Apple’s Spin on the Personalized Siri Apple Intelligence Reset

Thursday, 12 June 2025

Michael Tsai, “Apple’s Spin on AI and iPadOS Multitasking”:

I do want to call out that, in multiple interviews, they are kind of setting up strawmen to knock down. They keep saying that people say Apple is behind in AI because it doesn’t have its own chatbot. To me, Apple has been clear that it has a different strategy, and I think that strategy mostly makes sense. I have never heard someone wish for an Apple chatbot. The issue is that everyone can see that Apple seems behind in executing said strategy, both that features didn’t ship on time and that the ones that did ship don’t measure up to similar features from other companies.

Secondly, they seem to be trying to debunk John Gruber’s claim that Apple showed vaporware at the last WWDC. But Apple’s assertion that there was actual, working software doesn’t contradict anything Gruber wrote. He put it at level 0/4 because there wasn’t even a live demo, just a pre-packaged video. If it can’t be demoed to the media in a controlled setting, even calling it “demoware” would be charitable. Wikipedia says, “After Dyson’s article, the word ‘vaporware’ became popular among writers in the personal computer software industry as a way to describe products they believed took too long to be released after their first announcement.” Is that not exactly what happened here?

The whole “Siri, when is my mom’s flight landing?” segment of last year’s WWDC keynote definitely shouldn’t qualify as “demoware” either. It was never demoed. Whether the feature was actually running, and actually capable of doing what they said it could, even just some of the time along a golden path, ultimately doesn’t matter. Even the keynote video didn’t show the actual feature working. It kept cutting away from the iPhone that was purportedly performing the feature back to presenter Kelsey Peterson at every single step. Apple’s internal rules for keynote demos say that the entire feature has to be real, and capturable in a single take of video. I’ve spoken to people who’ve been in keynotes, and many more who’ve done WWDC session videos. Apple has strict rules about everything being real. That doesn’t mean they always show the feature in a single take in the final cut of the presentation, but it has to be possible, just like it would have to be in a live stage presentation. But that Siri demo in last year’s keynote is almost like a series of screenshots. We never see Peterson speak to Siri and then watch the results come in. There’s not one single shot in the whole demo that shows one action leading to the next. It’s all cut together in an unusual way for Apple keynote demos. Go see for yourself at the 1h:22m mark.

I spoke this week, off the record, to multiple trusted sources in Apple’s software engineering group, and none of them ever saw an internal build of iOS that had this feature before last year’s keynote. That doesn’t mean there wasn’t such a build (see next paragraph). But none of my sources ever saw one, and they found that to be exceptionally unusual, because they’re in positions where they believe that if there had been such a build, their teams would have had access to it. Most rank and file engineers within Apple do not believe that feature existed in an even vaguely functional state a year ago, and the first any of them ever heard of it was when they watched the keynote with the rest of us on the first day of WWDC last year.

But at this point, based on Federighi’s and Joswiak’s public statements (e.g. in their interview with Joanna Stern this week), and some other things I’ve heard from little birdies this week, I’m willing to stipulate that there was, let’s call it, “working code” for the personalized Siri feature a year ago. I can’t verify that, but I’m willing to stipulate it if only for the sake of argument, and to put to rest any notion that the feature was completely imaginary at this point one year ago, which clearly isn’t the case. At least one reason why the feature, as presented in last year’s keynote, was edited like it was is that the latency was so bad. Whatever state it was in, it couldn’t be shown in a single take. Which means that, as shown in the 2024 keynote, it broke the “rule” that demos in the new pre-recorded format should be as rigorously “real” as they would be if the demos had to be performed live on stage.

I’m quite certain Apple’s executives believed this feature could be shipped at some point in the iOS 18 year. It’d be crazy to announce the feature if they didn’t believe they could ship it, and Apple’s executives aren’t crazy. I’m also quite certain that eventually there was a truly functional implementation of the now-abandoned “v1” of the more personalized Siri, but it was unreliable with no path forward to make it reliable. (I think it was far worse than “not up to Apple’s high standards” — it was clearly unshippable.)

But, as Tsai notes, from our perspective it doesn’t really matter it was real, in any sense, a year ago. It’s still vaporware at this point. Vaporware doesn’t mean “completely fabricated”. It just means “promised but hasn’t shipped”. Tsai links to this Mastodon tweet from Russell Ivanovic:

“This narrative that it was vaporware is nonsense”. Craig Apple. My guy. You announced something that never shipped. You made ads for it. You tried to sell iPhones based on it. What’s the difference if you had it running internally or not. Still vaporware. Zero difference.

Also, Apple is sticking with the euphemism “in the coming year” for when we can expect to see these next-gen personalized Siri features. Gurman reported today that they’re shooting for next spring. I confirmed with Apple at WWDC that “in the coming year” means “in 2026”. I don’t know why they’re sticking with that euphemistic phrasing, which to many people’s ears makes it sound like in the next 12 months, which might include this fall. Effectively, Apple could ship this feature in December 2026 and still hit their self-imposed “in the coming year” deadline — but that phrasing has left many people with the impression that the deadline is June 2026 and maybe it’ll ship at the end of this year. Just say “next year” instead of “in the coming year”. It’s very obvious that this year’s WWDC keynote went back to an underpromise/overdeliver mindset. But “in the coming year” is raising some users’ hopes misleadingly.