What happened
Google has started rolling out a computer use capability in the Gemini app for Mac, TNW | Apple reported. The feature lets the assistant complete tasks that span several steps by clicking, typing and completing forms on a user's behalf. Google announced it on Thursday. Gemini works in its own background window and shows a live preview inside the chat, so the user's screen and cursor remain available for other work.
Users pick which apps Gemini may access, can pause it at any point and have to sign off on important actions. Google's examples include moving weights and addresses from PDF files into a shipping quote form, converting dashboard figures into a Google Slides presentation, editing batches of photos and drafting resale listings from images stored in a folder.
Access is limited to subscribers on Google's paid AI Plus, Pro and Ultra plans using a Mac with Apple Silicon. As with the Gemini Spark agent, the feature is unavailable in Europe and the UK. Google cautioned that Gemini can err, for example by erasing files permanently or sending them to others. The tool is built to request permission first and can view only connected folders. Backups of altered files are removed after 24 hours or once a new task begins.
Context
Google first added computer use to Gemini 3.5 Flash for developers in June, and it is also embedding Gemini at the operating-system level of its Googlebook laptops. Rivals are pursuing desktop agents as well: OpenAI's ChatGPT can capture what users type on a Mac to inform its memory, and agents running on the OpenClaw framework have contributed to stronger Mac mini demand.
简报
一封邮件,尽览一周科技要闻。
每周精选值得一读的报道。可随时退订。
