Apple's WWDC: Siri AI finally works — and Tim Cook's career bookended by the same broken promise

Apple's WWDC: Siri AI Finally Works—Tim Cook's Career Bookended by the Same Broken Promise

Tim Cook's final WWDC as CEO closes a 15-year loop that perfectly captures his tenure: Siri, the voice assistant that defined the opening keynote of his era in 2011, now actually works—just as he's leaving.

When Steve Jobs introduced Siri at the first iPhone 4S event in October 2011, the pitch was transformative. Users wouldn't need to learn syntax or memorize commands. They'd simply talk to their device. Ask "what's the weather going to be like today" or "do I need an umbrella" and it would understand context and intent. The promise was natural, conversational AI that would handle whatever you threw at it.

That announcement happened the day before Jobs died. Cook, as the newly installed CEO, introduced Siri that night. His last WWDC keynote yesterday announced the same product, finally functional.

What took 15 years was not Apple's effort but the technology itself. The industry needed LLMs, transformer architectures, and the computational scale that emerged only in the past few years. When Siri launched in 2011, it was the best voice assistant available—genuinely impressive for its time. But the fundamental gap between the promise and the capability was real. For over a decade, competitors didn't eat Apple's lunch because Google Assistant and Alexa had the same problem. Nobody was solving the hard problem.

That changed with ChatGPT and the broader LLM wave. By 2024, when Apple Intelligence launched with ChatGPT integration, Siri was undeniably behind. Now, with the new Siri rolling out in iOS 27, Apple has caught up. The company spent roughly 12 minutes detailing the expanded assistant yesterday—including better reasoning models (AFM3 Cloud Pro, fine-tuned on Gemini), deeper system integration, and a dedicated Siri app.

Cook was CEO during what amounts to an AI winter. From 2011 to 2022, the Siri team faced the same reality every year: the underlying technology wasn't moving fast enough. They weren't inventing new capabilities because the raw material—modern deep learning—didn't exist yet. Apple remained the best of a mediocre category for over a decade. The company operationalized itself brilliantly under Cook, building the App Store into a cash engine and maintaining margins. But Cook's era happened to span the one period in recent tech history when the frontier of AI was essentially flat.

The bookend is poetic: Siri as the symbol of Cook's first major announcement as CEO, and Siri as the centerpiece of his last. Not because he failed, but because the world finally caught up.

The parental control framing

Apple devoted 12 minutes of a notably short keynote to child safety and parental controls—ask-to-browse permissions, age-gated media, contact management, and communication controls. The allocation of time signals where Apple sees leverage.

The company is not a fear-based marketing organization. When climate concerns emerged, Cook didn't give speeches about rising sea levels or ecological collapse. Instead, Apple worked privately on renewable energy infrastructure and then talked only about solutions. It's doing the same thing now with phone addiction and the fertility crisis (the topic Derek Thompson has been covering as potentially linked to smartphone use and dopamine feedback loops). Apple doesn't diagnose the problem publicly; it offers solutions and lets parents decide.

There's also a direct business case: many parents are hesitant to buy smartphones for children at all. The niche market for GPS watches without screens or social media exists because parents want control and safety. By positioning the iPhone as fully lockable—by design, not just in settings—Apple converts a liability into a selling point. You can buy your kid an iPhone because you are in control.

The Gruber-Gurman timeline dispute

In March 2026, Bloomberg's Mark Gurman reported that Apple planned to let rival AI assistants integrate with Siri in iOS 27. Users with Claude, Gemini, or ChatGPT installed could route queries directly through the Siri voice button, the same way they've been able to do since ChatGPT integration launched in 2024.

As of yesterday's keynote, John Gruber publicly questioned whether Apple had actually announced this. His read: Apple either ran out of time, forgot to announce it, or scrapped the feature entirely in the last month and rebuilt from scratch. Gruber was taking shots at Gurman's reporting.

Gurman fired back with screenshots of integrations that match his original reporting—a model picker in the Siri app, the ability to configure which assistant handles queries.

The core gap between them is about user experience, not facts. Hardcore AI users hoped the Siri button would fully remap to their model of choice by default. Currently, you still have to say "hey Siri, ask ChatGPT" and then click okay on a popup—what one observer called "playing telephone." The new flow is clearer, but the button itself may not fully delegate to your chosen service without a voice command prefix.

There are also screenshots of a model picker dropdown, but it's unclear whether that setting persists across queries or resets regularly. Details will emerge as the beta rolls out over the coming weeks.

AI features and the detection problem

Apple also introduced spatial reframing, a camera feature that lets users adjust the angle of a photo after capture using AI. If you shot a photo at an awkward angle, the tool straightens it by reconstructing the geometry. This sparked mixed reactions.

Some users see it as creep—the camera is no longer just capturing reality but altering it. Apple has long positioned the camera app as "what you see is what you get," though that claim has been somewhat loosened over years of computational photography: artificial depth of field, tone mapping, and image upscaling already alter the output.

Others view spatial reframing as a natural extension of the camera's existing toolkit, no different than tone mapping or noise reduction. You can simply not use it.

The tension is real but may become moot. As one observer noted, soon you'll take a photo in low light, the phone will upres it, apply filters, color-grade it, spatially reframe it, remove unwanted people in the background, and what you get bears little resemblance to the scene you captured. The endpoint is a fundamentally synthetic photo.

Instagram and other platforms are beginning to tag AI-generated or significantly AI-modified images. Apple could easily extend that to spatial reframing and other features—adding metadata to flagged images. There was some hope Apple would lean into detection and disclosure more aggressively yesterday. It didn't.

The Google Cloud compute deal

Apple is extending its private cloud compute infrastructure onto Nvidia GPUs running on Google Cloud. The partnership allows AFM3 Cloud Pro—Apple's reasoning model, fine-tuned on Gemini—to run on Google infrastructure while maintaining Apple's privacy guarantees.

This creates a branding tension. Apple has long marketed itself as the privacy-first alternative to Google, the "listening" company. Now it's vending its most powerful AI models through Google's infrastructure.

There's also an environmental concern. Apple has made renewable energy central to its brand story—data centers powered by clean energy. But Google signed a deal with Colossus, a data center operator that does not run on renewable energy. If Apple Cloud Pro models end up running on those Colossus nodes through Google Cloud, there's a mismatch between Apple's public environmental claims and its actual footprint.

Apple has never claimed it sources only from renewable providers—just that its own data centers run clean. But if this becomes a long-term model, expect pressure for Apple to invest in its own clean-energy infrastructure rather than rent from providers that don't meet its standards.

The bigger picture

Siri in 2011 promised the future. It delivered a product that worked for a narrow set of commands in a narrow set of contexts. The gap between promise and capability was real and stayed real for 15 years because the underlying technology didn't exist.

Now the technology exists. Siri works because LLMs work. Apple got there, albeit later than some competitors, because it prioritized its walled garden and its own privacy model over speed to market.

Cook's tenure spans exactly the period when that tradeoff mattered least—when the frontier of AI was relatively flat and late entrants could catch up quickly once the breakthrough arrived. The irony is also clean: Cook leaves just as his most visible opening promise finally becomes true.