Anthropic Pulls Ahead in the AI Agent Race

In the tight race between well-funded developers of conversational AI, Anthropic has suddenly made some real gains.

The startup on Tuesday announced new software that completes tasks for people, such as building a website or editing a spreadsheet. The feature, called “computer use,” means that Anthropic’s Claude model can actually view users’ screens and take actions for them like moving a cursor, clicking buttons and typing text.

(Computer use works by taking a screenshot of your screen, deciding what actions need to be taken, executing on those actions and then taking another screenshot to decide what the AI model should do next, a spokesperson explained to me.)

If this sounds familiar, it’s probably because we’ve reported that OpenAI has been working on similar software for months—and Anthropic seems to have beaten them to the punch, despite Anthropic’s reputation for moving a bit slower due to safety concerns.