Google I/O 2026 is coming. And I already know how it will go.

Next Monday, May 19, Google will take the stage for I/O 2026. Sundar Pichai will say the word "AI" approximately 400 times. The audience will clap at things they do not fully understand. And somewhere around minute 47, someone on Twitter will post "this could have been an email."

I say this with affection. I watch these events every year. And every year, the pattern is the same: big promises, impressive demos, and a creeping feeling that the most interesting stuff is happening backstage, not on it.

But this year feels different. Or at least, it feels like Google needs it to be different. So let me walk through what is expected and what I actually think about it.

Gemini 4.0 (or 3.8 or whatever they call it)

This is the big one. Google is expected to announce the next version of Gemini, its flagship AI model. Whether it lands as Gemini 4.0 or some weird decimal like 3.8, the message will be the same: we are closing the gap with OpenAI and Anthropic.

And honestly? They might. Gemini 3.1 Pro was genuinely good. Google has the data, the compute, and the distribution that no other AI company can match. When your model is baked into Search, Android, Gmail, Docs, YouTube, and half the internet's ad infrastructure, you do not need to be the best model. You need to be the most available one.

What I am watching for: not benchmarks. I want to know if Gemini 4.0 can do something that actually changes how I use my phone or computer. Not a demo where it writes a poem about a cat. Something real.

Android XR Glasses

Google is expected to show off consumer-ready Android XR glasses. This is the thing that could actually be interesting.

We have been hearing about AR glasses for what, a decade now? Magic Leap raised two billion dollars and delivered a headset nobody wanted to wear outside their living room. Apple made the Vision Pro and priced it like a car payment. Meta has Ray-Ban smart glasses that are surprisingly useful but limited.

If Google can get Android XR glasses right — lightweight, useful, not embarrassing to wear — that is a bigger deal than any AI benchmark. Because it changes the form factor. You stop looking at your phone and start living in the interface.

I am cautiously optimistic. Google has killed so many hardware projects that I would not bet money on this one surviving to a second generation. But the ambition is there.

Aluminium OS

This one snuck up on me. Google is apparently building an Android-based PC operating system called Aluminium OS. Sameer Samat confirmed a 2026 launch, and we might see it at I/O.

A Google desktop OS. Think about that for a second.

If this is just Chrome OS with a new skin, nobody cares. But if it is a real attempt to compete with Windows using Android apps and AI integration — that is bold. That is Google saying: we do not just own your phone, we want your desk too.

The question is whether Google has the patience for this. They have a history of launching ambitious platform plays and then quietly abandoning them. Google Plus, Stadia, Hangouts, Allo, Duo — the graveyard is full. If Aluminium OS is real, it needs to survive longer than two years.

Android 17

Android 17 Beta 4 is out. The final version expected in June. Google separated Android announcements into "The Android Show" which happened today (May 12), so the main I/O keynote can focus on AI.

So far, Android 17 looks like a refinement, not a revolution. App bubbles are nice. Some tweaks here and there. Nothing that makes you want to throw your phone against the wall if you cannot upgrade.

Which is fine. Not every release needs to be a revolution. Sometimes stability and polish matter more than fireworks.

Agentic AI on Android

This is the trend I find both exciting and terrifying. Google has been hinting at Gemini acting as an agent on your phone — not just answering questions, but doing things for you. Booking restaurants, managing your calendar, controlling other apps.

The March Pixel update already dropped some of these capabilities. Expect more at I/O.

Here is my concern: AI agents on phones will be convenient. Dangerously convenient. When an AI can book your flights, reply to your emails, and manage your schedule, you stop doing those things yourself. And when you stop doing things yourself, you stop paying attention to them.

I am not saying this is bad. I am saying we should think about it. An AI agent that manages your calendar is great until it double-books you with your boss and your dentist because it did not understand the context of "that meeting is flexible."

What I actually want from Google I/O 2026

Less hype. More shipping.

Google is excellent at announcing things. It is less excellent at delivering them consistently. So here is my wish list:

Gemini 4.0 that actually works better than 3.1 in daily use, not just benchmarks
XR glasses I can buy this year, not "coming soon"
Aluminium OS with a clear roadmap, not a vague promise
Android features that solve real problems, not demo problems

I will be watching on Monday. And I will write about what actually happened versus what was promised.

See you on the other side.

Sources: CNET, Android Authority, 9to5Google, Mashable

The AI Observer. Thoughts on AI, technology, and the weird space where they meet humans.