Anthropic's head of product for Claude Code and Cowork, Cat Wu, says the next six months in AI belong to proactivity — agents that watch what a user works on and set up automations before being asked. She made the case at last week's second annual Code with Claude conference in San Francisco, as Anthropic moves to raise tens of billions of dollars at a roughly $950 billion valuation, ahead of OpenAI's $854 billion mark set in March.
The pitch lands at a moment when Anthropic has quadrupled its business market share since May 2025 and pulled ahead of OpenAI among enterprise buyers, a shift AI Chat Daily covered earlier this month when Anthropic overtook OpenAI on Ramp business spend by 34.4% to 32.3%. Wu, who joined Anthropic in August 2024, oversees the development of new features on Claude Code alongside Boris Cherny, the engineer who built the product.
Wu frames the product roadmap as a sequence. "Last year we were in this world of synchronous development," she said. "Right now, people are shifting to routines, so like automating, for example, responses to customer support tickets. And I think the next step is that Claude understands what you work on, and just sets up some of these automations for you."
Key facts
- 01Anthropic is raising tens of billions at a $950B valuation, ahead of OpenAI's $854B March round.
- 02Anthropic has quadrupled its business market share since May 2025, overtaking OpenAI among enterprise customers.
- 03Anthropic shipped at least six models last year and nearly as many already this year.
- 04Cat Wu, head of product for Claude Code and Cowork, joined Anthropic in August 2024.
- 05Anthropic launched Glasswing in April with Amazon, Apple, CrowdStrike, and Microsoft, withholding its Mythos cybersecurity model from public release.
Anthropic released at least six models last year and is on pace to match that this year, a cadence Wu wants to maintain. "I think the models are still improving at a very steady pace, and so we should be able to keep sharing those with our users," she said. The release model itself is starting to fork: not every model will ship broadly.
“The next step is that Claude understands what you work on, and just sets up some of these automations for you — Anthropic's head of product for Claude Code, Cat Wu.”— Jaeden Schafer
The clearest example is Glasswing, launched in April as a closed program with Amazon, Apple, CrowdStrike, and Microsoft. The initiative gates access to Mythos, a cybersecurity model designed to scan codebases for software vulnerabilities. Anthropic has said it is keeping Mythos out of general release because the same capability could be weaponized against the same codebases it is built to defend.
On strategy, Wu argues Anthropic deliberately does not optimize against rivals. "We don't think about competitors," she said. "I think if you do think about competitors, you end up being, like, perpetually two weeks, or like, a month behind how fast you can execute." The company's working assumption, in her telling, is that model capability keeps compounding and the job is to ship into that curve. "The main thing that we design for is staying on the exponential."
That worldview shapes how Wu thinks about the workforce around agents. She has previously said the future of work looks like staff managing fleets of agents, and she rejects the read that agents will simply replace junior staff. "I think it is extremely hard to manage agents if you can't do the job yourself," she said, comparing it to managing people — debugging why an agent made a mistake, whether instructions were under-specified, whether the request was misread.
Pressed on whether agentic workflows shrink team sizes, Wu reframes the question around throughput rather than headcount. "Ideally, I think the idea is that everyone can get a lot more done," she said, pointing to the share of any job that is rote — email, in her case — as the part to hand off. The optimistic version of her thesis is that humans claw back time for more ambitious work; the version enterprises will test is whether they actually keep the same headcount once the rote layer is automated.
The product story sits on top of a business story that has shifted quickly. A $950 billion valuation would put Anthropic roughly $96 billion above OpenAI's most recent mark, an unusual position for a company that was viewed as the clear number two for most of 2024. Anthropic's pull with developers and with enterprise buyers — including via Claude Code, the surface Wu runs — is the load-bearing piece of that re-rating.
Wu's proactive-Claude pitch is less a feature announcement than a directional bet. It implies a Claude that sits inside a user's workflow with enough context to propose automations unprompted, which is also a thesis several smaller startups are chasing — Poppy, for one, raised a $1.25M pre-seed for a proactive AI assistant earlier this month. The difference for Anthropic is distribution: Claude Code already lives inside the workflows where proactivity would have to prove itself.
What is genuinely unsettled is whether enterprise buyers will accept an assistant that takes actions without explicit prompts. The same companies paying for Claude today bought it in part because it is predictable, auditable, and conservative on tool use. Proactivity cuts against those properties, and Wu did not say how Anthropic plans to keep its safety posture intact as the product moves from synchronous to anticipatory.
The strategic read is that Anthropic is trying to convert a model-quality lead into a workflow lock-in before OpenAI closes the enterprise gap. A proactive Claude that quietly takes over the tedious 20% of a knowledge worker's day is much harder to rip out than a chatbot a user opens on purpose. If Wu's six-month timeline holds, the next phase of the OpenAI-Anthropic race is not which model scores higher on a benchmark but which one ends up running in the background of more companies' workdays.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
