With the tech industry focused on AI models, Anthropic is experiencing a notably successful year.
The company may surpass its main competitor soon, as it aims to raise tens of billions in a funding round that would value it at around $950 billion, compared to OpenAI’s $854 billion valuation in March. Business customers increasingly prefer Claude over ChatGPT. A report showed Anthropic recently surpassed OpenAI in business customers, quadrupling its market share since May 2025.
Cat Wu, Anthropic’s head of product for Claude Code and Cowork, has been instrumental in this success. Since joining in August 2024, Wu has guided Claude from an informational chatbot to a coding tool and more. Wu, overseeing feature development, often collaborates with Boris Cherny, Anthropic’s technical staff core member and Claude Code creator, earning them the nickname Anthropic’s “Batman and Robin.”
Wu spoke at last week’s Code with Claude conference in San Francisco about product strategy and future hopes for Claude’s experience.
This interview has been edited for length and clarity.
**When considering product strategy, how much is reactive to competitors?**
We focus on staying at the exponential frontier, ensuring AI keeps improving. Thinking too much about competitors can put you behind in execution.
**Anthropic released several models last year and continues to keep pace. Do you foresee maintaining this development pace?**
We hope to continue (laughing). Models are improving steadily, allowing us to share them with users. Deployments might vary, like with Glasswing, but the goal is safe and widespread intelligence sharing.
*[Glasswing is an initiative by Anthropic launched in April, where selected partners, like Amazon, Apple, and Microsoft, accessed its new cybersecurity model, Mythos. Unlike other models, Mythos isn’t released publicly due to potential misuse.]
**In a previous interview, you mentioned the future involves managing fleets of agents. Could agents eventually outperform humans?**
Managing agents is challenging without expertise. Managers need domain knowledge, akin to managing people, to understand and debug agent errors.
**Is the ultimate goal to reduce team sizes since agents can handle tasks?**
Ideally, everyone becomes more productive. AI agents can handle tedious tasks, freeing up time for creative projects.
**What excites you most for the next six months?**
The focus is on proactivity. Moving from synchronous development to automating tasks like customer support responses. The next step is Claude understanding workflows and setting up automations.
When purchasing through our article links, we may earn a small commission. This does not affect our editorial independence.