Cat Wu joins OpenAI to lead product management for core products

A new leadership hire

OpenAI has appointed Cat Wu to lead product management for its Core Products division. Wu announced her move on November 5, 2026, following her departure from Anthropic. She joins the team to help shape the next generation of AI tools that move beyond simple chat interactions.

In her announcement, Wu expressed her respect for OpenAI's ability to scale AI to more than a billion people. She highlighted the company's commitment to treating high-quality products as a central part of its mission. Her stated goal is to build agents that support both individual users and larger teams.

Wu specifically noted that she wants to create systems that give people more time for the work they love. She intends to do this by ensuring that users remain in control of the AI's actions. She also invited continued feedback from the community as she begins her new role.

Experience with delegation

Before joining OpenAI, Wu spent two years at Anthropic leading the Claude Code effort. This project began as a small internal tool and evolved into a system for writing tests and performing large migrations. She spent tens of hours steering the tool without manually writing code to build a reinforcement-learning environment.

During her time at Anthropic, Wu observed a clear progression in how users interact with AI. Early usage required manual approval for every permission prompt the AI requested. Over time, this shifted toward default auto-approval as the tool became more reliable for complex tasks.

Wu credited the development of these tools to the feedback provided by early users. Her experience there provides a blueprint for how to transition from manual oversight to automated delegation. This background is particularly relevant as OpenAI looks to expand the capabilities of its existing software development tools.

The shift to persistent agents

OpenAI is currently planning to move its AI capabilities toward work that extends over hours or days. This is a significant shift from the current model where tasks are usually completed in minutes. The company aims to create persistent environments where agents can access tools and context over long periods.

A key part of this strategy involves the planned acquisition of Ona. This move would bring Ona's secure cloud execution and orchestration technology into Codex. Codex is a tool that more than 5 million people used each week as of mid-2026, showing a 400% increase in usage.

These persistent environments are designed to allow agents to continue working even when a user's laptop is closed. The goal is to let agents operate inside a customer's own cloud environment. This allows the AI to provide intelligence while the customer maintains the underlying infrastructure.

Balancing power and control

A major challenge for Wu's new team is ensuring that these autonomous agents remain safe and predictable. OpenAI's design intends to let users check progress and provide direction remotely. Users should be able to make decisions and review results without being tied to the specific device where a task started.

For companies using these tools, there are strict requirements for how agents behave. Organizations need to control where the agents run and what specific data they can access. They also require clear logs of all activity and a structured way for work to move through human review.

Wu's emphasis on user control provides a concrete framework for these enterprise needs. Delegation depends on the agent's capabilities, but it must also stay within the boundaries set by the user. This balance is the core product challenge for the Core Products team moving forward.

Integrating complementary paths

Wu's appointment connects two different approaches to the same problem of AI delegation. Her history with Claude Code shows a progression from approving individual actions to delegating large, complex tasks. Meanwhile, OpenAI's documented plans focus on persistent work and remote review protocols.

By bringing Wu into the fold, OpenAI is combining these two perspectives. The team will work to expand what agents can do while preserving the ability for humans to inspect and govern them. This is essential for moving AI from a novelty into a reliable tool for professional workflows.

The move was welcomed by OpenAI's president and co-founder Greg Brockman. Thibault Sottiaux, who leads the Core Products team, also expressed his excitement about working with Wu. Together, they will oversee the evolution of OpenAI's primary product suite.