ClaudeCode’s female boss swiftly backs OpenAI—just left Anthropic yesterday!

  Xinzhiyuan reports 

</h3

Case solved!

The former lead of the Claude Code product who announced her resignation yesterday Cat Wu, actually went to OpenAI?

Yesterday, after posting her farewell letter, she joked with netizens that she would “rest after AGI is achieved,” and everyone thought she was going to start a business.

But! Just today! She turned around and announced that she is joining OpenAI to lead the core product team.

OpenAI President Greg Brockman retweeted a welcome message within three minutes.

The speed was insane—so fast that Cat Wu’s bio didn’t even have time to remove Anthropic.

Cat Wu said plainly that she admires OpenAI “bringing AI to over a billion people.” Her next step is to “create an Agent that backs you and your team while keeping control in your hands.”

The comment section blew up instantly, with OpenAI inference guru Noam Brown, developer experience lead Romain Huet, and other executives quickly lining up to welcome her.

Netizens exclaimed: this transfer move is literally “Figo‑level” — the Barcelona captain moving to arch‑rival Real Madrid, leaving everyone’s jaws on the floor.

On the eve of the listing, the elite squad loses five members in a row

This is because, just ten days ago, Cat Wu was still the “anchor” of the Claude Code team.

 Anthropic’s talent earthquake came as a surprise.

Looking back at her sudden defection, Anthropic, which is racing toward an IPO, has seen its core product team undergo a talent earthquake lasting nearly two weeks.

On October 24, Claude Code early core engineer Thariq Shihipar announced his departure to start a company.

He specifically listed a series of accounts of the remaining key creators at the end of his farewell tweet, urging everyone to keep following, with Cat Wu at the top.

And ten days later, Cat Wu left too??

Moreover, since Anthropic secretly filed its IPO prospectus on October 26, within just 12 days, the two flagship teams Claude Code and Cowork, have already lost five core contributors.

Netizens are joking: everyone is timing their resignations and seamless transitions to new employers at this moment—could it be becausetheir stock options justvested?

Among these five key figures, early engineer Thariq, engineering manager Anthony, and team lead Sunil Shah chose to go solo,while the other two core leaders both jumped to the rival.

On October 29, Cowork design lead Jenny Wen announced her resignation and moved to join Elon Musk’s xAI;

On November 5, the head of Claude Code and Cowork’s product lead, Cat Wu, officially announced joining OpenAI.

Last summer, Cat Wu’s job‑hopping story was already dramatic.

In July 2025, Claude Code just went viral, and Cat Wu, together with Boris Cherny, was poached by the star‑studded startup Cursor.

But the two of them stayed at Cursor for less than two weeks before both packing up and returning to Anthropic, where they continued to turn Claude Code into a flagship generating over $2.5 billion in annual revenue.

That quick resignation and return made her realize: in application‑layer startups, your hands are always tied to someone else’s model;to define the next‑generation Agent, you must stay in a lab that has its own top‑tier large model, letting the model and the product grow together..

This time, her longtime partner Boris stayed at Anthropic, while Cat Wu headed to OpenAI on her own.

The compelling factor she wrote down clearly was:OpenAI holds a billion‑scale consumer user base that Anthropic lacks..

After proving herself in a community of a few million programmers, embedding an Agent into the daily lives of hundreds of millions of ordinary people worldwide is obviously what excites a product fanatic the most.

To welcome this product chief, OpenAI has already set the stage internally,the clue is hidden in the account mentioned in Cat Wu’s announcement tweet.

As the head of the OpenAI Codex and ChatGPT engineering teams, Tibo, two days before Cat Wu’s announcement of his joining, vented on X.

He lamented, “The current agents are unbelievably powerful, but using them is still maddening.”

He then added meaningfully, “I can’t wait to tackle these challenges and regain that smooth flow state.”

Now everyone realizes, this is a triple combo.

Nov 3 Tibo and the developer team rolled out the brand-new internal Git backend Pando, cutting large repository push latency by 80%, handling 1 PB of code data per day, paving a high‑speed road for massive agents to write code simultaneously.

Nov 4, OpenAI swiftly acquired Pierre, who has been underpinning Codex’s foundation, and Tibo exclaimed, “Let’s reinvent everything together.”

On the morning of November 5, the third piece of the puzzlearrived!The arch‑rival Anthropic’s product lead, who knows best how to help users regain a “smooth experience,” was poached.

Behind OpenAI’s moves lies a year of anxiety on the enterprise and office front.

Although ChatGPT dominates the consumer chat market, in the most lucrative programmer coding and white‑collar desktop Agent tracks, OpenAI has watched Claude Code and Cowork steal a large chunk of customers.

Even OpenAI co‑founder Karpathy publicly praised CC as useful.

Both companies’ base models are similarly smart, yet the Agent built by Anthropic feels more intuitive, the secret lies in the secret weapon that Cat Wu brought to OpenAI.

First, the ability to turn geek tools into mass‑market products—a breakthrough sense of crossing circles.

In January of this year, Cat Wu discovered that many programmers were actually using their code‑writing terminal tools to organize invoices and write weekly reports. She and her team then pulled an all‑nighter for ten days and created the desktop productivity product Cowork, enabling non‑technical office workers to command AI to handle miscellaneous tasks.

On the day it launched, it sent Wall Street’s traditional software stocks tumbling.

Second, it’s about giving AI the sense of “working diligently but never overstepping.”

Just a few days before she left, she released her final project at Anthropic, Claude Duet, an proactive agent that quietly monitors your to‑do list in the background, preparing tasks for you before you even ask.

But as soon as it involves sending messages, editing files, or any serious operation, it will first obedientlypop up a windowasking you to nod.

It both covers the bases and stays within limits, and she emphasized this in the onboarding post just now.

Watching OpenAI and xAI snap up a product lead and a design director respectively, the tide of the talent war in Silicon Valley has already shifted.

A couple of years ago, the giants were waving billion‑dollar checks to poach each other’s staff, all targeting pre‑training gurus and reinforcement‑learning scientists.

But by the end of 2026, everyone realized: the models are already smart enough, yet if the products they produce are laggy, hard to use, or mess up users’ files, no one will pay for them.

Whoever can tame the model into a handy, user‑friendly tool holds the real money‑printing machine.

Now, the talent war in Silicon Valley is about grabbing people who understand users.

With ChatGPT and Codex in hand, how will the former CC queen stir the battlefield?

Reference materials:

https://x.com/_catwu/status/2118137612577968402

https://x.com/gdb/status/2118138320375136695

https://x.com/polynoamial/status/2118140090631495703

Edited by: Aeneas Moses