Really good follow-up conversation between Nikhyl and Dianne Na Penn (former podcast guest, runs product for Anthropic's research and Labs teams) about who makes it as a PM at Anthropic (and who doesn't).
Dianne was Anthropic's first technical product manager, joining before Claude was even publicly available. Since then, she and her team have developed every version of Claude, from Claude 2 through today's 5.5 family. She's also helped incubate MCP, computer use, skills, Claude Code, Claude Design, and much more.
I sat down with Dianne Na Penn, who leads product for Anthropic’s research and Labs teams. She joined in 2023 as the company’s first technical product manager, before the Claude app existed. Since then her team has product managed every model from Claude 2 through the Claude 5.5 family, and incubated Claude Code, MCP, and Claude Design.
Most of what gets written about Anthropic is about the models. This conversation was about the people. How she hires PMs, how she ramps them, and how she keeps them.
I find how PMs operate at a frontier lab an excellent glimpse of the future. The job there works differently from the one most of us know. A new hire is encouraged to thoughtfully debate with the chief scientist. Planning happens a month at a time. What gets you hired is the hard thing you pushed through at your last company, not the size of your last title. Some of that is specific to a lab. Most of it is a preview of the PM job everywhere, with lessons to steal for the skills you should build and the career moves ahead of you.
Below is what stood out, and what each one means outside the lab. The full episode has more, including how she keeps a team at this pace without burning it out.
A bit more on each of these below.
The First Two Months Are Supposed to Be the Hardest
Every new hire on Dianne’s team gets the same line in their onboarding plan: the first two months should feel like some of the hardest of your career, and that means you’re doing it right. The point of the hard start isn’t to see who breaks. It’s to show, quickly, what working at this pace requires, and to build the habits that make it possible. A new PM gets a real area, usually a top priority, and the expectation that they move at the company’s speed rather than their own.
What makes it hard is that nobody hands you the structure. There’s a demanding first assignment, but no scaffolding around it and no sequence of sign-offs. The answers are all there. Most channels are public, and her advice to a new hire is to scroll back two months and read what happened, then show up saying here’s what I think is going on and here’s my view, rather than asking the room to stop and catch you up. A spin-up buddy helps. Mostly it takes the creative hustle to go find things out.
The people who settle in fastest are the ones who take that on. The ones who struggle tend to arrive with the reasons they couldn’t do X, Y, and Z, usually because they’re waiting for a structure that isn’t coming. She can tell which way it’s going within the first month or two, but the ramp isn’t a test. It’s training.
Outside the lab, the same shift is underway. The listening tour is getting shorter, and the expectation is impact in the first few weeks. For a manager, the part worth copying is to give a hard assignment early, point to where the answers live, and make clear that going to get them is the job.
New Hires Are Expected to Push Back
The habit that trips up PMs from large companies is deference. If your VP says X, you do X, all the way down the chain, and nobody asks whether X is still the right thing. A research lab runs closer to academia, where a first-year PhD student might question their advisor. The PMs who make it at Anthropic are good at thoughtful debate. They can walk up to researchers and execs with important user feedback, point out gaps, and question assumptions.
There’s a right way to do it, and most of it is a soft skill. The first few ideas a new person has were probably considered two or three years ago, and there’s a history of why they didn’t happen. So the homework comes first: read what was tried and understand why it died. Then push, but in a way that is curious and builds trust. What she doesn’t want is the hot take, the “you’re wrong” that never gets into the details. What she does want is someone who has done their own thinking, worked out the so-what two or three steps ahead, and keeps at it when the first idea doesn’t land. Her phrase for the people who do this well is that they join the discourse: they add to it, build on it, change it in some way. Questioning assumptions gracefully, and arriving with a point of view of your own, is the skill.
The culture makes it easier than it sounds. Most channels at Anthropic are public, and it runs as a yes-and culture: someone you didn’t know was in the conversation will interject with “have you thought about X.” Kindly, she adds. Nobody is waiting to be asked.
Managing up used to mean keeping the people above you informed. Now it means having done the reading and being willing to hold an opinion against them. That’s uncomfortable for anyone trained in a hierarchy, and it’s the shift a senior hire has to make early.
What Her Best Hires Have in Common
Her favorite profile is the first PM at a startup. That person was also the first customer success hire, did business development at some point, and maybe even the IT person. The job forced them to wear every hat and figure it out. Founders fit the same mold. So do people with a track record of doing something hard inside a big company, where inertia is the obstacle. Each of them is a story about pushing through inertia, with nothing to do with a ladder of bigger titles. That grit, that get-stuff-done attitude, tends to be more powerful than almost anything else.
Her logic is simple. People with that story stay curious, go find the answers they don’t have, and act without waiting for permission. It’s also why so many careers on her team have wandered, as she puts it, instead of climbing the same job one level at a time.
For anyone interviewing, that’s the story to lead with. Not the scope you were handed, but the thing you got done when the company was in the way.
Plan a Month at a Time
In the early years Anthropic ran a full annual OKR process. The OKRs got thrown out quickly, because most assumptions about usage and adoption turned out to be an underestimate. Nobody wanted to spend three months on an annual plan that goes out of date within two weeks. What her team does now is simpler. A short set of assumptions about where the world will be in three to six months, and a one-month plan for making progress against them. The long view comes from the research org and the mission, which keep the team from reacting to whatever a competitor shipped last week. Life sciences is the current example. The stated ambition is to compress decades of progress in biology into a few years. That isn’t a quarterly target. Getting scientists access to the models and deciding which training experiments to run is monthly work.
Before committing to a project, she asks what the models will be able to do three months from now, and six. A fine-tuning API is her example. The team debated offering one. But by the time a customer figured out how to post-train a model and deploy it, the base models would have moved on two or three generations and they would have to start over. The habit underneath is staying stubborn on where the models are headed, even when a reasonable request pulls the other way. Stubborn is her word for the few commitments that survive the monthly replanning.
The version for the rest of us: write down what you’re stubborn about for the next six to twelve months, and treat everything else as a plan that gets rewritten monthly.
Less to Unlearn
Two groups on her team stand out. The first is early-career PMs, who she describes as extremely effective and high ceiling. They have no process to unlearn. They don’t wait for a sign-off before the next step. They decide they own a capability, talk to the customers, talk to whichever researcher they need, and go.
The second is forward-deployed engineers who transferred into product. They’ve been exceptional. They already know the users and the systems those users run, and they bring that knowledge back to the research team. It’s the same hands-on work she did herself in 2023, when she wrote the claude.ai system prompt because there was nobody else to do it.
Together, the two groups point to a new way into product, something like a modern APM program. Google’s version rotated new PMs through teams. This one starts with time next to customers and the systems they use. If you’re trying to get into product and the PM job isn’t open, find your way into a customer-facing role at an interesting company and make the move to product from there. If you’re already senior, expect that some of what you know will have to be unlearned.
The episode goes further than this. Dianne walks through the bootcamp her team runs for every PM at Anthropic, how she decides whether a problem needs a change to the model or a fix in the product, her own move from strategy work into the details of execution, and the doc she keeps on working at this pace without burning out. If you’re thinking through your next move, or how to run your team, as we head into 2027, it’s worth the hour.
Have your own career question? Get personalized guidance at Nikhyl.AI. It’s where the questions keep coming, and where I’ll keep sharing what I’m learning.
