices that might hinder their specific KPIs if it serves the broader Anthropic mission, demonstrating a high degree of alignment and collective purpose. This mission-centric approach is crucial for navigating the inherent chaos of rapid AI development.
The critical role of 'product taste' and adaptability
As AI capabilities advance, the ability to decide *what* to build becomes paramount. This 'product taste' is the most valuable skill, guiding the user experience and ensuring the right features are prioritized. While an engineering background can be beneficial for understanding development effort, the core skill remains discerning which ideas are worth pursuing and how to best implement them. The AI landscape is constantly shifting, requiring individuals to adopt a first-principles approach, continuously learn, and adapt to new technological advancements to remain effective.
Empowering users and internal teams with AI tools
Anthropic is building tools like Cloud Code and Co-work to empower users and internal teams. Cloud Code, available via CLI and desktop, focuses on code-related tasks and visual development, while Co-work handles non-code outputs like presentations and documents. The company emphasizes connecting data sources to Co-work for better context and uses AI for tasks like generating slide decks and summarizing customer information. Internally, custom apps built with Cloud Code are widespread, automating repetitive tasks for sales and other departments, thereby increasing efficiency and focusing on higher-impact activities.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●People Referenced
Common Questions
Product Managers in AI companies now need to prioritize rapid iteration and getting features to users quickly, often within days or even hours. The focus shifts from long-term roadmaps and coordination to reducing time-to-market and identifying core features that work out-of-the-box.
Topics
Mentioned in this video
The company where Cat Wu works as Head of Product for Cloud Code and Co-work. Known for its rapid pace of shipping features and its focus on developing AI technology.
Mentioned as one of the companies powered by Work OS, and also as a competitor that Anthropic was initially behind.
Mentioned as one of the companies powered by Work OS.
Used as an analogy for Work OS, comparing it to 'Stripe for enterprise features'.
Mentioned as a platform where Anthropic receives tens of thousands of issues and requests for features.
Mentioned as a tool where design systems or slide formats can be saved and accessed by AI.
A CRM that is integrated with custom-built Anthropic apps to pull customer context for generating tailored sales decks.
A platform integrated with Anthropic's custom sales tools to pull customer context for personalized decks.
A supporting sponsor that automates compliance and risk management for companies, helping them earn and prove trust.
Mentioned as one of the companies that uses Vanta for compliance and trust.
Mentioned as one of the companies that uses Vanta for compliance and trust.
Mentioned as one of the companies that uses Vanta for compliance and trust.
Mentioned as one of the companies that uses Vanta for compliance and trust.
Mentioned as a competitor to Whimo, suggesting that Whimo's premium pricing is justified by its superior user experience.
A ride-sharing service Cat Wu uses daily and calls a '10x product' for providing a productive and less pressured commute experience.
Head of Product for Cloud Code and Co-work at Anthropic. She discusses the evolving role of PMs, rapid product iteration, and the company's mission.
Mentioned as a past guest on the podcast who discussed the importance of a product's character and constitution.
The subject of the movie 'Free Solo', known for climbing El Capitan without a harness. Cat Wu admires his achievement.
A presenting sponsor of the season, it's a platform designed for B2B SaaS companies to integrate enterprise features like SSO, audit logs, and RBAC.
Mentioned as one of the companies powered by Work OS.
Mentioned as one of the companies powered by Work OS.
An advanced AI model developed by Anthropic that is still in preview due to its power. It is not the sole reason for Anthropic's fast shipping pace.
An AI-powered coding tool developed by Anthropic, used extensively by the team. Its source code was recently leaked due to human error.
The AI model developed by Anthropic. Mentioned in the context of its source code leaking and its usage in various Anthropic products like Cloud Code and Co-work.
A term used in relation to using Claude subscriptions with third-party products, which Anthropic has restricted, causing some user upset.
Described as the 'core OS' of Anthropic, used extensively for communication and integrating AI tools. It's praised for its hackability and reliability.
A version of Anthropic's models (Sonnet 4.6) that enabled advanced code review capabilities.
A platform mentioned in the context of customers using it instead of Anthropic's Cloud for Enterprise.
A platform mentioned in the context of customers using it instead of Anthropic's Cloud for Enterprise.
A version of Anthropic's models (Opus 45 and 46) that enabled advanced code review capabilities.
A TV show Cat Wu enjoys for its focus on the purity of pursuing a singular engineering goal, similar to the dedication seen in her work.
A movie about Alex Honnold climbing El Capitan without a harness, which Cat Wu admires for its depiction of pure achievement and mental focus under extreme pressure.
The name is referenced humorously when discussing the difficulty of tracking all the rapid product launches.
A book recommended by Cat Wu that examines past technological revolutions (Industrial, Computer) and their impact on workers, offering lessons for current transitions.
A book recommended by Cat Wu, focusing on economic development and the policies that contribute to successful economies.
A short story collection recommended by Cat Wu, touching on themes of coming of age and self-discovery, with a connection to AI.
More from Lenny's Podcast
View all 25 summaries
96 minWhy half of product managers are in trouble | Nikhyl Singhal (Meta, Google)
83 minHard truths about building in the AI era | Keith Rabois (Khosla Ventures)
113 minAnthropic’s $1B to $19B growth run: how Claude became the fastest-growing AI product in history
100 minAn AI state of the union: We’ve passed the inflection point & dark factories are coming
Found this useful? Build your knowledge library
Get AI-powered summaries of any YouTube video, podcast, or article in seconds. Save them to your personal pods and access them anytime.
Get Started Free