Siri Finally Woke Up: Apple's WWDC 2026 AI Glow-Up (and the Twist Nobody Saw Coming)

June 10, 2026 Ray de Guzman

Siri spent a decade as AI's punching bag. WWDC 2026 was the comeback attempt.

For most of a decade, Siri was the assistant you compared other assistants to so you could feel good about them. While ChatGPT and Gemini were planning trips and writing code, Siri was setting timers and offering to Google things for you.

So at WWDC 2026 -- Tim Cook's final keynote as CEO -- Apple stood up and announced a brand new "Siri AI." It has promised this glow-up twice before and missed. This was the third "no, really, this time."

Here's the twist: it might actually be true this time. And the reason why is the best plot twist in tech this year.

The 30-second version:

Siri finally talks like a real assistant, sees your screen, and does things across your apps.
The smart part is secretly powered by Google's Gemini -- yes, the same AI running Android.
A pile of features are genuinely useful and also shamelessly borrowed from rivals.
Apple's real edge isn't being the smartest. It's being the most private. More on that below.

Siri finally got a real brain

The upgrade is real, and it comes down to three things old Siri couldn't do:

It holds a conversation. It remembers what you just said, goes back and forth, and gives you actual answers instead of a wall of links.
It sees your screen and uses your apps. The demo that landed: someone pointed Siri at an Instagram post of a landmark and just asked for directions. No copying an address, no app-switching. Siri read the screen, got it, and handed back the route.
There's a standalone Siri app where you can scroll back through past chats. Which is, you know, the ChatGPT app. Apple looked at the whole chatbot world and said "we'll have what they're having."

Why it matters: this is the jump from "assistant that answers" to "assistant that does." That's the line every AI company is racing to cross right now.

The plot twist: it's running on Google's Gemini

Apple buried this, so let me say it plainly: the smartest part of the new "Apple Intelligence" is Google's Gemini. The same Gemini I wrote about when it took over Android.

Here's how it actually works, without the acronym soup. Apple's assistant runs on three brains:

A small, fast brain on your phone for quick, private stuff.
A private Apple cloud for heavier requests, built so it provably isn't logging or snooping on you.
And for the genuinely hard thinking, a custom version of Google's Gemini -- but routed through that private Apple bubble, so Google never actually sees your data.

Sit with the irony. Apple spent years branding "Apple Intelligence" like a secret in-house superpower, and the big reveal is that the superpower is its biggest rival's model.

But here's why it's actually a smart move: building a top-tier AI model costs billions and years, and Apple started late and cautious. So instead of losing that race, it rented the best brain available, ran it on its own hardware, and made the whole pitch about privacy instead of raw intelligence. Anyone can rent Gemini. Not everyone can promise -- and prove -- they're not peeking at your data. That's the bet.

What it can actually do (and is it really an "agent"?)

Beyond chatting, Siri now reaches into your life a bit:

Messages suggests replies.
Phone can pull up context from your Mail and Messages mid-call.
Safari finally wrangles your 40 open tabs.
Passwords update in a tap.

Quick reality check, because "agentic AI" is the buzzword of the year: most of this is Siri being a very smart helper that does one step for you -- read this, then do that. It's not yet an AI that runs your whole multi-step errand end to end, books the thing, fills the cart, and handles the curveballs. Useful, real, but keep the hype calibrated. (Hold that thought -- it matters for the Android comparison.)

The cool stuff that's also gloriously borrowed

Photos can bend reality. A new tool called Reframe rebuilds your photo in fake-3D so you can reposition the camera after you took the shot -- slide that pole out from behind your friend's head. There's also Extend (the AI invents more of the photo) and a smarter Cleanup (erase photobombers). Genuinely cool. Also something Google Pixel and Samsung shipped two years ago. Welcome, Apple, cake's in the back.
Shortcuts you build by talking. Apple's automation app was always powerful and completely impenetrable. Now you just describe what you want ("every morning text my partner the weather and my first meeting") and the AI builds it. This is the quiet winner that opens automation to normal humans.
Dictation that cleans up your rambling. Talk, and it fixes your punctuation, caps, and "ums" live. Love it. Also, this is Apple casually flattening the entire business model of apps like Wispr Flow and Willow. Rough day to be that startup.
Image Playground, take two. Last year's AI image generator flopped so hard Apple seemed embarrassed. This is the redo and the renewed plea to please generate something this time. Respect the persistence.
Liquid Glass gets a rollback. Last year's polarizing design can now be dialed back down. An entire WWDC segment dedicated to undoing the previous WWDC. Iterative, baby.

The for-builders bit, in one breath

Apple is letting any app tap that on-the-phone AI brain for free, offline, without sending your data anywhere. Why it matters: a wave of apps can get smarter and more private without paying for AI by the token. The catch: it's the small brain, so anything heavy still routes through Apple's cloud on Apple's terms.

The fine print Apple glossed over

English only at launch. Other languages wait.
It may skip the EU and China for a while thanks to regulation -- two of the biggest markets on Earth.
Privacy is the real headline. Apple's exec Craig Federighi hammered "privacy in AI is non-negotiable," and the verifiable-privacy angle is the one thing here rivals can't easily copy.
The good stuff ships this fall with iOS 27, which runs all the way back to the iPhone 11.

How this stacks up against Android's Gemini era

A few weeks ago I wrote Gemini Intelligence: Android's Agentic Era Starts This Summer about Google baking an AI agent straight into Android. And now here's Apple doing its own version on the same stage.

The delicious irony: I ended that piece wondering who'd build on top of Google's Gemini. Turns out one answer is Apple. Google quietly won the round by powering both sides. Buy an iPhone or a Pixel this fall and you're talking to Gemini either way -- it's just wearing a different outfit.

But the two play it very differently, and that's the interesting part now that they share a brain:

Android let the agent run. Multi-step errands, browsing, doing the whole task end to end. Bigger ambitions, looser leash.
Apple put the agent on a leash. It sees your screen and helps across apps, but it's cautious, permission-first, and wrapped in privacy promises. Smaller ambitions, way smaller chance of going rogue.

In the Android piece I floated that the future is a general-purpose AI baked into your phone doing the boring errands, plus a personal, opinionated agent doing the high-context creative work. WWDC just confirmed half of it: the phone-level assistant is now baseline on both platforms, and both run on Gemini. The competition isn't "who has the smartest AI" anymore -- it's who has the better approach and who you trust with your data.

The personal agent I built for my own work still does the thing neither Siri nor Gemini will: the high-context, opinionated stuff a giant one-size-fits-all assistant can't. WWDC didn't change my setup. It just made the baseline a lot better for everyone else.

So which approach are you betting on for the next year -- the agent that runs free, or the one that proves it's playing it safe?

If you like being early on this stuff

If you got this far, you're exactly the kind of person I make this for.

YouTube (@nomaditsu) is the main thing -- breakdowns of new AI tools, what's actually worth your time, and the systems I build to make this stuff useful instead of just hype.

Instagram (@nomaditsu) is the daily layer -- quick experiments, behind-the-scenes, and the messy middle that doesn't make it into an article.

Different speed, different vibe. Pick whichever fits how you like to keep up.

Item added to your cart