Illustration: AI
Cat Wu, who led product for Claude Code at Anthropic, has joined OpenAI to head product management for its Core Products division (unite.ai). She announced her departure from Anthropic on November 3 and her new role two days later. OpenAI president Greg Brockman publicly welcomed her arrival.
Who is Cat Wu, and why did her move draw attention?
Wu joined the Claude Code team about two years ago, when it was a small internal tool at Anthropic. Today, Claude Code is one of the best-known AI coding agents: it writes and edits code, runs commands and handles multistep tasks based on a user’s description. In her farewell message, Wu thanked the Claude Code, Cowork and Tag teams.
She said her goal in the new role was to build agents that help people and teams while keeping users in control (unite.ai). At OpenAI, she will work on the core products team led by Thibault Sottiaux, who previously headed Codex, a direct competitor to Claude Code.
This is not her first attempt to leave: in the summer of 2025, Wu and Claude Code creator Boris Cherny moved to Anysphere, the company behind the Cursor editor, but returned to Anthropic about two weeks later.
Competition shifts from models to products
The move coincided with Anthropic’s preparations for an IPO: the company filed an updated prospectus on November 2, and analysts are already forecasting its credit rating after the listing. Industry observers are focusing on something else: as leading companies’ models become more similar in capability, competition increasingly hinges on whose product is easier to use in everyday work. Over the past year, Claude Code has become OpenAI’s main rival in coding agents, and OpenAI is hiring someone who helped build it.
What this means for Kazakhstan
Developers and startups in Kazakhstan use both Claude Code and Codex, which are available by subscription and through APIs. The raem.kz editorial team also works with both: they underpin its news feed automation and website development. For these users, Wu’s move signals that the AI coding agent market remains in flux: features, prices and terms from both companies will change quickly.
The practical takeaway is simple: avoid tying workflows to a single tool, keep tasks and agent instructions in a portable format, and compare agents on your own real tasks every few months.
For AI teams in Kazakhstan building their own models and services, there is another lesson here: the winner is not whoever has the strongest model, but whoever turns it into a product people use every day.
