Google reframes itself around agents at I/O 2026

verified

Thursday, May 21, 2026

The News

At Google I/O 2026 in Mountain View on May 19–20, Google unveiled Gemini Omni, the faster Gemini 3.5 Flash, and Antigravity 2.0, its agent-first coding platform built for autonomous AI tasks, with CEO Sundar Pichai opening the keynote on Gemini launches including Antigravity . Pichai revealed the company is now processing more than 3.2 quadrillion tokens per month, a massive increase from 480 trillion tokens per month at I/O 2025 . Antigravity 2.0 features an Editor view that functions like a familiar IDE interface with an agent sidebar, and a Manager view that serves as a control center for orchestrating multiple agents working in parallel across workspaces . The Gemini app now has over 900 million monthly active users, up from 400 million a year ago. Full announcements are catalogued on Google's blog and the Pichai keynote.

The Read

Google didn't ship one product at I/O; it shipped a thesis — that the developer surface area is the agent harness, not the model. The Antigravity 2.0 / Manager-view structure is a direct repositioning against Cursor and Claude Code, both of which still treat the IDE as primary and the agent as a sidecar. The reason it works is buried in the token-volume disclosure: Google was processing half a trillion tokens a day internally across its AI developer tools in March, doubling every few weeks, and is now processing more than three trillion a day — that's not user growth, that's agents calling agents inside Google before any of this ships externally. The competitive question isn't "is Gemini 3.5 Flash good enough." It's whether anyone outside the hyperscalers can absorb a 7× year-over-year jump in agent token consumption without renegotiating their model contracts mid-year.

Counterview

Antigravity is still a Google-stack lock-in play; teams running multi-model production agents have more reason than ever to keep their harness vendor-neutral.

Do This Week

Pull the last 30 days of your team's model-API spend and break it into "human-in-the-loop calls" vs. "agent-to-agent calls." If the agent-to-agent share has crossed 40% and is still growing, you are on the same curve Google described — start sizing what your spend looks like at 3× and have the answer before your next renewal.

For Developer