Catch up on 12 major I/O 2026 moments

Source: Google Date: 2026-05-28 URL: https://blog.google/innovation-and-ai/technology/ai/io-2026-keynote-moment-videos/

Summary

Google I/O 2026’s headline cluster was model and agent launches: Gemini Omni (multimodal video generation from any input), Gemini 3.5 Flash (positioned as the new default for agent workloads globally), and Gemini Spark (a 24/7 background personal agent across Gmail, Docs, and Slides). The remaining nine moments covered search agents that monitor the web on your behalf, a generative UI layer in Search, a universal shopping cart spanning Google’s product surface, Android XR eyewear for fall 2026, SynthID watermarking expansion into Search and Chrome, and Gemini for Science connecting to 30+ life science databases. The common thread is Google deploying agents and generative UI as first-class product surfaces, not research previews.

Implications

• ACP/host-slot thread. Gemini Spark and the Search information agents are Google’s answer to background agent slots — persistent, credentialed, operating across services. This is a meaningful data point on how a vertically integrated platform handles the host-slot problem without needing a neutral protocol.
• Open-frontier vendors thread. Gemini 3.5 Flash becoming the default for agent work globally is competitive pressure on open-weight vendors whose value proposition is cost-at-inference. Google’s distribution advantage (Gemini app, Search AI Mode, API) compresses the window for challengers.
• Model-layer holding thread. Gemini Omni’s video-from-any-input framing extends the multimodal frontier further than text+image. If video becomes a first-class reasoning input/output, the capability gap between frontier and open-weight models widens again.
• Watch: whether Gemini 3.5 Flash’s agent pricing undercuts the mid-tier open-weight API market, and how quickly Gemini Spark’s background execution model is adopted vs. Google’s prior agentic launches that stalled in preview.