OpenAI announced that it is launching a research preview of Operator, an AI agent that can take control of a browser and perform tasks.
Instead of relying on specialized APIs, the system uses screenshots for visual input and virtual mouse and keyboard actions to complete tasks.