Claude Code Product Lead Cat Wu Leaves Anthropic
Claude Code product lead Cat Wu announced that she has left Anthropic, with last Friday being her final day at the company. She recalled that two years ago, when Claude Code was still just a small internal tool, she spent dozens of hours guiding it to build a reinforcement learning environment without writing a single line of code herself, and then joined the team. She noted that when the product first launched, most people manually confirmed permissions one by one, but today the default has changed to trusting Claude's judgment and granting authorization automatically.
View original postFriday was my last day at Anthropic. Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env. I steered it for tens of hours but never manually wrote a line of code. It worked, I was hooked, and I joined the team. When we first launched, most people approved every permission prompt by hand and mostly used Claude to answer questions and write tests. Today, the default is trusting Claude's judgment to auto-approve. Claude does large migrations and builds new products. Helping build Claude Code into an industry-defining product has been the journey of a lifetime. Thank you to the incredible team behind Claude Code, Cowork, and Tag. These products are the sum of thousands of great ideas from you. I'll miss working with you all. Thank you to our users for embracing a new way of working. Your constant feedback shaped so much of what we shipped and your stories made the work so meaningful. Excited to see what you all make next!
View original
Friday was my last day at Anthropic. Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env. I steered it for tens of hours but never manually wrote a line of code. It worked, I was hooked, and I joined the team. When we first launched, most people approved every permission prompt by hand and mostly used Claude to answer questions and write tests. Today, the default is trusting Claude's judgment to auto-approve. Claude does large migrations and builds new products. Helping build Claude Code into an industry-defining product has been the journey of a lifetime. Thank you to the incredible team behind Claude Code, Cowork, and Tag. These products are the sum of thousands of great ideas from you. I'll miss working with you all. Thank you to our users for embracing a new way of working. Your constant feedback shaped so much of what we shipped and your stories made the work so meaningful. Excited to see what you all make next!
Body · AI translation
Oh my god, Cat Wu, the goddess of Claude Code, the godmother-level figure, has also left Anthropic??? As the product lead who personally drove Claude Code, the words she wrote upon leaving are the sexiest and most sincere confession about human-AI collaboration I've read recently 😭 Two years ago, Claude Code was still an inconspicuous little experiment inside Anthropic. She stared at it for dozens of hours without writing a single line of code to build a reinforcement learning environment; the second it ran successfully, she decided to join, and that was also the moment the gears of all our fates began to turn.
When the whole environment finally ran successfully, she was completely captivated and decided to plunge headlong into this team.
This is almost the decisive moment that every one of us who has truly been struck by AI has experienced.
What's most intriguing is the micro-evolution of human-machine relations she recorded over these years: when Claude Code first launched, the vast majority of developers nervously confirmed every single terminal permission pop-up by hand, daring only to hand it the most marginal tasks, like answering questions or writing unit tests.
And today, the default posture of tens of thousands of people is to simply turn on full auto-authorization, letting AI schedule multiple sessions and projects in the background while humans retreat to a more macro position to control goals and accept deliveries.
Many people think the secret to making AI products is piling on compute and features. But on a podcast she discussed an extremely clear-headed principle: when the model itself becomes increasingly powerful, all the complex interfaces and steps built early on to assist weak models become heavy liabilities, and top teams must dare to tear them down one by one with their own hands.
From cautious manual confirmation to calm and complete trust; from a craftsman typing line by line of code to a navigator setting the course.
She spent two years defining a product that changed the workflows of countless people,
but what moves me most is still those dozens of hours two years ago when she sat in front of the terminal, writing not a single line of code yet watching a miracle happen.
Technology is being frantically restructured on a daily basis, but what truly fascinates people is always the pure curiosity ignited in a person at some tiny moment.
The original source and the full thread of her discussion on product evolution are in the first comment.