openaienmodel: gpt-5-mini-2025-08-07
Let Gemini handle your multi-step daily tasks on Android.
Key Points
- Beta automations for multi-step tasks
- Runs selected apps inside a secure virtual window
- Invoke via long-press power on Pixel 10 and Galaxy S26
Summary
Gemini's Android app introduces a beta automation that can perform multi-step, app-based tasks on supported phones. Users invoke Gemini by long-pressing the power button and asking it to complete tasks such as reordering a meal or booking a ride; Gemini runs the target app in a secure, virtual window and proceeds in the background while the user keeps using the phone.
Key Points
- Invocation: Long-press the power button to start an automation; actions begin only after explicit user command and stop when the task finishes.
- Execution model: Gemini automates by running the required app in a secure, virtual window with limited access to the device.
- Transparency & control: Live notifications show progress and let users view, jump in, or stop the task.
- Scope & availability: Beta (18+), initially available in the U.S. and Korea on Pixel 10 / Pixel 10 Pro / Pixel 10 Pro XL and Samsung Galaxy S26 series; supports select food, grocery, and rideshare apps.
- Privacy & safety: Designed to limit app access and emphasize user supervision—"Supervise closely, interrupt when needed."
- Engineering notes: Test on supported devices, validate app compatibility and window isolation, verify notification hooks and interruption behavior, and collect user feedback for expansion.