How Anthropic Ships Faster Than Anyone — Cat Wu's Playbook

Lenny just dropped an interview with Cat Wu — Head of Product for Claude Code and Cowork at Anthropic — and it's the kind of conversation that makes every other product team look slow.

The big idea Wu shared on the Lenny show: stop waiting for the model to be ready. Build the product now, and the model will catch up.

That's a backwards way of thinking about product development if you came up in the "validate first, build second" world. But in AI-native, it's the only way that actually works. Let me break down what stood out and where I'd add my own experience.

Anthropic's ship cadence

Wu describes Anthropic's release timeline going from months to weeks to days. That's not a vague claim about being fast — that's a real, measurable compression. They're treating the cycle the way most product teams treat a hotfix.

What I love about this is how unsexy the explanation is. Wu doesn't talk about magic processes or new frameworks. She talks about mission alignment — every team agrees on the same north star, so you don't burn three weeks debating priorities every quarter.

I see this in my own business all the time. The teams that ship are the ones where nobody has to ask, "wait, are we even doing this?" That question alone kills more products than any technical limitation.

"Just do things"

The operating principle Wu mentioned that I keep thinking about: just do things.

In her words, action over strategy. Experimentation over comprehensive frameworks. In an AI-native company, the rate of change is so fast that any 90-day plan is basically dead on arrival. So they don't write 90-day plans. They write 9-day experiments.

I've been running my own funnel and SaaS work this way for the last year. Every time I sat down to "plan a roadmap" for the year, I'd ship less than when I just wrote one thing on a sticky note and built it that day. The plan was the bottleneck, not the work.

Building before the model is ready

This is the line that flipped a switch for me. Anthropic ships products with incomplete functionality because they're betting on the next model release closing the gap.

Most companies wait for the capability to be there first, then build. Anthropic builds first, and lets the model catch up. The result: the product is already in users' hands the day a new model drops, instead of starting from scratch.

If you're building anything AI-adjacent, this is the playbook. Don't wait for the perfect foundation — build the rough version now and ride the curve.

The introspection trick

Wu mentioned a tactic I'd never heard before: prompt Claude to analyze and explain its own mistakes. She called it underutilized — and I think she's right.

I tested this on a chatbot I'd trained on a marketing framework, and the response when I asked, "where did you go wrong on that last answer?" was way more useful than re-running the prompt. The model can self-critique in a way it can't self-improve in a single shot.

Personality as a moat

The last thing Wu said that landed for me: Claude's personality is a moat.

It sounds soft, but think about it — most AI products are functionally interchangeable on raw capability. What differentiates them is how they feel to use. Claude has a distinct voice. ChatGPT has a different one. That voice is the thing that makes users stay or leave, and you can't copy a voice.

I've been telling people the same thing about funnels for years. The features don't matter — the personality of the brand inside the funnel is what converts. AI is just proving the same lesson at a bigger scale.

The Bottom Line

The takeaway from Cat Wu's interview isn't a list of frameworks. It's a posture: ship fast, prompt the AI to teach you what you missed, and let personality be the thing that keeps people coming back. Most product teams won't do this because it feels too loose. The ones that do are the ones lapping the field.