Apple’s annual WWDC saw its latest edition begin on 8th June with a keynote highlighting all of the company’s newest innovations in software. Following the frankly underwhelming performance since the dawn of the AI age of all Apple’s operating systems, most people (myself included) were hoping to see Apple’s software finally catch up to their hardware. This event also marks the final year Tim Cook will be part of Apple, with John Ternus taking over his role as CEO barring any unexpected issues.
Now then, despite what your personal opinions are, AI has become the primary focus of every major tech company. As such, they skipped through much of the new features and refinements of all the new OS’ to focus on Siri AI – essentially delivering what was promised to users back in 2024.
Apple Silicon deserved better
I’m not Apple, so I won’t blaze through the optimisations they allegedly made with these new versions of iOS and macOS. Because, honestly, these might be the most important changes announced at the entire event.
Apple claims apps now launch up to 30 percent faster, newly taken photos load up to 70 percent faster and AirDrop transfers can be up to 80 percent faster. Those are obviously Apple’s own figures, produced under Apple’s own testing conditions, but the improvements do not feel entirely imaginary. Having run the developer betas since they were released, these are easily the most stable first developer betas I have used in years. That is not exactly a high bar, but I have experienced fewer catastrophic bugs, fewer random battery drains and considerably less of the general sense that my phone is being held together with tape.
This is the kind of work Apple’s software has needed for years, arguably since the beginning of the Apple silicon era. The company has spent the last half-decade creating some of the most powerful and efficient consumer chips in existence, then paired them with operating systems that often felt strangely incapable of taking full advantage of them. A MacBook could export a complicated video project without breaking a sweat, yet Spotlight might refuse to find a document whose exact name you had just typed in.
The Spotlight’s on Search
That particular problem appears to have finally received some attention. Search across Spotlight, Photos and Mail has been rebuilt to be more stable and efficient, with a new semantic index that does not just look for exact words but attempts to understand what the indexed content actually means. Developers can also make their apps’ information available to this index, allowing both Spotlight and Siri to retrieve it more naturally.
The obvious reason for building this is Siri AI. If Siri is supposed to find the restaurant your friend mentioned three weeks ago, identify the photograph containing your passport number or retrieve a playlist from a third-party app, it needs a reliable map of everything stored across your device. The old search system was nowhere near dependable enough for that.
The rather nice side effect is that ordinary search should become better too. Even users who never ask Siri to read their emails or perform actions across apps should benefit from a faster, more accurate Spotlight. The same applies to the less glamorous work Apple has done around background processing, memory management and the way tasks are scheduled across its chips. AI may have provided the motivation, but the infrastructure required to make AI work benefits the entire operating system.
In other words, Apple’s desperation to make Siri usable may have accidentally forced it to fix everything Siri depends on.
Deja Vu
And Siri certainly needed fixing.
Apple has gone from barely acknowledging artificial intelligence at WWDC a few years ago to making it the centre of the entire keynote. Siri AI is essentially the product Apple told us was coming in 2024, except this time it appears to exist.
That distinction matters. Apple’s original demonstrations promised a Siri that understood personal context, knew what was on the screen and could perform actions across applications. What users eventually received was a scattered collection of writing tools, notification summaries and image-generation features, while the Siri people actually wanted was repeatedly delayed. And goodness me, those tools were just abysmal. I cannot think of a single time I actually used them.
The new version finally attempts to fulfil that promise. It can search through messages, emails, photographs and files, understand what is currently displayed on screen and carry information between applications. It can answer broader questions using information from the web, remember previous conversations and be accessed through a dedicated app when speaking to an invisible orb no longer feels appropriate.
Why Siri AI Probably Won’t Fail
It is also built with substantial help from Google. Earlier this year, Apple and Google announced that the next generation of Apple Foundation Models would be based on Gemini models and Google’s cloud technology. That was probably the most sensible decision Apple could have made.
Apple’s original plan seemed to assume that it could quietly develop frontier-level models internally while maintaining its unusually strict approach to privacy. In retrospect, that was always an extremely difficult combination. Training a genuinely competitive general-purpose model requires a scale of data, infrastructure and specialist expertise that Google has spent years accumulating. Apple trying to recreate all of that from scratch, while simultaneously refusing to operate like an advertising company, left it starting several laps behind everyone else.
Using Gemini as a foundation allows Apple to concentrate on the part it is actually good at: integrating technology into its hardware and software without exposing every technical decision to the user. Requests can still be handled on-device where possible, while more demanding tasks are sent through Apple’s Private Cloud Compute system. Apple therefore gets access to a much stronger base model without simply placing the Gemini app at the centre of the iPhone or sending users’ personal information directly to Google.
Less Impressive, more Useful
The demonstrations were notably simpler than those shown at Google I/O. There were no grand promises that Siri would autonomously plan your entire week, browse dozens of websites and negotiate the purchase of a car while you slept. Based on what Apple showed, Siri AI is probably less capable than Google’s most advanced agents.
I am not convinced that is necessarily a problem.
Apple’s demos focused on the sorts of requests people might genuinely trust their phone to perform: finding an address buried in a message, identifying information from a photograph, organising an event or acting on something currently visible on screen. They were far less technically spectacular, but considerably closer to what most users expect from a personal assistant.
There is an important difference between an AI that can do something and an AI that you are comfortable allowing to do it. A chatbot getting a trivia question wrong is irritating. A system with access to your messages, files, passwords and applications getting an action wrong is potentially disastrous. Apple appears to understand that earning trust is as important as demonstrating intelligence, although we will only know how cautious Siri really is once more people begin using it.
Beyond Siri, the remaining updates are largely quality-of-life improvements, which is exactly what Apple’s operating systems needed.
Liquid glAss
Liquid Glass, introduced last year with the apparent goal of making every menu harder to read, has been refined throughout the system. Users can now adjust how transparent or tinted the interface appears, which is about as close as Apple ever comes to admitting it may have made a mistake. Menus have also been adjusted to improve contrast and make information easier to distinguish from whatever happens to be sitting behind it.
What About the Children!
The new parental controls are considerably more substantial. Screen Time has been redesigned, parents can approve new contacts, control which applications are available at different times of day and require permission before a child visits particular websites. Communication Safety can also intervene when explicit or violent content is detected.
These tools are genuinely useful, but their timing is interesting. Combined with the introduction of the cheaper MacBook Neo, they make Apple’s renewed interest in families and education extremely obvious. Apple once had a far more dominant position in schools, before cheaper Chromebooks became the default option for many institutions. A lower-cost Mac paired with genuinely strong parental controls gives the company a much more convincing route back into that market.
Of course, getting children onto Apple products also means introducing them to the ecosystem as early as possible. A child who grows up using iMessage, iCloud, AirDrop and a Mac is considerably more likely to remain an Apple customer when buying their own devices. The features are good for parents, but they are also excellent customer-acquisition tools. Both things can be true.
Catching, not Caught up
Safari received the type of AI features I would actually like to see more of. It can organise related tabs into groups, save them for later and monitor a webpage for changes through Notify Me. Rather than attempting to replace browsing with an automatically generated summary of the internet, it helps manage the browsing I am already doing.
That is my favourite category of AI: technology that makes it easier for me to do something, rather than insisting on doing the entire thing for me.
WWDC 2026 was not Apple suddenly leaping ahead of the rest of the industry. Siri AI appears more restrained than its competitors, many of the headline features are arriving in beta and Apple is still delivering promises made two years ago. It would be ridiculous to pretend otherwise.
But this was the first WWDC in several years where Apple appeared to understand what had gone wrong. Its operating systems needed refinement more than another redesign. Siri needed to become useful before it became revolutionary. Apple Intelligence needed models capable of supporting the promises placed upon it. Users needed more control over Liquid Glass, not more glass.
The most encouraging aspect of the event was not any individual AI demonstration. It was that Apple appears to have recognised that powerful hardware is meaningless when the software running on it feels neglected.
Perhaps the company’s fear of being left behind in AI was finally enough to make it care about the fundamentals again. Whatever the motivation, everyone benefits if the result is faster devices, reliable search and an assistant that doesn’t just spit out “I’m sorry, I didn’t get that” every time I talk to it.