Back to Journal
Google I/O 2026: Everything Announced (And Where It's Heading)

Google I/O 2026: Everything Announced (And Where It's Heading)

Ray de Guzman

Google I/O 2026 was a two-day firehose (May 19-20, Mountain View) and the theme was unmistakable: AI just stopped being a feature you open and started being an agent that works in the background while you do other things. Gemini models got faster, the Gemini app got a personality transplant, Search got its biggest box upgrade in 25 years, Android grew a heads-up display for agent activity, and smart glasses finally got a ship date.

Here's the entire keynote in one scannable read, grouped by what it actually affects in your day-to-day. Sources are linked inline so you can dive into any rabbit hole.

Watch the full keynote · Developer keynote · Official I/O hub

Gemini models: faster frontier, native video out

Gemini 3.5 Flash is the new frontier-class model, claiming to surpass 3.1 Pro on coding, agentic, and multimodal benchmarks while running ~4x faster (output tokens/sec) than other frontier models. Rolling out today across the Gemini app, Search, Antigravity 2.0, and the Gemini API.

Gemini Omni is a new family that fuses reasoning with creation. Omni Flash takes image, audio, video, and text as input and outputs video grounded in real-world knowledge that you can edit downstream. Available on AI Plus, Pro, and Ultra via the Gemini app, Google Flow, and YouTube Shorts. Flow and Flow Music shipped as mobile apps (Flow on Android beta, Flow Music on iOS).

Provenance for everyone: SynthID detection is leaving the Gemini app and showing up in Search and Chrome, and Google added C2PA Content Credentials so you can see whether content came straight out of a camera or was touched by tools.

Sources: 9to5Google recap · Developer keynote post

The Gemini app: new look, new agent, new digest

Neural Expressive is the app's new design language. Pill-shaped prompt box, single plus menu, fullscreen nav drawer, fluid animations, haptic feedback, and a new type system. Responses now lead with the most important info in bold, with inline images, narrated videos, timelines, and interactive visualizations. Gemini Live ditches the fullscreen modal for an inline experience. Rolling out now on Android, iOS, and web.

Gemini Spark is the headline: a 24/7 personal agent that actually takes actions on your behalf. Plugged into Gmail, Docs, and Workspace at launch, with third-party tools coming via MCP over the summer. Available next week to Google AI Ultra subscribers in the US.

Daily Brief is the personalized digest version: scans your Gmail, Calendar, and Tasks and hands you a prioritized day plan with suggested next steps. Rolling out today to AI Plus, Pro, and Ultra in the US.

Source: 9to5Google

Google AI plans: pricing got rewritten

Google is moving the Gemini app from daily prompt caps to a compute-used model -- a simple text prompt costs less than a complex video or coding job. Limits refresh every five hours until you hit your weekly cap.

Google AI Ultra dropped to $100/month with 5x higher Gemini app usage than AI Pro. The old $250 tier is now $200 with the exact same capabilities. (Yes, that means everyone on Ultra just got a significant price cut.)

Source: 9to5Google

Search: the biggest box upgrade in 25 years

Sundar called it the biggest Search upgrade in over two decades. A new intelligent Search box expands as you type, anticipating intent instead of just autocompleting. AI-powered query suggestions go beyond keyword matching.

The bigger story: Search agents. Background agents that monitor the web 24/7 for whatever matters to you -- blogs, news, social, plus Google's real-time finance, shopping, and sports data -- and notify you when something changes. Rolling out to Google AI Pro and Ultra this summer.

Source: Google's Search blog post

Shopping: Universal Cart

Universal Cart is Gemini's shopping hub across the Gemini app, YouTube, and Gmail. Add a product, and the cart proactively surfaces deals, price drops, price history, and stock alerts. Building a custom PC? It flags incompatible parts and suggests alternatives. Built on Google Wallet, so it knows your payment perks, loyalty info, and merchant offers. US-only this summer in Search and the Gemini app, with YouTube and Gmail to follow.

Source: 9to5Google

Google Workspace: "Live" everywhere

  • Gmail Live: conversational email search. Pro/Ultra, US English, summer, Android + iOS.
  • AI Inbox: coming to AI Plus and Pro.
  • Docs Live: create and edit documents conversationally. Pro/Ultra, summer, English globally.
  • Keep: a similar voice-organized mode that turns free-flowing thoughts into concise notes. AI Pro/Ultra, summer, Android.
  • Google Pics: a new AI image generation and design app.

Source: 9to5Google

YouTube

  • Ask YouTube: complex search + follow-ups across the YouTube catalogue with a structured, interactive response. Premium subscribers in the US can try it at youtube.com/new.
  • Gemini Omni in Shorts Remix and the Create app: video-to-video creation lands natively in the Shorts pipeline.

Source: 9to5Google

Android XR: smart glasses, finally with a ship date

Google is calling the category intelligent eyewear and splitting it in two:

  • Audio glasses: spoken Gemini help in your ear. Launching first, later this fall. Hardware by Samsung and Qualcomm, frames by Gentle Monster and Warby Parker.
  • Display glasses: screens that surface info right when you need it. No firm ship date yet.

Both pair with Android phones AND iPhone. Live demos showed real-time translation, navigation, and visual lookups.

Sources: Android XR official post · PCMag's smart glasses hands-on

Android Halo: agent activity, always visible

Android Halo is a new system-level UI surface that gives you at-a-glance visibility into what your agent is working on -- a subtle indicator at the top of any screen, so you can let Gemini Spark or another agent run a task without having to leave what you're doing to check on it. Available later this year.

Source: 9to5Google

For developers: agents move from demo to default

  • Antigravity 2.0 + the new Antigravity CLI: Google's agent-first development platform now lets you spin up specialized subagents for complex workflows, with built-in cross-platform terminal sandboxing, credential masking, and hardened Git policies.
  • Google AI Studio: now includes native Kotlin support so you can vibe-code Android apps directly inside Studio.
  • Gemini 3.5 Flash in the API: the new frontier model is immediately available to build on.

Source: Developer keynote blog post

Google Play and Android

  • Google Play: new tools to grow reach beyond the store, simplify day-to-day operations, and safeguard against abuse. Full I/O 2026 Play recap on the Android Developers Blog.
  • Android 17 + Gemini Intelligence: task automation across apps lands as a system-level feature. Plus updates to Android Auto, the new Googlebook laptop platform, and core experience polish. Full breakdown on the Android Show: I/O Edition page.

Quick scoreboard

Available now This summer This fall / later 2026
Gemini 3.5 Flash, Gemini Omni Flash (AI Plus+), new Gemini app design, Daily Brief (US), Google AI Ultra at $100, Antigravity 2.0, Flow + Flow Music apps Gemini Spark (US Ultra, next week then expanding), Search agents (Pro/Ultra), Universal Cart (US), Gmail/Docs/Keep Live, Ask YouTube (US Premium) Android XR audio glasses, Android Halo, display glasses (date TBD)

Best long-form recaps if you want to go deeper

My Take

The feature that actually excited me: building Android apps without the usual hell

For years, building an Android app meant learning Kotlin properly, setting up Android Studio, wrestling with Gradle, navigating the Play Store developer dashboard -- before you'd even written a single line of real logic. The barrier wasn't the idea, it was the setup cost.

Google AI Studio now lets you describe what you want and it builds a native Android app using Kotlin and Jetpack Compose. You can preview it in a browser-based Android emulator while you're building, then install it on your actual phone via USB. That part is fully live.

As for the Play Store: technically you can push to an internal testing track -- AI Studio handles the app bundle and record creation. But public publishing (even just sharing with friends) is still described as "on the roadmap" by TechCrunch. So the dream of "build it and ship it" is mostly there, but not fully cooked yet. Still -- the install-to-my-own-phone path alone would unblock everything I've wanted to build personally. This is the announcement I'm watching most closely. Source: Android Developers Blog · TechCrunch

Then I went on Twitter, and the vibe shifted

The initial excitement didn't survive contact with the feed.

The core complaint: Gemini 3.5 Flash is positioned as the main model, and it's Google's fast/cheap tier -- not a raw intelligence leap. Google claims it's "comparable to the best models" on benchmarks, but that's not the same as leading them. No clear "GPT-4-level jump" moment. For a company that was supposed to be AI-native from the start, the response felt like "we caught up" more than "we pulled ahead."

The pricing narrative also landed weird. Calling $100/month for Ultra a deal is technically true (it was $250) -- but you're still paying $100/month for a product that isn't clearly best-in-class. The compute-used model sounds flexible until you realize limits are reset-based and you can burn through your weekly cap on one complex video.

Gemini Omni vs Seedance 2.0: what I actually saw in the comparisons

Gemini Omni Flash looked impressive in Google's own demos. Video-to-video editing, multi-turn editing, native audio -- the feature list is real.

Then I saw the side-by-side comparisons with Seedance 2.0 (ByteDance) and Kling 3.0, and the gap is noticeable. Seedance and Kling are producing footage that looks more cinematic -- better lighting physics, more believable motion, character consistency that holds across cuts. The kind of footage you could actually use in a real project without it screaming "AI video."

Gemini Omni is catching up, and the integration into the YouTube Shorts pipeline is genuinely useful for the creator workflow. But for anyone whose bar is "would I put this in a video?" -- Seedance and Kling are still the tools to beat on raw quality. Gemini Omni vs Seedance 2.0 comparison

The pattern I'm taking away

Google's play isn't to win on raw model performance -- it's to win on surface area. Gemini is now inside Search, Gmail, Docs, YouTube, Android, the Play Store pipeline, and your phone's notification bar. You may not choose Gemini because it's the smartest model. You'll end up using it because it's already there.

That's a different kind of moat. And for someone building tools and content on an Android phone, deeply inside the Google ecosystem -- it matters more than benchmark rankings.

If this is your kind of rabbit hole

If you got this far, you're probably the right kind of person to follow the rest of what I'm doing.

YouTube (@nomaditsu) is the main thing. Long-form essays and tutorials on AI tools, creator workflows, and the systems I'm actually building and shipping.

Instagram (@nomaditsu) is the daily layer. Experiments in progress, behind-the-scenes from Bangkok, and the messy middle of building in public.

Different cadence, different texture. Pick whichever fits how you like to consume.

Journal

View all
1 3
View all