Google used its annual I/O developer conference on 19 May 2026 to signal a decisive shift in how it thinks about AI — from a tool that answers questions to one that takes action. Under the banner of an “agentic Gemini era,” the company unveiled new models, redesigned interfaces, a cross-product shopping cart, intelligent eyewear, and a sweeping set of developer tools, all tied together by a single thesis: AI should be working in the background even when you are not.
Scale Before the Announcements
CEO Sundar Pichai opened with a scale check. Monthly tokens processed across Google’s surfaces have grown from 480 trillion last year to more than 3.2 quadrillion — a sevenfold year-on-year increase. The Gemini app has more than doubled its user base in a year, reaching over 900 million monthly active users across 230 countries. AI Overviews in Search now has 2.5 billion monthly active users, while AI Mode — positioned as the biggest Search upgrade in Google’s history — has crossed 1 billion monthly active users in under 12 months.
Gemini 3.5 Flash: Fast, Cheap, Frontier
The headline model announcement is Gemini 3.5 Flash, which Google claims outperforms Gemini 3.1 Pro across almost all benchmarks while running four times faster than comparable frontier models, at less than half the price. Pichai framed the cost case bluntly: companies processing a trillion tokens a day could save more than US$1 billion annually by shifting 80 percent of their workloads to Flash. Gemini 3.5 Pro, described as showing strong improvements internally, is expected to follow next month.
Gemini Omni: Any Input, Video Output
Gemini Omni Flash is a new model that accepts any combination of text, images, audio and video as input and generates high-quality video output. Editing works conversationally — each instruction builds on the last, with consistent characters and physics across turns. Users can transform real footage, apply styles, swap backgrounds, or drop a character into a real-world environment by referencing their own photos and clips. All Omni-generated videos are watermarked with Google’s SynthID technology and are verifiable in Search and Chrome. Omni Flash is available today to Google AI Plus, Pro and Ultra subscribers globally, and is also rolling out free to YouTube Shorts users.
The Gemini App: A New Look and Two New Agents
The Gemini app is getting its most significant redesign to date with a new design language called Neural Expressive, featuring fluid animations, vibrant colours and haptic feedback, rolling out globally today. The conversational Gemini Live experience is now embedded directly in the main app, allowing users to switch between typing and open-ended voice conversation without interruption.
Two new agents debut alongside the redesign. Daily Brief is an opt-in agent that compiles a personalised morning digest from Gmail, Calendar and other connected apps, prioritising tasks and suggesting next steps. It begins rolling out to AI Plus, Pro and Ultra subscribers in the US today. Gemini Spark is more ambitious: a 24/7 cloud-based personal agent running on Gemini 3.5 and the Antigravity harness, deeply integrated with Workspace tools. Because it runs on Google Cloud rather than on-device, it continues working after the laptop is closed. Spark can set recurring tasks, learn workflows, synthesise meeting notes and create polished documents. New MCP connections to Canva, OpenTable and Instacart launch today, with more partners to follow. Spark begins rolling out to trusted testers this week, with a US Ultra subscriber beta next week.
Search Gains Agentic Capabilities
Search is adding three agentic layers this summer. Information agents run in the background 24/7 to surface relevant information at the right moment, starting with AI Pro and Ultra subscribers. Generative UI capabilities — dynamic layouts and interactive visuals built by Gemini 3.5 Flash on the fly — will be available to all users at no charge. For longer-running tasks, Search can build persistent custom dashboards, described as mini apps for specific needs, available initially to Pro and Ultra subscribers in the US.
Google Workspace: Voice, Pics and AI Inbox
Three voice features arrive in Workspace this summer for AI Pro and Ultra subscribers. Gmail Live enables voice-activated inbox search — users can ask what their flight gate number is and get an instant synthesised answer. Docs Live turns free-form verbal thinking into structured first drafts, pulling relevant context from Gmail, Drive and the web with permission. Keep gets a voice input mode that auto-organises verbal notes into structured lists in the background.
Google Pics is a new image creation and editing tool built on the Nano Banana model that treats every element in an image as an individual object, enabling precise edits — colour changes, object swaps, text translation — without regenerating the whole image. It integrates with Slides and Drive and supports real-time collaboration. Pics launches to a limited trusted tester group today and will roll out globally to AI Pro and Ultra subscribers this summer.
AI Inbox, which intelligently prioritises emails and surfaces to-dos in Gmail, was previously limited to Ultra subscribers. It is now expanding to AI Plus and Pro subscribers in the US, with new features including personalised draft replies and one-click task management.
Google Shopping: Universal Cart
Universal Cart is Google’s new cross-merchant shopping hub. Added items are tracked across Search, Gemini, YouTube and Gmail. The cart monitors price history, flags product incompatibilities — useful for purchases like PC components from multiple retailers — and surfaces payment card perks and loyalty rewards. Checkout via Google Pay is available now with Nike, Target, Walmart, Sephora, Wayfair and Shopify merchants including Fenty and Steve Madden. A new Agent Payments Protocol (AP2) enables agents like Gemini Spark to make purchases within user-defined guardrails. Universal Cart rolls out in the US this summer.
Developer Tools: Antigravity 2.0 and Managed Agents
Google Antigravity, its agent-first development platform, is expanding into a full ecosystem. Antigravity 2.0 is a new standalone desktop application for orchestrating multiple agents in parallel, with scheduled tasks and integrations across AI Studio, Android and Firebase. A new Antigravity CLI serves terminal-first developers; Google is encouraging Gemini CLI users to migrate. Managed Agents in the Gemini API lets developers spin up an agent with a single API call — it reasons, uses tools and executes code in an isolated Linux environment, powered by the same infrastructure behind Google’s own products.
Google AI Studio is getting a mobile app, native Android app-building support, and a one-click export to Antigravity for local development. A new $100/month AI Ultra tier targets developers specifically, offering 5x the usage limits of the Pro plan alongside priority Antigravity access and 20TB of storage. The existing $250 Ultra plan drops to $200 with identical capabilities.
Creative Tools: Flow, Flow Music, Pomelli
Google Flow, the AI filmmaking studio, gains Gemini Omni Flash for conversational video editing, a new Flow Agent that can brainstorm, batch-edit and organise assets, and a no-code Tools feature for building and sharing bespoke workflows. Google Flow Music adds section-by-section editing, full-track style transformation and Omni-powered music video creation. Both products get mobile apps — Flow on Android (beta) and Flow Music on iOS. Pomelli, the Google Labs branding tool for small businesses, introduces a Pomelli Agent that generates brand identities, brand books and full websites from a single setup conversation.
Hardware: Intelligent Eyewear This Fall
Google confirmed two categories of intelligent eyewear built on the Android XR platform with Samsung. Audio glasses deliver spoken assistance in the ear and arrive first, this fall, in two ranges: designs by Gentle Monster and designs by Warby Parker. Features include turn-by-turn navigation, hands-free calls and texts, real-time translation of speech and written text, photo capture with on-device editing, and multi-step task delegation. Display glasses, which project information onto the lens, are coming later. Both types are compatible with Android and iOS phones.
Science and Infrastructure
Gemini for Science is a new collection of experimental tools on Google Labs aimed at accelerating research workflows: Hypothesis Generation (multi-agent idea evaluation built with Co-Scientist), Computational Discovery (parallel hypothesis testing built with AlphaEvolve) and Literature Insights (paper synthesis built with NotebookLM). Validation papers for two of the underlying tools were published in Nature on the same day as I/O.
On infrastructure, Google’s eighth-generation TPUs — the 8t for training and 8i for inference — now power the full stack. The 8t enables distributed training across more than 1 million TPUs globally. Pichai also confirmed an expected capital expenditure of US$180 to $190 billion this year, approximately six times 2022 spending levels.
Project Genie, DeepMind’s world-generation model, now connects to Google Street View imagery, allowing users to create interactive environments anchored in real-world locations. It is rolling out to AI Ultra $200 subscribers globally. Android Halo, a new status layer in Android, will show users at a glance what Gemini Spark or other agents are doing in the background, arriving later this year.
Quick Reference: What Was Announced and What It Means for You
[Model] Gemini 3.5 Flash
What it is: Google’s new flagship AI model — faster and cheaper than its predecessor, yet more capable across almost every benchmark.
What it means for you: If you use any Google AI product, your responses will be snappier and more accurate. For businesses and developers, it dramatically reduces the cost of running AI at scale.
[Model] Gemini Omni Flash
What it is: A new model that takes any mix of text, images, audio and video as input and generates high-quality video. You edit it conversationally — no timeline, no manual cuts.
What it means for you: Content creators and marketers can produce polished video from their own footage using plain-language instructions. Think of it as a video editor you talk to.
[Gemini App] Neural Expressive (New Design)
What it is: A full redesign of the Gemini app with fluid animations, richer colours and haptic feedback. Responses now adapt their visual format — using images, timelines and graphics — rather than returning walls of text.
What it means for you: The app becomes noticeably more polished and easier to scan. Responses are designed to communicate, not just answer.
[Gemini App] Daily Brief
What it is: An opt-in morning agent that pulls from your Gmail, Calendar and tasks to create a personalised daily digest — with prioritised to-dos and suggested next steps.
What it means for you: Instead of opening five apps to figure out what your day looks like, you get a single, organised briefing. It is the closest Google has come to a proactive personal assistant for everyday users.
[Gemini App] Gemini Spark
What it is: A 24/7 cloud-based AI agent that works on your behalf across Gmail, Docs, Slides and third-party apps via MCP connections. It keeps running even after you close your laptop.
What it means for you: You can hand off a recurring task — say, parsing your credit card statement every month for hidden subscriptions — and Spark handles it in the background. It asks before taking high-stakes actions like sending emails or spending money.
[Search] Agentic Search
What it is: Three new layers added to Google Search: background information agents that surface relevant content at the right moment, generative dynamic layouts built on the fly, and persistent custom dashboards for ongoing research tasks.
What it means for you: Search starts to feel less like a lookup tool and more like a research partner that tracks topics over time and builds resources specifically for your questions.
[Shopping] Universal Cart
What it is: A single shopping cart that works across Search, Gemini, YouTube and Gmail. It tracks prices, flags product incompatibilities, surfaces card perks and enables Google Pay checkout across multiple retailers.
What it means for you: No more jumping between retailer sites to compare and checkout. Your cart follows you across Google, and an agent can eventually shop on your behalf within limits you set.
[Workspace] Gmail Live, Docs Live & Keep Voice
What it is: Voice-first features across three Workspace apps. Gmail Live answers spoken inbox queries. Docs Live turns verbal thinking into structured first drafts, pulling context from your Drive and Gmail. Keep organises voice brain-dumps into tidy notes automatically.
What it means for you: Thinking out loud becomes a legitimate way to write. Commuters, people with accessibility needs, and anyone who types slowly stand to gain the most.
[Workspace] Google Pics
What it is: A new AI image creation and editing tool that treats every element of an image as an individual, selectable object. You can move, recolour or swap specific things without redoing the whole image. Integrates with Slides and Drive.
What it means for you: Marketers and communicators get precise image editing without needing Photoshop skills. It is especially useful for social media assets, event flyers and presentation visuals.
[Workspace] AI Inbox
What it is: An intelligent Gmail layer that prioritises urgent emails, generates draft replies, surfaces relevant Docs or Sheets linked to tasks, and lets you mark items done in a single click. Now available to AI Plus and Pro subscribers.
What it means for you: If your inbox is a source of anxiety, AI Inbox does the triage. It will not replace reading your emails, but it significantly reduces the time spent deciding what to act on first.
[Creative] Google Flow & Flow Music
What it is: Google Flow is an AI filmmaking studio that now includes Gemini Omni for conversational video editing, a creative agent for brainstorming and batch editing, and a no-code tools builder. Flow Music adds granular song editing, style transformation and AI music video creation. Both get mobile apps.
What it means for you: Independent filmmakers, YouTubers and musicians gain production tools that were previously only accessible with expensive software and specialist skills.
[Creative] Pomelli
What it is: A Google Labs branding tool for small businesses that now includes an AI agent to define your brand identity from scratch, then automatically generate a brand book and launch a full website.
What it means for you: Small business owners and solo entrepreneurs can build a coherent brand presence without hiring a designer or agency. It is one of the more practical SME-focused tools in Google’s Labs portfolio.
[Hardware] Intelligent Eyewear
What it is: AI-powered glasses in partnership with Gentle Monster and Warby Parker, launching this fall. Audio glasses deliver spoken Gemini assistance for navigation, calls, translation, photo capture and hands-free task delegation. Display glasses, which show information on the lens, follow later.
What it means for you: For frequent travellers or people who want help without looking at their phone, these are the most practical wearable AI devices Google has announced. The fashion-brand partnerships suggest Google is serious about making them something people actually want to wear.
[Dev] Google Antigravity 2.0
What it is: An expanded agent-first development platform with a new desktop app, CLI, SDK and enterprise integrations. Managed Agents in the Gemini API lets developers deploy a capable agent with a single API call.
What it means for you: If you build on Gemini, this is Google’s clearest signal yet about where its developer ecosystem is heading — away from single-turn prompts and towards persistent, orchestrated agents.
[Science] Gemini for Science
What it is: A suite of experimental research tools on Google Labs: Hypothesis Generation (multi-agent idea evaluation), Computational Discovery (parallel hypothesis testing) and Literature Insights (paper synthesis via NotebookLM). Enterprise versions are available on Google Cloud.
What it means for you: Primarily relevant to researchers and R&D teams, but the broader significance is that Google is positioning Gemini as a genuine scientific instrument — not just a productivity tool.
[Science] Project Genie + Street View
What it is: Google DeepMind’s world-generation model now uses Street View imagery to anchor its interactive environments in real-world locations. Users can reimagine any US location in a chosen visual style — underwater, stone age, black-and-white film.
What it means for you: A curiosity for most, but an early signal of where Google’s world modelling technology is heading — with direct applications in robotics, autonomous vehicles and eventually AR/VR experiences.
Share your thoughts