Claude Code Product Lead Cat Wu Joins OpenAI as Anthropic Loses Three Core Members Ahead of IPO

By Letter List

On November 5, Cat Wu, who had just resigned as the product lead for Claude Code, announced her next stop:

She will join OpenAI to lead product management for core products, working with Thibault Sottiaux (Tibo) and his team.

Just the day before, Cat Wu had announced that she had already left Anthropic on October 30.

This was actually her second departure. In July 2025, she had jumped ship to Cursor together with Claude Code founder Boris Cherny, and just two weeks later, the two of them both returned to Anthropic.

Now this pair has parted ways again: Boris stayed at Anthropic, while Cat Wu turned around and joined its "sworn enemy" OpenAI.

It is worth noting that Cat Wu is not the only core member to leave recently. On October 23, Anthony Morris, the creator of the Claude Code desktop client, resigned; the next day, core engineer Thariq Shihipar also announced his departure.

Anthropic has just made public its IPO prospectus, and a capital feast that could create enormous wealth is close at hand, yet Claude Code has lost three key figures in just two weeks.

And OpenAI has gained its star product lead.

If You Leave, I Leave Too

Cat Wu was one of the earliest participants in Claude Code.

(Cat Wu's homepage, where the bio has already changed)

In August 2024, she joined Anthropic, initially responsible for product management in the research direction. In the fall of the same year, Boris Cherny developed an early prototype of Claude Code. At the time, the tool was still called Claude CLI, and it could access files, run terminal commands, and directly modify code.

When Boris first posted a demo video to the company's internal Slack, it received only two or three likes, and Cat Wu was one of the most active testers.

It was then that she began using Claude CLI to build reinforcement learning training environments. She spent dozens of hours continuously giving it instructions, and without writing a single line of code herself, she ultimately completed the task successfully, which made her fall in love with the product.

In Anthropic's official oral history of Claude Code released this year, "The making of Claude Code," she recalled that she often sent Boris long essays of product feedback at the time. Boris also replied quickly; often, just after Cat Wu raised a request, Boris would tell her that the relevant feature had already been implemented (or the bug had already been fixed).

Cat Wu gradually became one of the most active internal users of Claude CLI, and eventually Boris took the initiative to invite her: "Do you want to join?"

So the two began working together formally.

In February 2025, when Claude CLI was preparing for public release, the situation was actually not very promising. Cat Wu revealed that the response from early trial users was rather lukewarm. Some thought the product was interesting but complained that there were too many bugs. The team ultimately decided to release it as planned. It was also at this time that the tool was officially renamed Claude Code and unveiled as a research preview.

At the time, few people probably imagined that this tool, which had received a lukewarm early response, would quickly grow into one of Anthropic's most successful products.

Cat Wu also gradually became one of the most representative figures of Claude Code, responsible together with Boris Cherny for the product's iteration and development. In July 2025, the two left Anthropic together and returned together just two weeks later. At that time, they always seemed to move as a unit.

In May this year, when TechCrunch interviewed Cat Wu, it even mentioned that outsiders called the two Anthropic's "Batman and Robin." Cat Wu was responsible for product direction, while Boris was responsible for technical development. The two often gave interviews and talks together to introduce Claude Code's development to the outside world.

Cat Wu's work was in fact not limited to coding products. She was also responsible for Claude Cowork for non-programmers and participated in the development of agent products such as Claude Tag.

In July this year, at the AI Engineer World's Fair, she also gave an interview together with another Claude Code core engineer, Thariq Shihipar, to developer Simon Willison.

In that conversation, Thariq also told the story of how he joined Anthropic.

In 2025, around the release of Claude Opus 4, a friend sent him a message strongly recommending that he try Claude Code. After trying it, Thariq was greatly impressed and immediately had the idea of joining Anthropic. After that, he participated in the development of core features of Claude Code such as context engineering and dynamic workflows.

In July this year, Thariq also published a technical article as an Anthropic engineer, explaining how the team redesigned the context mechanism for the new generation of Claude models, reducing the Claude Code system prompt by more than 80%.

For Claude Code users, he was also a fairly active face in the developer community, frequently introducing new features publicly and sharing engineering experience.

But on October 24, Thariq suddenly announced his departure.

He recalled that he had personally watched Claude Code grow to a scale far beyond imagination, and that he deeply identified with Anthropic's mission. He planned to rest for a while first, then move on to something new.

At the end of his farewell letter, he also specifically recommended several Claude Code team members worth continuing to follow, including Cat Wu and Boris Cherny.

As it turned out, ten days later, Cat Wu also announced her departure.

Another departee, Anthony Morris, was the creator of the Claude Code desktop app.

Early Claude Code was mainly operated through the command line, which posed a certain barrier for users unfamiliar with terminals. Anthony pushed for the development of a desktop app, hoping to enable more people to directly use AI to complete software development tasks.

In April this year, he announced the launch of a redesigned Claude Code desktop app, making it more convenient for users to run multiple AI programming tasks simultaneously.

Until mid-October, he was still publicly responding to user feedback on the desktop app and promising to continue improving features.

But on October 23, he also left Anthropic.

In his farewell statement, Anthony wrote that he had always believed everyone can be a creator. Being able to participate in Anthropic's mission was one of the most important experiences of his career. Leaving his friends, the product he had poured his heart into, and the company's mission was the hardest decision of his career.

However, he now finally has time to play "GTA 6."

All three farewell statements expressed affirmation of Anthropic, did not publicly accuse the company, and did not disclose specific reasons for leaving.

In February this year, Anthropic disclosed that Claude Code's annualized revenue run rate had exceeded $2.5 billion, and its weekly active users had doubled from the beginning of the year.

The latest IPO prospectus, although it did not separately update Claude Code's revenue, further revealed its commercial value.

In the third quarter of 2026, Claude Enterprise users who did not use Claude Code spent an average of less than $75 per month; among Claude Enterprise users who did use Claude Code, the top 5% by spending had an average monthly expenditure about 20 times that of the former.

In the same period, Anthropic's overall revenue reached about $17 billion, with cumulative revenue of $33.2 billion in the first three quarters.

From its debut in February 2025 to becoming one of Anthropic's most important revenue growth engines, Claude Code took just over a year.

Now Anthropic is sprinting toward an IPO. Yet precisely at this time, several key members involved in creating it have left the company one after another.

Cat Wu went directly to OpenAI, arriving at Anthropic's "most hated" former employer and competitor.

A new colleague at an old rival

According to Cat Wu's own post, she will be responsible for product management of core products at OpenAI, working as a colleague of Tibo.

"Cyber godfather" Tibo probably needs little introduction. In May this year, OpenAI underwent a round of organizational adjustment, bringing businesses such as ChatGPT, Codex, and the API into a unified core product team, with Tibo responsible for core products and platform business. OpenAI President Greg Brockman was responsible for overall product strategy, former ChatGPT head Nick Turley moved to enterprise business, and Ashley Alexander took charge of consumer products.

According to foreign media reports, OpenAI has confirmed that Cat Wu will report to Tibo and work with Nick Turley and Ashley Alexander. However, OpenAI has not yet disclosed which products she will specifically be responsible for.

Tibo also reposted Cat Wu's announcement post, welcoming her to OpenAI and looking forward to working together in the future. He then joked in the comments: "CatGPT."

Cat Wu's joining happens to coincide with a stage in which OpenAI is continuously adjusting its product organization and business direction.

In the organizational adjustment in May this year, Brockman explicitly stated that OpenAI needed to integrate ChatGPT and Codex into a unified product experience while competing for both consumer and enterprise markets; in June, OpenAI announced that Codex had more than 5 million weekly users, of whom about 20% were not developers, covering groups such as analysts, marketers, designers, and researchers.

By the DevDay conference on September 29, OpenAI further disclosed that Codex and ChatGPT Work together had more than 35 million weekly users.

This tool, originally aimed at programmers, is being promoted to a broader range of work scenarios.

Just last week, OpenAI also announced a new billing adjustment: starting in December, the advanced feature usage quotas for Plus, Pro, and Business users in ChatGPT and Codex will be merged and calculated uniformly by tokens. Official emails received by some users also revealed that OpenAI will run a small-scale test of integrating Chat and Work in the coming weeks.

It is worth noting that Anthropic had already integrated Claude's chat and Cowork features as early as September this year. The two companies are advancing similar Agent product experiences.

OpenAI's recent adjustments overlap considerably with Cat Wu's previous work at Anthropic.

In her joining statement, Cat Wu specifically mentioned OpenAI's user scale. She said she appreciated that OpenAI has already brought AI to more than a billion users, and looked forward to working with the new team to develop Agents that can help individuals and teams complete work while keeping users in control.

As coding Agents expand into daily work and team collaboration scenarios, the product layouts of OpenAI and Anthropic are becoming increasingly similar, and the enterprise market has become an important area of competition between the two sides.

On November 5, according to foreign media reports, OpenAI President Greg Brockman revealed that enterprise customers now contribute more than half of the company's revenue.

As for whether this job change was facilitated by OpenAI actively poaching, how long the two sides had been in contact, and what compensation and equity package Cat Wu received, no public information is currently available.

On the eve of the IPO, they don't want money?

Anthropic is sprinting toward a trillion-dollar valuation. Choosing to leave at this time inevitably makes people wonder: don't they want to wait until the IPO to cash in their wealth?

On November 2, Anthropic updated its IPO prospectus. According to the latest reports, some investors suggested setting its IPO valuation at $1.5 trillion to $1.6 trillion, a shrinkage of about 20% to 25% from the previous expectation of $2 trillion.

Even after shrinking by hundreds of billions of dollars, this is still an astronomical figure. But Cat Wu and others do not seem eager to wait until the day of the IPO.

In fact, leaving Anthropic does not necessarily mean giving up the gains brought by the IPO. At Silicon Valley tech companies, equity incentives are often an important component of employee compensation. For employees who joined earlier, as the company's valuation rises, the value of this portion of assets may far exceed their salaries.

According to the latest prospectus, the equity incentives Anthropic grants to employees mainly include stock options and restricted stock units (RSUs).

Stock options give employees the right to purchase company stock in the future at an agreed price. If the company's valuation rises substantially, employees can obtain shares by exercising the options and enjoy the difference between the exercise price and the stock value; RSUs are awards in which the company promises to deliver shares or the corresponding value to employees after specific conditions are met, and usually do not require employees to pay additional exercise costs.

Both types of awards generally have vesting requirements, and employees need to serve the company for a certain period before gradually obtaining the corresponding rights. However, for stock options that have already vested, employees can usually still exercise them within the post-departure period specified in the grant agreement; shares that have actually been obtained generally will not be reclaimed solely because of departure.

Anthropic's prospectus discloses that in 2024, the company granted a cumulative total of about 1.28 billion stock options, with an average exercise price of only $1.30 per share. As of September 30, 2026, the company had about 1.082 billion vested and exercisable options, and based on the estimated fair value of the stock at that time, their total intrinsic value was as high as $72.576 billion.

However, this money is only the theoretical spread on all of the company's vested options. It does not mean that employees have already obtained an equivalent amount of cash, nor can it be used to calculate the personal wealth of Cat Wu and others.

But it is enough to show that Anthropic's equity incentives have already accumulated enormous potential value.

In April this year, according to foreign media reports, Anthropic completed a round of employee share sales, allowing some employees to sell company equity to investors. This transaction used the pre-money valuation of about $350 billion from the February financing round. Some investors were even unable to buy the expected number of shares because not enough employees were willing to sell.

In other words, at least some Anthropic employees already had the opportunity to convert their equity into cash before the IPO.

However, for employees who have not yet obtained shares, the situation is not so simple.

Anthropic's currently unvested RSUs usually need to meet two conditions at the same time: first, continued service, usually four years; second, the occurrence of an IPO or a qualifying change of control or other event. Therefore, even if an employee has already worked for two years, it does not necessarily mean they unconditionally own the corresponding shares.

If they choose to leave early, whether they can retain awards that have already met the service conditions, and how much unvested equity they will lose, all need to be judged according to the individual grant agreement. How much equity compensation Cat Wu, Thariq Shihipar, and Anthony Morris specifically received, and whether they have already cashed out, is currently not public information.

They may also give up a considerable amount of future gains because of leaving early.

On November 4, Cat Wu personally responded to a speculation related to the IPO.

A netizen asked Bloomberg finance columnist Matt Levine: if an employee leaves just before the IPO, can they bypass the post-listing stock lock-up period and sell shares immediately?

Cat Wu replied in the comments section: "This was not a factor in my decision to leave."

Anthropic's prospectus actually stipulates that some former employees are still subject to the lock-up period on share sales after listing, and cannot automatically avoid the restriction by leaving early.

In addition, unlike China, which allows post-departure non-compete restrictions to be agreed upon within the legal scope, California in principle prohibits such agreements. Employees can usually join competitors directly, but must still comply with trade secret and confidentiality obligations.

Anthropic also acknowledged in its prospectus that competitors have directly solicited its employees, and departing employees can usually join competing companies immediately.

Anthropic can even foresee that this talent competition will continue.

In the same section, Anthropic further lists employee wealth accumulation as a potential talent retention risk.

The company specifically warns that as the listing is completed, the equity held by some senior employees and early employees may have fully or largely vested, and its value may also account for a large part of personal wealth. This may make it harder for the company to continue retaining and motivating them.

What is certain is that Anthropic is growing from a smaller research institution into a large tech company. As personnel and business expand, the company needs to handle more complex organizational management, commercialization, and talent competition issues.

Cat Wu's move to OpenAI this time also gives the talent retention risk in Anthropic's prospectus a concrete real-world case.