Cat Wu announced on X on November 3, 2026, that she had left Anthropic, where she served as head of product for Claude Code, closing a two-year stretch in which the coding tool moved from an internal experiment toward more autonomous software development.
She thanked the teams behind Claude Code, Cowork and Tag, describing the products as the sum of thousands of ideas from colleagues. She also credited users’ feedback with shaping much of what the teams shipped. Wu characterized helping build Claude Code into an industry-defining product as the journey of a lifetime.
From Research Product Management to Claude Code
Wu’s March 19, 2026, account of her product-management work showed how she reached that role. She joined Anthropic in August 2024 as a product manager on the Research PM team, which connected researchers with real-world customers to improve models. Before that, she worked as a product engineer at Scale AI and Dagster, then as a venture capitalist, continuing to write code that automated company-announcement monitoring and identified open-source projects gaining momentum.
When Claude Code became available internally in fall 2024, she used it to build Streamlit applications that analyzed large-scale user feedback, run evaluations to assess potential benchmarks and create reinforcement-learning environments to better understand training. Wu reported that these projects took hundreds of hours of prompting with Sonnet 3.5 (new), without her writing any of the code by hand.
She also tested successive models with the same exercise: adding a table tool to Excalidraw. The attempts began with Sonnet 3.5 (new) in October 2024, when Claude still failed. Wu reported that Opus 4 occasionally succeeded in June 2025 and that Opus 4.6 later completed such feature requests reliably enough for live demonstrations before thousands of professional developers.
Her account connected that progress to changes in the team’s working methods. Designers shipped code, engineers made product decisions and product managers built prototypes and evaluations, while short self-directed experiments and demos displaced longer documentation-first cycles. Claude Code on Desktop, the AskUserQuestion tool and todo lists were among the features that emerged from those experiments.
The Autonomy Shift She Leaves Behind
Wu’s farewell contrasted Claude Code’s early reliance on manually approved permission prompts and uses such as answering questions and writing tests with the current default she described: trusting Claude’s judgment to auto-approve, conduct large migrations and build new products. Anthropic’s auto-mode announcement, published August 7, 2026, and updated October 22, 2026, documents the product mechanics behind that shift.
The initial announcement set August 14, 2026, for new sessions on Pro, Max and Team plans to run in auto mode. The October update said the default would roll out during the following week for new sessions in most Claude apps on those plans. Instead of asking users to approve each action, auto mode routes tool calls through a classifier intended to block irreversible or destructive actions and actions aimed outside the user’s environment, while allowing safe actions to proceed.
When the classifier blocks an action, Claude can seek a safer route or ask for approval. Three consecutive blocks, or 20 across a session, return Claude Code to manual approvals. Anthropic cautions that the classification systems do not eliminate risk and recommends that users review Claude’s actions themselves for high-stakes changes to production infrastructure.
For Claude Code users, the practical change is a redistribution of oversight: routine actions can move forward with fewer interruptions, while high-stakes production changes still call for direct human review. That balance between autonomous execution and meaningful supervision remains central to the product Wu leaves behind.
Credit: Source link
