What changes for Mac users
Google has introduced the Computer Use feature in the Gemini app for Mac. According to The Decoder, the user describes a goal and the model handles clicking, typing and navigating menus and forms to complete complex tasks involving local files, desktop applications and websites.
The distinguishing element compared with a simple conversational assistant is the actual execution of actions on the computer. Google claims that everything takes place in an isolated background window, with a live preview in the chat: the screen and cursor thus remain free, and the user can continue working without interruption.
Controls and use cases cited
The source material lists several safeguards: the user chooses which applications Gemini can access, can follow every step in real time, pause the operation at any time and pre-approve actions considered critical. These are Google statements reported by the source, not independent verifications of actual functioning.
Among the examples cited are batch editing of photos and 3D scenes, filling out shipping forms starting from PDFs on the device, and building presentations. These are illustrative scenarios provided by the company: the material does not indicate measured performance, error rates or operational limits found in external tests.
Availability and comparison with OpenAI
The launch begins today in English in the United States; the Gemini app for Mac is available for download. The material does not indicate timelines for extension to other languages or markets, nor a schedule for additional countries.
The source notes that OpenAI has also been working on Computer Use features for some time and that, in its assessment, it would be further ahead with the Astra model. This is the publication's judgment, not a verifiable fact in the text: comparison metrics between the two systems are in fact missing. The significance of Google's move should therefore be read as opening a competitive front on agents that operate directly on the desktop, an area that has so far been less covered than dialogue-based assistants.
Aspects not covered by the source remain to be clarified: which system permissions the isolated window requires, how local file data are handled, and which applications are actually compatible at launch. In the absence of these elements, the news confirms the launch of the feature but does not allow assessments of its practical reliability.
