Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseInsight · On RadarThe Verge→ original

Gemini for Mac Learns to Control Your Computer: Files, Apps, and Websites

The Gemini app for macOS can now control your computer, carrying out multi-step tasks across files, apps, and websites. Google suggests handing it form filling, presentation building, and listing items for sale. Users choose which apps Gemini can access and follow its work in a background window with a live preview. The feature is currently available only in English and only in the US, so everyone else will have to wait.

AI-processed from The Verge; edited by Hamidun News
Gemini for Mac Learns to Control Your Computer: Files, Apps, and Websites
Source: The Verge. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceThe Verge ↗
Editor-in-Chief
Key metric #1Access is granted per app, not to the entire computer at once
Key metric #2Apple is testing restrictions on "full disk access" in the macOS 27.3 beta
The rollout is underway nowEnglish only and US users only

Chatbots have long been able to advise on how to fill in a form or put together a presentation. Now Google is trying to have the AI do it itself. The Gemini app for macOS has received a computer use feature: the assistant can carry out multi-step tasks across local files, desktop programs, and the browser. This is a shift from "tell me how to do it" to "do it for me," and it matters for developers and product managers: AI agents are moving out of cloud windows and onto the user's real desktop.

What Gemini on Mac Can Actually Do

Google describes the feature as help with tasks that require several steps in a row. The examples the company cites include speeding up workflows, filling in forms, creating presentations, and listing products for sale. A typical scenario looks like this: you have a folder with photos and a description of an item, and you need to create a listing. Previously you copied text from the chat into a form by hand. Now the assistant can open the site itself, enter the data, and move through the fields.

An important detail is that the work does not happen blindly. Gemini has its own background window, and a live preview of what is happening appears right in the chat. The user can see which buttons the agent clicks and what it types, and can stop it if something goes wrong. Background mode also means the assistant does not have to take over the whole screen while you are busy with something else.

Access Control and Apple's New Rules

The key question for any agent with access to a computer is what it can see and do. Google says the user decides which apps Gemini gets access to. This is closer to granting permissions to specific programs than to handing over a master key to the system.

The timing is also notable. According to The Verge, Gemini is launching this feature just before Apple introduces new restrictions for apps that request "full disk access" on Mac. The reason is the potential risks from AI agents with computer use. A preliminary version of these settings appeared in the macOS 27.3 beta, released this week. The source provides no details on how the restrictions will affect Gemini or other agents, so it is too early to draw conclusions about compatibility.

For the market, this is a signal: the platform where agents run is starting to set the rules for them. If Apple tightens file system access, every developer building agents for macOS will have to think through more precise permissions and show users exactly what the program is doing.

Limitations and What Practitioners Should Do

The feature is not available to everyone yet. The rollout is underway, but only in English and only for users in the US. If you live in another country or work with Russian-language interfaces, you cannot yet evaluate how well it performs. Google's announcement also does not give figures on accuracy or task completion speed, so it is too early to talk about the agent's reliability in complex scenarios.

What teams should take into account:

  • If you are building a desktop product with agents, use the Gemini model as a reference: per-app access and a process that is visible to the user. This approach will likely become the expected norm, especially given Apple's restrictions.
  • For testing, choose safe scenarios: forms, presentation drafts, preparing listings. Payments and important documents are better left out for now.
  • Keep an eye on the macOS 27.3 beta. Disk access rules may change before agents become mass-market.

Competition in the segment of agents that control computers is only picking up speed, and Gemini on Mac shows that major players are moving it from demonstrations into everyday consumer apps. How well such assistants handle real routine work will become clear after the first weeks of use in the US.

Story

Key Questions Answered

What Gemini on Mac Can Actually Do?

Google describes the feature as help with tasks that require several steps in a row. The examples the company cites include speeding up workflows, filling in forms, creating presentations, and listing products for sale. A typical scenario looks like this: you have a folder with photos and a description of an item, and you need to create a listing. Previously you copied text from the chat into a form by hand. Now the assistant can open the site itself, enter the data, and move through the fields.

What are the key details on Access Control and Apple's New Rules?

The key question for any agent with access to a computer is what it can see and do. Google says the user decides which apps Gemini gets access to. This is closer to granting permissions to specific programs than to handing over a master key to the system. The timing is also notable. According to The Verge, Gemini is launching this feature just before Apple introduces new restrictions for apps that request "full disk access" on Mac. The reason is the potential risks from AI agents with computer use.

What are the key details on Limitations and What Practitioners Should Do?

The feature is not available to everyone yet. The rollout is underway, but only in English and only for users in the US. If you live in another country or work with Russian-language interfaces, you cannot yet evaluate how well it performs. Google's announcement also does not give figures on accuracy or task completion speed, so it is too early to talk about the agent's reliability in complex scenarios.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…