Google I/O 2026 begins today at 10 AM Pacific (17:00 UTC) from the Shoreline Amphitheatre in Mountain View, with CEO Sundar Pichai delivering the keynote. Based on pre-event reporting from Gadget Bridge and TechGenyz, agentic AI is the throughline across nearly every product announcement. Google is embedding agents across Android, Search, desktop computing, XR, and Workspace simultaneously — distributing capability across the entire product ecosystem rather than concentrating it in a single interface.
Project Remy and Agentic Gemini
The most significant expected announcement is Project Remy, a proactive agent capable of multi-step operations: handling emails, organizing appointments, making reservations, and executing workflows without step-by-step human prompting. According to Gadget Bridge, this represents Google’s clearest move into the same territory as OpenClaw and Anthropic’s Claude Cowork, but through a different approach. Where OpenClaw runs as a local orchestrator and Cowork operates within Anthropic’s own interface, Remy would live inside Google’s existing product surfaces.
A Gemini Desktop Agent is also expected, allowing Gemini to interact directly with desktop applications. Combined with Project Mariner (Google’s browsing agent, first previewed in late 2025), this positions Google with agent capability across browser, desktop, and mobile contexts.
Android 17 and Gemini Intelligence
Android 17 is expected to ship with what Google internally calls “Gemini Intelligence,” an AI framework integrated into core OS workflows rather than layered on top as a standalone feature. TechGenyz reports expected capabilities include AI-powered multitasking, context-aware actions that span multiple apps, AI-generated adaptive widgets, and proactive scheduling and booking features.
The hardware requirements hint at the compute demands: Gadget Bridge reports Gemini Intelligence will require a minimum of 12GB RAM and high-performance CPUs for local execution. That requirement alone filters out a large portion of the active Android install base and signals that Google views on-device agentic capability as a premium feature, at least initially.
Aluminum OS and the Desktop Bridge
Google is expected to announce Aluminum OS, a desktop operating system based on Android that replaces the ChromeOS/Android split. Gadget Bridge reports it will power a new line of “Googlebook” laptops manufactured by Acer, ASUS, Dell, HP, and Lenovo. If Gemini Intelligence is baked into Aluminum OS the same way it’s being integrated into Android 17, Google would have a unified agent-capable operating system spanning phones, tablets, and laptops.
The Platform Play
The strategic logic is distinct from what OpenAI and Anthropic are doing. OpenAI is consolidating around ChatGPT as the primary surface for agent capability, with Greg Brockman now overseeing a unified ChatGPT/Codex/API product organization. Anthropic is tightening Claude access around its own surfaces and billing tiers. Google is going the opposite direction: distributing agentic AI across an ecosystem of surfaces that already have billions of users.
The bet is that agents embedded in the OS, the browser, the email client, and the search engine will generate more engagement and data than agents accessed through a single chat interface. For builders, the question is whether Google’s distributed approach creates more integration points (and more opportunity) or more lock-in to Google’s specific agent APIs and hardware requirements.
The keynote livestream is available on Google’s YouTube channel. NCT will cover confirmed announcements in subsequent cycles as they break.