Gemini automates apps on Pixel 10 Pro, Galaxy S26 Ultra
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google's Gemini task automation on Pixel 10 Pro and Galaxy S26 Ultra is the first version letting the AI control apps for the user, currently limited to a small subset of food delivery and rideshare services and still in beta
- Ordering a teriyaki dinner took approximately nine minutes, with Gemini occasionally making wrong turns before self-correcting; the reviewer found it surprisingly accurate when it did complete tasks
- The feature defaults to running in the background and stops short of final confirmation so users can review its work, and never completed an order without permission across five days of testing
- In a flight-booking test, Gemini used calendar and email access to suggest an 11:30 or 11:45 AM Uber for a 1:45 PM flight and scheduled the ride in about three minutes with no further input
- Sameer Samat, Google's head of Android, told the reviewer Gemini uses a visual reasoning approach because developers haven't yet adopted Model Context Protocol (MCP) or Android's app functions
- The reviewer argues current apps — designed for humans with ads, photos, and clutter — are a stopgap; AI-first interfaces built on databases would be far more efficient than reasoning through human-centric UIs
Why it matters: Google is the first major player to ship end-to-end app automation on consumer phones, but the review surfaces a key concession from Android chief Sameer Samat: visual reasoning through human-designed UIs is a stopgap until developers adopt MCP or Android app functions, putting pressure on the entire app ecosystem to rebuild for AI.
Ask SkimNews
