Overview
- Google positioned Gemini Intelligence as Android’s system AI layer for cross‑app automation, and says it will launch first on Pixel 10 and Galaxy S26 phones later this year.
- Official requirements include Gemini Nano v3 on‑device model, AI Core support, a flagship chip, at least 12GB of RAM, AVF and pKVM, five OS upgrades, six years of security updates, and quality bars that exclude recent models like Pixel 9 and Galaxy Z Fold 7 that are listed on Nano v2.
- A teardown of the Google app reveals a skills system, a task scheduler, and an always‑running foreground service, along with server‑side device and language checks that let Google gate and expand access without app updates.
- Leaked screens point to an always‑on agent called Gemini Spark that can run multi‑step tasks across apps and create reusable “skills,” with onboarding warnings that the experimental feature may access linked apps and websites and could share sensitive data or make purchases without confirmation.
- Reports suggest a staged rollout after Google I/O next week, raising practical questions for users about privacy controls and whether they will need to upgrade hardware to get the full experience.