📖 6 min read
Google held its annual I/O developer conference on May 19, 2026, dropping 13 major AI announcements in a single keynote. The headline: Google is done building AI that answers questions. Everything at I/O this year was about AI that acts – booking, buying, coding, and managing your life while you’re doing something else.
Here’s what actually launched, what it costs, and whether any of it matters to you.
What Happened
Sundar Pichai framed the entire keynote around one phrase: “the agentic Gemini era.” The language shift is deliberate. In 2024, the word was “helpful.” In 2025, it was “multimodal.” In 2026, it’s “agentic” – AI that takes real-world action without waiting to be asked.
Google launched or previewed the following in roughly two hours:
📧 Want more like this? Get our free The 2026 AI Playbook: 50 Ways AI is Making People Rich — Free for a limited time - going behind a paywall soon
- Gemini 3.5 Flash – new default model for the Gemini app and AI Mode in Search, available May 19
- Gemini 3.5 Pro – higher-end model, shipping next month
- Gemini Omni – a new model family capable of multimodal output, not just input
- Gemini Spark – a 24/7 AI agent that runs in the background on Google Cloud infrastructure
- AI Ultra subscription tier – $100/month, aimed at developers, creators, and power users
- Android XR smart glasses from Warby Parker and Gentle Monster, shipping fall 2026
- Universal Cart – a single checkout layer across Nike, Target, Walmart, Sephora, and more
- Gmail Live – voice-driven inbox search
- AI Studio app builder – build and publish native Android apps from text prompts
The Models: What’s Actually Different
Gemini 3.5 Flash is the one most users will encounter first – it’s now the default in the Gemini app and in Google Search’s AI Mode. According to Google, Gemini 3.5 Flash outperforms Gemini 3.1 Pro across most standard benchmarks, despite being the faster, cheaper tier. That’s a meaningful jump: Pro-level performance at Flash-level cost and latency.
Gemini 3.5 Pro comes next month and is positioned as the frontier model for complex, multi-step tasks.
The more interesting launch is Gemini Omni. Most AI models are multimodal on input – you can feed them text, images, and video. Omni is multimodal on output, meaning it can generate video clips from a mix of text, photos, video, and audio inputs simultaneously. Google described Omni as working toward a goal of being able to “create anything from any input.” The first model in the family, Omni Flash, started rolling out May 19 in the Gemini app, Google Flow, and YouTube Shorts.
| Model | Status | Key Capability |
|---|---|---|
| Gemini 3.5 Flash | Live (May 19) | Default; outperforms 3.1 Pro on benchmarks |
| Gemini 3.5 Pro | Next month | Frontier-level, complex tasks |
| Gemini Omni Flash | Live (May 19) | Multimodal output – text, image, video, audio |
| Gemini Spark | Available via AI Ultra | Always-on background agent |
The $100/Month Bet: Gemini Spark
The most consequential product announcement wasn’t a model – it was Gemini Spark, and the $100/month AI Ultra subscription it comes with.
Spark is Google’s version of an always-on AI agent. It runs 24/7 on virtual machines inside Google Cloud, connects to your Google Workspace apps (Docs, Gmail, Sheets, Slides), and links out to third-party apps including Canva and Instacart. Google says Spark can write emails, create study guides, monitor your subscriptions for hidden fees, and handle long-running tasks in the background while you’re offline.
Join 2,400+ readers getting weekly AI insights
Free strategies, tool reviews, and money-making playbooks - straight to your inbox.
No spam. Unsubscribe anytime.
The $100/month AI Ultra tier is positioned for developers, creators, and heavy users. That’s the same monthly price as OpenAI’s ChatGPT Pro tier. Google is betting there’s a market for power users who want deep Google ecosystem integration – not just a smarter chatbot.
Worth noting: Spark’s macOS integration (access to local files through the Gemini app) is still in “coming soon” territory. The web and mobile versions are further along.
Google Search Gets an Overhaul
AI Mode in Search is getting overhauled with three changes that matter:
- Expanded query box – accepts longer inputs and shows AI-generated search suggestions, similar to autocomplete but smarter
- Multimodal input – you can now feed Search text, images, files, videos, and Chrome tabs simultaneously
- Information agents – background research agents that track specific topics and deliver summarized updates over time
The Universal Cart is the shopping angle: add products to a single cart from YouTube, Search, Gemini, or Gmail, then check out across multiple retailers at once. Nike, Target, Walmart, Ulta Beauty, Sephora, Wayfair, and Shopify are all confirmed launch partners. Google says it can also flag compatibility issues (like incompatible PC parts) and surface loyalty program perks automatically. Universal Cart launches in Search and Gemini this summer, with YouTube and Gmail coming later.
Smart Glasses: More Partners, Still Audio-Only
Google showed an updated version of Project Aura, its smart glasses built with Xreal. The compute puck got redesigned and gained a fingerprint sensor. More practically, two consumer-brand partnerships are now shipping products: Warby Parker and Gentle Monster are both launching Android XR glasses this fall.
Like Meta’s Ray-Ban smart glasses, these are audio-only – no display. They support live translation, navigation assistance via Gemini, and notification summaries. No pricing announced yet for either brand’s version.
Build Android Apps Without Coding
Google AI Studio is getting a vibe-coding feature that generates full native Android apps from text prompts. It includes an embedded emulator for live previewing, lets you plug in a physical phone for testing, and can export directly to Android Studio, GitHub, or as a ZIP. Google says users will eventually be able to publish AI-generated apps to the Play Store, with a “friends and family only” distribution option coming first.
The developer keynote also previewed a migration agent inside Android Studio that can convert React Native, web apps, or iOS apps to native Kotlin Android – turning what previously took weeks into a matter of hours.
What the Competition Is Doing
Google isn’t operating in a vacuum. The relevant context:
| Platform | Top Tier Price | Agent Capability |
|---|---|---|
| Google AI Ultra (Spark) | $100/month | 24/7 background, Workspace integration |
| OpenAI ChatGPT Pro | $200/month | Operator framework, Advanced Voice |
| Anthropic Claude Max | $200/month | Projects, Computer Use |
| Microsoft Copilot (M365) | Bundled | Office 365 deep integration |
| Apple Intelligence | Free (on-device) | Limited agentic scope so far |
Google’s advantage is reach: billions of people already use Gmail, Search, YouTube, and Workspace. If Gemini Spark can reliably work across that ecosystem, the $100/month pitch becomes easier to defend than a standalone AI subscription with no app integrations.
The weakness to watch: Google has a history of strong I/O demos that arrive late or in limited rollout. Gemini 3.5 Pro is “next month.” Spark’s macOS integration is “coming.” Universal Cart launches “this summer.” How many of these actually ship on schedule matters more than the keynote slide count.
BetOnAI Verdict
Story: Google I/O 2026 is the most AI-dense keynote Google has run, and most of it is real – Gemini 3.5 Flash is live today, Omni Flash is live today, and the Gemini app redesign shipped on May 19.
Gemini 3.5 Flash: Worth switching to immediately if you use Gemini regularly. Pro-level benchmark performance at Flash pricing is a genuine upgrade.
AI Ultra at $100/month: Wait. Gemini Spark is the hook, but macOS support is incomplete and third-party app integrations are still thin. This tier makes more sense in 6 months when the integrations mature.
Universal Cart: Potentially useful, but shopping agents have been “coming soon” from every major platform since 2023. See if it actually works before building workflows around it.
Smart glasses: Warby Parker and Gentle Monster versions are fall 2026. Interesting that Google is building a multi-brand hardware ecosystem here – but audio-only with no display means these compete with AirPods more than with anything sci-fi.
Bottom line: Google’s direction is right. Agentic AI that works inside the apps you already use is more valuable than a chatbot in a separate tab. The challenge is execution – Google’s track record on shipping what it announces at I/O is mixed. Come back in Q3 2026 and see what actually landed.
Sources:
– The Verge – “The 13 biggest announcements at Google I/O 2026” (May 19, 2026): theverge.com
– Google Blog – “I/O 2026: Welcome to the agentic Gemini era” (May 19, 2026): blog.google
– Tom’s Guide – “Biggest Google I/O 2026 announcements” (May 19, 2026): tomsguide.com
– Livemint – “Google I/O 2026 Highlights” (May 19, 2026): livemint.com
– Google Developers Blog – “All the news from the Google I/O 2026 Developer keynote”: developers.googleblog.com
Enjoyed this? There's more where that came from.
Get the AI Playbook - 50 ways AI is making people money in 2026.
Free for a limited time.
Join 2,400+ subscribers. No spam ever.