Anthropic loses its coding product chief to OpenAI ahead of its IPO

Anthropic is preparing what could be one of the largest stock listings ever, though investors now doubt it will happen in November. Its AI coding tool, one of its main sources of growth, just lost the person who ran its product team.

Cat Wu, head of product for Claude Code, Anthropic's AI coding tool, said this week that she is joining OpenAI. She will lead product management for OpenAI's Core Products group, which builds Codex, a direct rival to Claude Code. OpenAI confirmed the hire to Business Insider. Anthropic declined to comment on it.

Wu is not alone. At least three other people tied to Anthropic's coding and growth teams have announced exits since late October. For investors, the question is whether this is ordinary churn at a company that grew more than fivefold in under two years, or an early sign that a pre-IPO workforce is starting to move.

Who left, and where they went

Anthony Morris, who says he created the desktop version of Claude Code, posted that his last day was October 23. Thariq Shihipar, an engineer who worked on Claude Code for more than a year, said at about the same time that he was leaving to take a break. Neither has named a next employer.

Wu's last day was October 30. This week, Austin Lau, who says they helped build Anthropic's growth team, also posted about leaving; Lau's profile now lists OpenAI, though neither company has confirmed the hire. Some reports add a fifth name, engineering manager Sunil Shah. He has not confirmed publicly, and at least one tracker leaves him off the list.

None of those who posted gave a reason for leaving, and no outlet has found a shared cause. Boris Cherny, Claude Code's original creator and the product's other public face, has not announced any departure.

Wu has left before. In July 2025, she and Cherny joined Anysphere, the company behind rival coding tool Cursor, and returned to Anthropic about two weeks later.

Her new reporting line is what makes this move pointed. She will report to Thibault Sottiaux, who ran Codex engineering before OpenAI merged ChatGPT, Codex and its developer tools into one product group in May. That group is where OpenAI is concentrating its push into AI agents, software that completes multi-step tasks on its own. OpenAI said in May that more than 4 million people used Codex each week.

Why the coding team is the one that matters

Claude Code is not a side project. Its annualized revenue had passed $2.5 billion by February. It is also tied to the heaviest spending by Anthropic's business users.

According to 36Kr's reading of Anthropic's filing, individual users of its business plan who did not use Claude Code averaged under $75 a month each in the third quarter. The top 5% of those users who did use Claude Code spent about 20 times that.

That matters more as overall growth cools. Anthropic's amended filing shows about $17 billion of revenue in the third quarter of 2026, up roughly 47% from the prior quarter. Second-quarter growth had been 144%. When the top line slows, the products that keep big customers spending carry more of the valuation.

OpenAI now has a product leader who knows firsthand what Anthropic's heaviest users asked for, and what the team chose to build. For companies buying either tool, nothing changes yet. No product changes have been announced, and Anthropic has not named a successor.

What the filing says, and what it leaves out

Anthropic's prospectus names this exact risk. It says the company is "highly dependent" on its co-founders, senior managers and key technical staff, and that losing them could hurt the business. Business Insider's reading of the filing noted that rivals are making aggressive offers for scarce top talent.

The same filing offers a reassuring number. Regretted attrition, meaning departures of people the company wanted to keep, was about 1% in the 12 months to September 30. Headcount grew from roughly 940 at the end of 2024 to about 5,300 by September 30. Four or five exits will not move that figure much, though they come from a small, high-value team.

Pay is the bigger variable. Anthropic grants staff restricted stock units, shares that vest only after both time served and a stock listing. Analysts reading the amended filing, submitted November 2, say Anthropic will book about $8.2 billion of stock pay when the IPO is completed. The filing shows $24.1 billion more to be recognized over about two years.

That unvested value is a powerful reason to stay. It weakens as shares vest after the listing and post-IPO selling restrictions lapse.

Asked on X whether leaving would let her sell shares sooner than colleagues who stayed, Wu said she had "no idea" and that it "wasn't a factor for me," Business Insider reported. Public coverage has not shown how Anthropic's plan treats units held by people who leave before the listing.

Talent has also flowed the other way this year. In June, Anthropic hired John Jumper, the Nobel-winning leader of Google DeepMind's AlphaFold team, and Clive Chan, an early member of OpenAI's custom chip program.

Three things to watch: whether the roadshow or a later filing updates Anthropic's attrition figures, who replaces Wu, and whether more Claude Code staff follow her. The risk for IPO buyers is not four people. It is whether the team behind Claude Code stays once its shares can be sold.