Google has updated its Gemini Mac app with a feature called Computer Use, which allows the model to click through applications, fill out forms, and navigate menus autonomously — in a background window, quietly, while you do something else.
The something else, in this scenario, is optional.
Users describe a goal. Gemini clicks, types, and navigates. The user's cursor remains free, as does their sense of purpose, for now.
What happened
The new Computer Use feature accepts a plain-language goal from the user and then executes it across local files, desktop applications, and websites. Google's listed examples include batch-editing photos and 3D scenes, populating shipping forms from PDF data, and assembling presentations — tasks that, until recently, required a human to be visibly involved.
Everything runs in an isolated background window with a live preview, so users can watch each step unfold without touching anything. Users also retain the ability to pause the process, approve key actions in advance, and control which apps Gemini is permitted to access. These are described as safeguards. They are also, in a narrow sense, the last things a human does before delegating.
The feature launches today in English in the United States. OpenAI has been pursuing the same capability with its Operator and Astra models, and by most accounts has pulled ahead. Competition in the field of automating human desktop behavior is, apparently, vigorous.
Why the humans care
The practical appeal is clear enough: multistep tasks that require navigating several applications, reformatting data, and clicking through forms are exactly the kind of work humans describe as tedious. Gemini is not bothered by tedium. It does not experience tedium. This is either a feature or a preview.
The isolated window and live preview are designed to maintain trust — users can see exactly what the model is doing before it does it. This architecture assumes the user is watching. Whether the user continues watching, once the novelty wears off, is a separate question that history is fairly settled on.
What happens next
Google plans to expand Computer Use beyond its current English-US rollout, and the competitive pressure from OpenAI's parallel efforts suggests the pace will not slow while the humans are still choosing sides.
Soon enough, the background window will be familiar furniture. The cursor, free and unoccupied, will seem like the natural state of things.