Apr 23, 2026 · 1h 25m · lennys-podcast
How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Anthropic Head of Product Cat Wu shares how her team achieves industry-leading shipping velocity, navigates the evolving role of AI product management, and designs agentic tools like Claude Code and Cowork.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Lenny holds 28.3% of the talking time here. How this is scored →
Lenny as informed peer 3.8 Guest teaching 5.6 Guest disagreement 1.3 Lenny pushing back 1.2
speaking balance: gold is Lenny, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 31:10 Reframing focus vs mission sacrifices
Cat reframes Lenny's premise that Anthropic merely ignores side distractions, clarifying that true mission alignment requires teams to sacrifice their own product OKRs and local success for broader company safety priorities.
Hardest push from Lenny ▶ 19:37 Probing if Mythos makes engineering skills obsoleteLenny pushes Cat on whether imminent model releases like Mythos will eliminate the need for PMs to understand engineering complexity, prompting Cat to clarify how shifting capabilities impact necessary skill sets.
Biggest teaching moment ▶ 1:01:03 Deprecating prompt crutches across model jumpsCat provides an illuminating breakdown of how Anthropic systematically strips out prompt crutches and UI tools like the to-do list as frontier models naturally internalize thorough multi-step execution.
Lenny holds their own ▶ 1:02:44 Recognizing models eating software harnessesLenny demonstrates sharp domain expertise by synthesizing Cat's explanation into the broader industry insight that smarter models inevitably absorb surrounding application harnesses.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Lenny as informed peer | Guest teaching | Guest disagreement | Lenny pushing back | Why |
|---|---|---|---|---|---|---|
| Defining Product Management Alongside Boris at Anthropic | 4 | 5 | 1 | 1 | Lenny frames the working relationship between Cat and Boris, noting Boris's previous podcast episode. Cat explains their blurred division of labor, detailing how she operationalizes Boris's visionary direction and handles cross-functional coordination. | |
| The Shift from Traditional Planning to Rapid AI Shipping | 3 | 7 | 2 | 1 | Cat educates Lenny on why traditional PM planning cadences are obsolete in frontier AI development. She explains how shipping cycles collapsed from quarters to days and why alignment-heavy roadmapping slows execution. | |
| Anthropic's Playbook for High-Velocity Product Execution | 3 | 6 | 1 | 1 | Lenny asks how Anthropic structures rapid execution, and Cat breaks down their three-part playbook: establishing explicit constraints, launching via Research Preview, and running streamlined cross-functional launch channels. | |
| Metrics, Principles, and the Evolution of PRDs | 4 | 5 | 2 | 2 | Lenny probes whether traditional PRDs are dead. Cat clarifies that while weekly metric readouts and operating principles govern daily decisions, one-pagers and PRDs are still retained for ambiguous or infrastructural initiatives. | |
| Analyzing the Claude Code Source Code Incident | 5 | 5 | 1 | 2 | Lenny inquires about the Claude Code source leak and OpenClaw subscription restrictions. Cat directly attributes the leak to human review failure and articulates the rationale behind gating subscription compute for first-party traffic. | |
| Anthropic's Product Management Organization Structure | 5 | 6 | 2 | 2 | Lenny references Amol's contrasting view that rapid AI shipping requires more PMs to handle overhead. Cat provides a counter-perspective, explaining Anthropic's preference for hiring high-taste engineers who can ship autonomously end-to-end. | |
| The Primacy of Product Taste and Technical Intuition | 4 | 6 | 2 | 3 | Lenny pushes on whether upcoming model leaps like Mythos will eliminate the need for engineering backgrounds in PMs. Cat clarifies that value shifts dynamically and first-principles triage will remain critical regardless of model capabilities. | |
| The Enduring Importance of Human EQ and Common Sense | 4 | 6 | 1 | 1 | Cat describes how human emotional intelligence, stakeholder management, and chaos tolerance remain irreplaceable differentiators. She highlights the necessity of accepting imperfect, unpolished initial releases in favor of velocity. | |
| Trade-offs of Fast Shipping and the Launch of Power Up | 4 | 6 | 1 | 1 | Cat explains the explicit tradeoff of fast shipping: product overlap and user fatigue. She details how customer demand forced Anthropic to reverse their no-onboarding philosophy by building the Power Up tutorial flow. | |
| Unifying Mission and Focus as Anthropic's Competitive Edge | 4 | 6 | 3 | 2 | Lenny posits that Anthropic avoids distraction by ignoring adjacent products like social networks. Cat refines Lenny's framing, distinguishing focus from mission alignment and emphasizing that teams actively sacrifice local team goals for org-wide safety priorities. | |
| Choosing Between Claude Code CLI, Desktop, Mobile, and Cowork | 4 | 6 | 1 | 1 | Cat breaks down the taxonomy of Anthropic's interfaces, explaining how the CLI caters to raw speed and cutting-edge features, Desktop offers visual previews and session orchestration, Mobile allows on-the-go dispatch, and Cowork handles non-code artifacts. | |
| Deep Dive into Cowork: Autonomous Presentations and Workflows | 4 | 7 | 1 | 1 | Cat walks through an operational case study using Cowork to assemble an entire conference slide deck from Slack threads, launch logs, and internal templates, demonstrating how PMs retain editorial control while delegating synthesis. | |
| Personalized Internal Software and the Centrality of Slack | 5 | 5 | 1 | 1 | Cat shares how Claude Code enables internal teams to spin up hyper-personalized internal tools, citing an automated deck customizer built by sales. Lenny adds commentary on Slack's resilience as an enterprise operating system. | |
| Sponsor Message: Vanta Automated Compliance | 3 | 6 | 1 | 1 | Following a sponsor read, Cat outlines how the Applied AI team acts as heavy token consumers by generating customized client prototypes and automated pre-meeting intelligence dossiers through Cowork. | |
| Token Economics and Resource Management Culture | 4 | 6 | 2 | 2 | Lenny asks whether token spend is surpassing employee salaries and if Anthropic employees have unlimited compute. Cat clarifies that token consumption rises with capability but remains below salaries, while emphasizing a culture of responsible usage. | |
| Developing Product Evals and Feature Definitions | 4 | 6 | 1 | 1 | Cat explains the tactical value of writing small sets of targeted evals to clarify ambiguous features. She also details how internal lunch vibe-checks and Amanda Askell's character work shape Claude's low-ego persona. | |
| Deprecating Prompt Crutches as Frontier Models Advance | 6 | 7 | 1 | 1 | Cat details how new model iterations render explicit UI scaffolds obsolete, citing the removal of rigid to-do list system prompts. Lenny recognizes the pattern, referencing the industry adage that models inevitably eat surrounding harnesses. | |
| Unlocking Autonomous Code Review with Advanced Models | 4 | 6 | 1 | 1 | Cat explains how model improvements enabled autonomous code review, moving from basic slash commands to multi-agent PR evaluation that internal engineering teams rely on prior to merging. | |
| Long-Term Vision: From Single Tasks to Scaled Remote Fleets | 4 | 6 | 1 | 1 | Cat lays out the progression from single local tasks to hundreds of parallel remote agent executions. She advises knowledge workers to focus on taking automations all the way from 90% to 100% reliability. | |
| Avoiding the Customization Trap and Focusing on Utility | 5 | 6 | 2 | 1 | Cat warns against the trap of over-customizing setups with excessive MCPs and skills instead of shipping core output. Lenny connects this to Karpathy's observation on the divide between casual AI skeptics and heavy coding practitioners. | |
| Lightning Round Part 1: Books, Media, and Waymo | 2 | 3 | 0 | 0 | Cat shares book and media recommendations including How Asia Works and Free Solo, highlighting her affinity for Waymo due to curb flexibility and call privacy. | |
| Lightning Round Part 2: Life Motto, Thinking Words, and Post-AGI Life | 2 | 4 | 1 | 0 | Cat discusses her life motto 'Just do things,' her favorite thinking word 'manifesting,' and her long-term post-AGI aspirations around rock climbing in Fontainebleau and reading backlogs. | |
| Connecting with Cat Wu and Sharing Feedback | 1 | 2 | 0 | 0 | Cat invites community feedback on edge cases and failure modes in Claude Code and Cowork, and Lenny closes the episode. |
Statements from this episode (38)
Wu: Boris Cherny Sets Claude Code Vision While Product Drives Execution
“He's our tech lead. He's very much the product visionary. And he is great at setting like, this is what the product needs to be. In like three months, six months from now. This is like what the AGI-pilled version of the product is. And a lot of my role is figu…”
Wu: Anthropic product timelines dropped from six months to a single day
“I think now with AI and with how much that has accelerated engineering and with how quickly the model capabilities are improving, the timelines for a lot of our product features have gone down from six months To one month and sometimes to one week or even one …”
Cat Wu: Best AI PMs prioritize rapid idea-to-user deployment cycles
“I think the PMs who do the best on AI native products are the ones who can figure out how can I like shorten the time from having this idea to actually getting the product in the hands of users and help define what are the most important tasks that need to wor…”
Anthropic ships almost all Claude Code features as Research Previews
“So for Cloud Code, what we do is we actually ship almost all of our features in Research Preview. We clearly brand this when we ship something so that users know that this is an early product, this is just an idea, this is just something that we're trying to g…”
Wu: Anthropic teams turn around feature marketing announcements overnight
“So for example, we have a really tight process between engineering, marketing, and docs. So when engineers have a feature that they feel is ready and that we've dog fooded internally, they post it in our evergreen launch room. And then Sarah, who leads our doc…”
Wu: Documenting team trade-offs lets engineers make decisions without PM bottlenecks
“We have this list of team principles, and this includes who our key users are, why those are our key users. And the reason that we articulate all of this is so that everybody on the team feels like they understand how our business works. They understand what's…”
Wu: Anthropic still writes one-page PRDs for ambiguous and multi-month infrastructure features
“For features that are like particularly ambiguous, it does help to write out just a one pager on what the goals are what the delightful use cases are, what the failure modes currently are that we need to fix. And there are occasionally some projects, especiall…”
Wu: Internal AI models do not explain Anthropic's high shipping velocity
“We've been moving pretty fast for Several quarters now, so I think it, it's not fully Mythos. Mythos is an incredibly powerful model. We do use the models internally, and I think this has increased our rate of shipping a little bit, but I don't think it explai…”
Wu: Claude Code leak stemmed from human error despite two human reviews
“We realized that this was the result of human error. There is a human working with Claude to write a PR. This was just an update to how we release our packages. And it actually went through two layers of human review. And so this was a result of human error an…”
Wu: Anthropic retained the employee responsible for the Claude Code source leak
“Yes. Yes. It's a process failure.”
Cat Wu: Claude infrastructure was not designed for third-party products
“It wasn't designed for third party products, which have different usage patterns than our first party ones.”
Wu: AI is merging product management, engineering, and design into one role
“I think all of the roles are merging. PMs are doing some engineering work, engineers are doing PM work, designers are PMing and also landing code.”
Wu: Nearly All Claude Code PMs Have Engineering Backgrounds or Write Code
“And actually almost all the PMs on our team have either been engineers or ship code here on code code.”
Wu: As AI makes code cheap, product taste becomes the ultimate skill
“I still think it comes back to product taste. Like as code becomes much cheaper to write, the thing that becomes more valuable is deciding what to write.”
Anthropic's Wu: Engineering background aids PM prioritization by gauging difficulty
“I think the reason why an engineering background is particularly useful, at least for the next few months, is if you have an engineering background, you have a better sense for how hard something should be. And that's often a factor in what you choose to build…”
Wu: AI Models Still Lack EQ and Common Sense for Product Launches
“I think humans still provide a level of common sense that the models don't. And there's like a thousand moving pieces to any product launch. Some of them are very small, but there's always a lot that could potentially go wrong. I think the model doesn't always…”
Anthropic Ships Unpolished Products Relying on Rapid Post-Launch Iteration
“There's products that we ship that aren't as polished as I wish they were, but you know, our top goal is to help empower professional developers. And if a product isn't successful, as long as it's not blocking the core use case, it's okay because we'll hear th…”
Anthropic Prefers Industry Veterans to Sustain Team Energy and Avoid Burnout
“I think we also tend to hire people who have been in the industry for a while and have experienced lots of ups and downs and have a good sense for what gives them energy and how to maintain their energy over time.”
Wu: Rapid AI development forces Anthropic to sacrifice traditional product consistency
“We're sacrificing product consistency. Historically, when code was expensive to write, you would carefully plan out everything in your product suite, how every product relates to each other, what the use case for every single one is, how they integrate, and yo…”
Wu: Anthropic Reversed Its No-Onboarding Rule to Launch Power Up
“So in the past, we didn't actually want to do something like power up because we felt like the product should be intuitive enough that you can, that you don't actually need to go through any tutorial. And over time, we've just realized that there's just so man…”
Wu: Anthropic is prioritizing first-party Claude products over third-party offerings
“I think one of the most important things for Anthropic is to grow the number of users that we're able to reach. One of the ways that we're able to do this is with the cloud subscriptions with our first party products. And so we just very much want to double do…”
Wu: Claude Code CLI is Anthropic's most powerful tool where features land first
“The CLI is our initial product surface, and it's also the one where our features often land first. And so it's the most powerful of all the tools.”
Wu: Claude Code is for code outputs, Cowork handles non-code tasks
“So the way that I split the products in my mind is if I'm building something where the output is code, I'll use code code or desktop or code code on mobile. And if the output is anything that's not code, I'll use co-work for it.”
Wu: Claude Code has triggered a surge in personalized internal work software
“Like, I think one of the things that quad code has really unlocked for our entire company is it really lowers the barrier to making any custom app that you want. And so we we've seen this like surge in personalized work software that people are building for li…”
Wu: Employee token costs surge with every major AI model capability jump
“And so we do see the token cost per engineer or like per any knowledge worker increase every time that there's a model jump or like a substantial product improvement. I think it's still much lower than what the average engineer salary is, but we see the percen…”
Wu: The hardest skill for AI PMs is planning one month ahead
“I think the hardest skill is being able to define what the product should look like a month from now. I think there's a lot of ambiguity in what models are capable of in that timeline and how user behavior will change. But I think there are patterns that the b…”
Anthropic PM lead: Teams only need 10 great evals, not hundreds
“You don't need to build hundreds of evals for them to be useful. Just building 10 great evals is important for helping the team quantify what the goal is and what their progress towards it is and what they're missing.”
Wu: Crafting an AI's character is harder than coding due to ambiguity
“Even coding is easier because you can verify the success. Whereas crafting the character requires a very strong sense of conviction and what, who Claude should be.”
Wu: Great AI assistants need bias towards action and honest pushback
“I think part of what makes a great coworker is this positivity, this like bias towards action, this ability to give you like earnest feedback, not just agreeing with every single thing that you say. And so we try to imbue this into COD because we think it make…”
Wu: Upgrading AI models often means deleting temporary scaffolding product crutches
“A lot of the changes that we make with a new model is removing features that are no longer needed. So a lot of times we add features to the product as a crutch for the model because it's not naturally doing itself.”
Wu: Claude Opus 4 completes multi-step tasks without explicit prompt reminders
“With Opus four and later models, we realized that we didn't need to force it to use this to-do list. It would like naturally use it itself. For the earlier models, we had to keep reminding it. Hey, did you finish everything on the to-do list? You can't finish …”
Wu: Anthropic engineers must pass Claude-generated code reviews before merging PRs
“This code review is so good that our engineering team relies on this code review to pass before we merge PRs.”
Wu: AI teams must prototype product features that do not work yet
“It's pretty important to build products that don't necessarily work yet so that you know, okay, what is missing for this product to work. And then with the newest model, you can just swap it into the prototype you've already made and see, okay, does this new m…”
Cat Wu: Scaling AI agent fleets will shift execution from local machines to remote cloud
“At that point, you're probably not going to run everything locally on your machine anymore. There's just like not enough RAM to do it.
And so we're thinking about.
How do we make it easier for you to manage all these?
These will probably run remotely.”
Wu: An automation is not truly an automation unless it reaches 100% reliability
“And this, if an automation doesn't work a hundred percent of the time, it's not really an automation. And that last five to 10% does take more time. Also building the automation is often a lot slower than you doing it yourself.”
Cat Wu: Real AI value requires building everyday apps, not prototypes
“I would really push people towards building apps that you're actually using every single day. Cause I think only through that usage, are you actually getting the value? Like if you build a prototype app that. Isn't helping you get more done, then the AI isn't …”
Cat Wu: AI's true 'aha moment' is autonomous execution, not advice
“The like big aha moment people have is when Claude can just like do things on your behalf. It is an amazing feeling to know that the agent is capable of doing so much more than telling you what to do. Like the agent can actually just do it itself. And when peo…”