Google on Tuesday, May 12, announced Gemini Intelligence, a proactive AI layer that will roll out to select Samsung and Google phones starting this summer, with broader device availability later this year. The headline capabilities — automate complex tasks, summarize web content, simplify form filling — are not, individually, new. The framing is.
Gemini Intelligence is the umbrella name for what Google is doing on Android while Apple Intelligence is being mocked for what it isn’t doing on iPhone. The product is positioned not as “AI features inside individual apps” but as “an AI layer inside the operating system.” That distinction matters because, on the operating-system layer, the AI can see everything, act across everything, and — crucially — initiate behavior the user did not specifically ask for.
That is the entire pitch. The phone doesn’t wait for you to tap something. It anticipates, summarizes, fills in, and presents you with the result. You stop tapping. Gemini does the tapping.
Whether this works is the next several months’ question. The previous generation of “AI on your phone” — Apple Intelligence’s iOS 18.1 launch eighteen months ago, the early Samsung Galaxy AI demos, the now-archived Humane AI Pin — set a low bar for “actually useful” and a high bar for “actively annoying.” Gemini Intelligence inherits both.
The launch lands in a week when Google was already telegraphing a full Gemini push at I/O the following week — Gemini 3.5 Flash, the multimodal Omni model, and Gemini Spark, the new general-purpose agent. The Android announcement is the consumer-facing surface those products will eventually flow through.
What we want to know:
- Which Samsung and Google phones make the launch list, and which sit out.
- How aggressive the “proactive” behavior is in default settings. (Aggressive = useful or annoying, depending on the day.)
- Whether the Gemini Intelligence layer can be turned off cleanly, or whether it’s the operating system now.
- What happens to battery life when an AI is doing speculative work in the background of every app.
The right consumer test, when Gemini Intelligence shows up on a real device, is whether it gets between you and the thing you wanted to do. If it does, this is the Humane AI Pin of operating systems. If it doesn’t — if it actually anticipates correctly more often than it doesn’t — Apple has a problem.