WebArena author Shuyan Zhou joins Meta Superintelligence Labs to build an AI browser

Shuyan Zhou, the author of WebArena, has announced that she is joining Meta's Superintelligence Labs and has named her core focus: building an AI browser, according to a September 22 report from the Chinese outlet QbitAI.

WebArena was a product of her PhD work. It evaluates agents in environments built to resemble the real web, where a model has to decide which page to open, where to click, what to type, and how to recover when a step fails — a different test from running agents on simplified pages or static datasets.

The motivation behind that work is personal. The report notes her goal of an AI browser driven by conversation that can complete real tasks such as buying tickets, turning 'say what you want' into 'the errand is already done.'

Across the industry, the browser is becoming a key entry point for personal agents: search, price comparison, checkout and form filling all happen on web pages. The teams that can reliably finish tasks in real web environments hold the front line of agent deployment.

That is also where the hard problems cluster. Real pages change, logins and permissions get in the way, tasks require long-horizon planning, and failures have to be detected and reversed — challenges that go well beyond language ability and into agent engineering.

Meta's Superintelligence Labs has been recruiting heavily in frontier models and agents, and the arrival of a researcher best known for web-agent evaluation suggests the direction is climbing in priority inside the company.

The question to watch is whether this research line moves out of papers and evaluation harnesses into Meta's existing product surfaces, and when ordinary users will actually feel the edge of an AI browser.