Anthropic is seeing some high-profile departures just ahead of its IPO.
Catherine “Cat” Wu, who has spent the last two years as Anthropic’s head of product for Claude Code, has announced she is leaving the company to join OpenAI, where she will lead product management for its Core Products team.
Wu shared the news on X, writing that she is joining OpenAI to lead product management for Core Products, and said she has a lot of respect for OpenAI for proving that scaling works and for bringing AI to more than a billion people. She said she is looking forward to working with Thibault “Tibo” Sottiaux and the rest of OpenAI’s product organization to build agents that keep users “in control” while freeing up their time. At OpenAI, Wu will reportedly work alongside Nick Turley, who runs enterprise products, and Ashley Alexander, who runs consumer products, reporting to Sottiaux.
Wu joined Anthropic in August 2024 on the Research Product Management team, where her job was connecting model research with real customer needs. That fall, when Claude Code first became available internally, she was among the earliest employees to use it to build internal tools, run evaluations, and prototype features — work she later credited with reshaping how she thought about product management in the AI era. By 2025, she had become Head of Product for Claude Code, and later took on Claude’s Cowork product as well, overseeing the tool’s evolution from an internal coding experiment into one of Anthropic’s central commercial products.
Alongside Boris Cherny, the engineer credited with writing much of Claude Code’s original code, Wu became one of the public faces of the product, with the pair nicknamed Anthropic’s “Batman and Robin” by some industry commentators. She was a frequent presence at developer conferences and in interviews discussing how Claude Code moved from manually approved permission prompts toward a model where the tool is trusted to auto-approve actions and run large, autonomous coding migrations on its own.
Before Anthropic, Wu was a partner at Index Ventures, the venture firm that later co-led Anthropic’s funding round in September 2025.
Part Of A Wider Pattern Of Departures
Wu’s exit is notable not just for her seniority, but for its timing. It comes just weeks after Anthropic confidentially filed for a US IPO that could value the company at up to $2 trillion, and follows a string of other high-profile exits from the Claude Code team specifically.
Earlier in the same week, Claude Code engineers Thariq Shihipar and Anthony Morris both announced they were leaving Anthropic within a day of each other. Shihipar, known on X by the handle @trq212, had become one of Claude Code’s most recognisable voices online, regularly fielding bug reports and community feedback and stepping in to publicly address backlash on the company’s behalf when controversies arose. Morris described himself as the creator of Claude Code’s desktop application. Neither detailed their next steps.
Anthropic has historically maintained one of the lowest voluntary attrition rates among major AI labs, a pattern some commentators have linked to the company’s tighter restrictions on employee share liquidity compared to rivals — restrictions that, as Andrej Karpathy had once suggested, have historically discouraged employees from leaving to start companies of their own. It also remains the only major AI lab with its entire founding team still intact. An IPO changes that calculus for rank-and-file employees too, since it finally gives vested staff a concrete path to cashing out equity, raising questions about whether Wu, Shihipar and Morris are an early wave of a broader exodus still to come.
The departures are also notable given how aggressively Anthropic has otherwise been hiring, including poaching chip and hardware specialists directly from OpenAI’s own teams.
Wu’s move to OpenAI is a particularly pointed one given the competitive dynamics involved: OpenAI’s own coding product, Codex, is widely viewed as Claude Code’s closest rival, making Wu’s hire a direct recruitment of competitive product leadership rather than a lateral move into an unrelated part of the business.
Claude Code has become central to Anthropic’s growth narrative heading into its IPO, with the company saying internally that it now uses Claude to generate nearly all of its own code, and positioning the agentic coding tool as a key pillar of its enterprise pitch to investors. Losing three people closely identified with building and publicly representing that product within the span of a couple of weeks, right as the roadshow approaches, is scrutiny Anthropic would likely have preferred to avoid.
Wu has not detailed her exact start date at OpenAI, but closed her farewell post by inviting continued feedback, saying there is “lots to learn, and lots to build.”