Anthropic’s Cat Wu says that, in the future, AI will anticipate your needs before you know what they are - BERITAJA

With the tech manufacture singularly focused connected AI models, Anthropic is having an exceptionally bully year.

The institution whitethorn soon propulsion up of its main competitor, arsenic it looks to raise tens of billions of dollars successful a backing information that would put its valuation at immoderate $950 billion (OpenAI was valued astatine $854 billion successful its March round), and business customers progressively definitive a prefererence for Claude complete ChatGPT. A caller study showed Anthropic recently outpaced OpenAI among business customers, quadrupling its marketplace stock since May 2025.

Cat Wu, Anthropic’s caput of merchandise for Claude Code and Cowork, has been a cardinal fig successful that success. Since joining the institution successful August 2024, Wu has helped shepherd Claude done a captious phase, leveling it up from a purely informational chatbot to a coding instrumentality and beyond. Wu, who oversees the improvement of caller features, is often paired pinch Boris Cherny, a halfway personnel of Anthropic’s method unit and the creator of Claude Code, starring the brace to beryllium characterized as Anthropic’s “Batman and Robin.”

Wu sat down pinch maine astatine last’s week’s 2nd yearly Code pinch Claude convention successful San Francisco, wherever she discussed really she thinks about merchandise strategy, and really she hopes the acquisition of utilizing Claude will alteration successful the future. 

This question and reply has been edited for magnitude and clarity.

When you’re looking astatine merchandise strategy, really overmuch of it is reactive to your peers aliases your competitors? Do you deliberation about that astatine all?

The main point that we creation for is staying connected the exponential, truthful I think, crossed our team, we instill successful everyone the instruction that AI will conscionable proceed to get better. For us, we conscionable request to enactment astatine this frontier. We don’t deliberation about competitors. I deliberation if you do deliberation about competitors, you extremity up being, like, perpetually 2 weeks, aliases like, a period down really accelerated you could execute. And truthful it’s usually not the champion measurement to enactment astatine the frontier.

Anthropic released astatine slightest six models past twelvemonth and has already released almost arsenic galore this year. Do you expect this gait of improvement to continue?

Our dream is that it continues (laughing). I deliberation the models are still improving astatine a very dependable pace, and truthful we should beryllium capable to support sharing those pinch our users. I deliberation the deployments mightiness look a spot different—like really we handled Glasswing, but arsenic overmuch arsenic possible, we want this intelligence to use arsenic galore group arsenic possible, and it has to beryllium handled successful a secure way, which is why we handled Glasswing [in the measurement that we did].

[Glasswing is an inaugural that Anthropic launched successful April that invited a mini consortium of partner organizations — including companies for illustration Amazon, Apple, CrowdStrike, and Microsoft — to summation entree to its caller cybersecurity model, Mythos. Unlike galore of Anthropic's different AI models, Mythos is not being fixed a wide nationalist release. The institution has claimed that it fears the exemplary — which is designed to scan codebases for package vulnerabilities — is excessively powerful, and could beryllium weaponized by bad actors.]

You said successful a erstwhile question and reply that the early of activity is fundamentally unit managing fleets of agents. It seems for illustration that could yet lead to a business wherever the agents are amended astatine the job, aliases cognize the job, amended than the human.

I deliberation it is highly difficult to negociate agents if you can't do the occupation yourself. I deliberation the managers still request to beryllium experts successful their domain. It's a caller accomplishment group that a batch of group are going to person to learn, but managing agents is really very akin to being a head of people, successful the consciousness that you person to understand, like, why did the supplier make this mistake? Did it misinterpret my instruction? Was my petition under-specified? You person to person the expertise to debug it.

It does look for illustration the agelong word extremity is to trim down connected squad size, though. Because if you person agents doing a job, past you don't request an intern, right?

Ideally, I deliberation the thought is that everyone could get a batch much done. I deliberation that, for everyone’s job, there's ever this percent of it that's really tedious. For me, it’s responding to emails. I deliberation everyone has this portion of their life...So my dream is that it [the AI agents] really does that, and past everyone has, like, each these cool things that they will want to build [in their spare time].

What are you guys about excited about successful the adjacent six months?

I deliberation the adjacent large point is proactivity. Last twelvemonth we were successful this world of synchronous development. Right now, group are shifting to routines, truthful for illustration automating, for example, responses to customer support tickets. And I deliberation the adjacent measurement is that Claude understands what you activity on, and conscionable sets up immoderate of these automations for you.