MOUNTAIN VIEW, CA – Right then. Settle down. I know it’s been a long year. The pound is still made of vibes and old tea leaves, the weather has been aggressively mediocre, and you’ve just about recovered from the trauma of last year’s Gemini update that accidentally renamed your dog ‘Gemini Advanced Platinum Plus.’
But fear not. It is May 19, 2026, which means it is time for the annual ritual where Google rents an enormous amphitheatre in California, stands in front of a lot of sweaty developers who desperately need some sun, and tells us that this time, everything is going to be different. This time, the robots are actually nice.
We have just witnessed the Google I/O 2026 keynote. And I have to say, if you thought the industrial revolution was a bit much, you are going to absolutely hate the next twenty minutes.
The ‘Uh-huh’ Heard Round the World
The lights dimmed at the Shoreline Amphitheatre. The smell of artisanal avocado toast and burning venture capital filled the air. Sundar Pichai took the stage looking like a man who has seen the future and is deeply, deeply tired of explaining to his parents what he actually does for a living.
He smiled. That slow, patient smile of a man who knows he is about to ask you to trust him with your grandmother’s medical records and your search history from 2008.
“Good morning, everyone,” Sundar began, his voice soothing, like a meditation app that has been hacked by a spreadsheet. “The last ten years have been incredible.”
Which is corporate code for: We started this as a search engine, then we accidentally built a surveillance state, but don’t worry, we’ve rebranded the surveillance as ‘helpfulness.’
He then gave us the stat that every CEO is legally required to give at a tech conference: “We are now processing over 3.2 quadrillion tokens per month.”
Three point two quadrillion. Now, I don’t know what a ‘token’ is. I think it might be a fragment of a word, or perhaps a piece of your soul. But whatever it is, Google has 3.2 quadrillion of them. To put that in British terms, that is roughly the number of cups of tea the British workforce consumes between 9:00 AM and 9:05 AM on a Monday morning.
Sundar told us we are entering the Agentic Era.
I hate that word. ‘Agentic.’ It sounds like a medical condition. “I’m sorry, Mr. Higgins, your software has become Agentic. We’ll have to put it on life support and charge you 40 quid a month for the privilege.”
Gemini 3.5 Flash: Faster Than a Speeding Letdown
We knew the AI was coming. You can’t have a Google I/O anymore without them shoving AI into every orifice of the operating system. But this year, they introduced Gemini 3.5 Flash.
According to Google, 3.5 Flash is “blazing fast” and offers “frontier-level intelligence.” Apparently, it now outperforms the previous model on the Terminal-Bench 2.1 benchmark. I have no idea what that is, but it sounds like the exam you have to pass to work at a train station.
The demo was typical Silicon Valley nonsense. A man on stage said, “Gemini, book me a holiday to Greece, find me a dentist that speaks Welsh, and write a screenplay about a sentient toaster.”
And the phone just did it. Instantly.
Now, I don’t know about you, but my current phone (which is supposedly ‘smart’) struggles to load the Tesco Clubcard app without crashing. Meanwhile, Google’s phone is out here booking Aegean cruises for a toaster that doesn’t exist yet.
The audience clapped. They always clap. They clap when the Wi-Fi connects. They clap when the slides change. These are people who have replaced their dopamine receptors with code commits.
Android 17: Or, ‘We Ran Out of Desserts’
We then moved to the operating system update. Android 17.
Remember when Android updates were named after delicious sweets? Cupcake, Donut, Eclair. We are now at Android 17. That is not a dessert. That is the age you are when you realise you have no A-Level revision done and your mum is going to kill you.
The headline feature of Android 17 is Gemini Intelligence. It allows the phone to “take action” on your behalf.
They demonstrated a feature called Rambler. You speak in absolute gibberish—slurred, tired, British gibberish—and the phone turns it into perfectly formatted text.
“Uh, yeah, sorry, running late, the M25 is, well, you know, it’s the M25, so, yeah, tell Dave I’ll be there in twenty, maybe thirty, unless I stop for a pasty, which I probably will.”
The phone types out: “I am delayed on the M25. I will arrive in approximately 30 minutes. Kindly inform Dave.”
It’s brilliant. It’s terrifying. It means we no longer have to be polite. We can just grunt at our wrist and the AI will apologise for us.
They also announced Android XR (Extended Reality). Finally, glasses that look normal. Warby Parker and Gentle Monster are making them. They don’t have screens; they are just audio glasses.
Finally! A pair of glasses that are as useless as my current prescription pair but cost six hundred pounds more! I can wear these glasses, and an AI will whisper in my ear. I’ve got that already. It’s called ‘my wife reminding me to take the bins out.’ I don’t need a TPU for that.
The Googlebook: Is It a Laptop? Is It a Chromebook? No, It’s Confused.
There was a lot of fuss about the Googlebook. Apparently, ChromeOS is dead. Long live Aluminum OS (or Aluminium, as we say in the civilised world, with the second ‘i’).
This is the merger of Android and ChromeOS. It’s a laptop that runs phone apps and looks like a laptop but acts like a tablet.
Look, Silicon Valley. We have had laptops for thirty years. They work fine. They have keyboards. They have hinges that sometimes break. We do not need a ‘magic pointer’ or a ‘dynamic widget.’ We need a battery that lasts longer than a game of Solitaire and a keyboard that doesn’t feel like typing on a deep-fried marshmallow.
But Google insists we need AI laptops. So, we’re getting AI laptops.
The ‘Oh, God, It’s Watching Me’ Segment (Gemini Spark)
Just when I thought I was safe, Sundar introduced Gemini Spark.
Gemini Spark is a 24/7 agent. It lives in the cloud. It never sleeps. It works while your phone is off.
It reads your emails. It checks your calendar. It looks at your shopping lists. And then it does things for you. It negotiates bills. It updates spreadsheets. It drafts emails to your investors.
Sundar said, with a straight face, “Spark operates autonomously, under your direction.”
No, Sundar. If it’s operating while my phone is off, and I’m asleep, that is not my direction. That is a tiny digital intern living in a server farm in Iowa, rifling through my Gmail, and deciding which of my friends is annoying.
They tried to calm us down with Android Halo. It’s a little notification that pops up when the AI is doing something.
“Gemini Spark is currently: Unsubscribing you from that newsletter you accidentally signed up for in 2015.”
“Gemini Spark is currently: Judging you for searching ‘Is cheese on toast a meal?’ at 2:00 AM.”
They also announced the Universal Cart. This is a shopping cart that lives across Search, YouTube, and Gmail. You see a jacket in a YouTube video, you say “Add to cart,” and it does it. You see a recipe in an email, “Add eggs to cart.”
It uses Universal Commerce Protocol (UCP).
Finally! A protocol for spending money I don’t have on things I didn’t know existed thirty seconds ago. The British economy is going to be absolutely humming thanks to late-night, wine-fuelled Universal Cart purchases.
A Brief Moment of Horror (The Omni Video)
Then came Gemini Omni. This is the video generator.
Google showed us a video generated by AI. It was a cat riding a skateboard in Ancient Rome. Photorealistic. The fur was perfect. The marble columns were flawless. The physics of the skateboard wheels turning on the cobblestones was mathematically impeccable.
And the entire audience of 5,000 developers sat in dead silence, because they all realised their jobs as “visual effects supervisors” and “stock footage providers” are now relegated to the same historical dustbin as the town crier and the Blockbuster video clerk.
Gemini Omni can also edit videos. You just type “Make the background a volcano,” and it does it.
Deepfakes? No, no. Google calls them “Synthetic media with imperceptible SynthID watermarks.” That’s fine, then. As long as the robot signs its name on the bottom of the fake video of the Prime Minister fighting a badger, it’s legally binding, right?
The Search Apocalypse
And finally, the bit that made my blood run cold. AI Mode in Search.
Google has redesigned the search box. It is no longer a box. It is a chat interface.
Sundar explained that instead of giving you links to websites, Google will now just… read the websites for you, summarise them, and keep you on Google.
He called it “seamless.”
I call it the end of the internet as we know it.
Why would I ever click through to the Manchester Evening News to read about a slow-speed chase on the M60 when Google can just synthesise the tone of the article using Large Language Models and whisper the gist of it to me through my Gentle Monster glasses while I’m brushing my teeth?
They launched AI Overviews last year, which famously told people to eat rocks and put glue on pizza. This year, they are doubling down. They are trusting the robot more.
There was a moment where Sundar tried to search for “best rainy day activities in London” and the AI didn’t just give results. It built an interactive mini-app. It had a live tube map, a rainfall radar, and a booking widget for the Natural History Museum.
I leaned over to the journalist next to me. “What happens if the AI gets it wrong?” I whispered.
He didn’t look away from the stage. “Then you spend three hours in the rain outside the closed London Dungeon, presumably.”
The British Verdict
So, there we have it. Google I/O 2026.
The TL;DR is: Google is putting a robot inside everything. The robot is very polite (American polite, not British polite—so it won’t apologise for existing, but it will hold the door open for you).
We have Gemini 3.5 Flash (fast), Gemini Omni (creative), and Gemini Spark (creepy).
We have Android 17 (functional), Aluminum OS (bendy), and Universal Cart (expensive).
Sundar Pichai left the stage to the sound of roaring applause and the distant sound of 7 billion people not reading this article because they’ve already asked their AI to summarise it into a single emoji.
As I packed up my laptop, I looked at my own smartphone. A perfectly serviceable device that, right now, only does what I tell it to do.
It doesn’t judge me. It doesn’t pre-empt me. And it certainly doesn’t spark.
I think I’ll keep it that way for as long as possible. At least until the M25 turns into a sentient AI agent and starts charging me a congestion fee just for looking at it.
Right. I’m off to find a pasty. The robots haven’t figured out how to eat those yet. Have they?
JOHN STAMOS