Apple tries to bury Siri’s AI vaporware past

Apple reannounces – After Siri’s 2024 AI promises didn’t make it into fall updates, Apple has spent the next two years reshuffling leadership and rebuilding its AI stack—aiming to deliver what it once demonstrated. At WWDC this week, the company put Siri AI, new Spotlight-style i

For two years. Siri’s big promised leap has sat in limbo—something Apple showed off with confidence. then failed to ship. The problem wasn’t subtle. At WWDC 2024. Apple described Siri taking on free-form. multistep tasks such as adding a photo to an email drafted to Madiha and Josh. showing photos of Stacy in New York wearing her pink coat. and searching Fitness+ for new yoga classes. When updates for iOS, iPadOS, and MacOS arrived that fall, “most of Siri’s new features were AWOL.”.

Apple later acknowledged the gap. In March 2025, it said the missing features were “unexpectedly challenging to implement” and expected to roll them out in “the coming year.” That timeline quietly stretched until “by the end of 2026.”

This week’s WWDC keynote changed the mood. Judging from the Siri update and other Apple Intelligence AI features unveiled in this week’s keynote at its WWDC developer conference. Apple is now trying to put the “epic vaporware fiasco” behind it—with a show-and-prove approach that tries to convince people the software isn’t just theater.

Apple also spent the intervening period not just on engineering—but on shifting the people at the center of AI leadership. Last December, Apple’s artificial intelligence chief, John Giannandrea, announced his retirement. His exit was part of a reshuffling that included the arrival of Amar Subramanya. a Google and Microsoft alum. as VP of AI. A few months later, Mike Rockwell—described here as the Apple Vision Pro godfather—was put in charge of Siri. Then. in January of this year. Apple and Google announced that Google’s Gemini AI model and cloud technology would be ingredients in future Apple AI experiences. including Siri.

By WWDC 2026, AI again dominated the keynote. Most notably, Apple effectively reannounced the Siri version that had fallen into limbo in 2024. This time the company didn’t treat it like a vague promise. The update is now dubbed “Siri AI. ” and in the keynote video Rockwell called it “profoundly more capable” than previous incarnations. Apple’s effort to show it was real came through in the presentation itself: the prerecorded keynote video had a cinema verité feel. including “pregnant pauses” as Siri AI processed requests before responding.

At subsequent WWDC briefings, executives leaned into live demonstrations of Siri and other AI updates. There was also a change in how Apple handled journalists’ access to developer software. In the past. Apple discouraged journalists who installed developer betas from publishing their impressions. preferring they wait for later. more polished public betas. This year, it “didn’t attempt to impose that moratorium.”.

Personal testing from the article’s author—running early AI features on an iPhone 16 Pro and iPad Pro—suggests Apple may be aiming for a rare kind of credibility: not just “look what we can do. ” but “watch what it does.” In early experiments. Siri AI handled multistep requests Apple has been touting. including “Text some photos of Aunt Liz to Marie” and “Find Mark Wilson’s Gmail address and add him as a contact.” The assistant also checked messages in the iPhone’s voicemail to help answer questions. acting as a reminder of what embedded assistants can reach inside a device. It could also answer questions based on what’s on-screen—for example. “Based on this article. who’s investing in SpaceX?”.

The most important shift for everyday users may be how Apple is positioning the AI. All of Apple’s new AI is integrated into iOS and iPadOS features such as Spotlight rather than being walled off inside a chatbot.

Still. there is a new product layer on top of that integration: a new Siri app. described here as a rough counterpart to ChatGPT. Claude. and Copilot. In the author’s tests, the app behaved with the cautiousness expected from Apple. It couldn’t generate images of specific humans at all, including Lincoln and Cleopatra. For people already satisfied with another AI bot—especially those paying for advanced features that Siri lacks—the app form may not be enough to persuade a switch. But it could draw interest from “casual users” who might not want the full commitment to competing paid assistants.

Apple’s privacy posture is also part of the pitch. Apple’s pledge of “privacy by design” is positioned as a practical advantage in an era where people are increasingly having sensitive conversations with AI about health—making assistants feel more intimate than traditional keyword search. Apple’s system hands off aspects of AI that can’t be processed on its devices to “Private Cloud Compute.” That infrastructure is described as not storing processed data and as not being readable by anyone. including Apple.

The competitive contrast is blunt. The author argues that Google, Meta, and OpenAI have strong incentives to use AI queries on free services for ad targeting purposes, calling that business model “inherently unnerving.”

Behind the scenes, Apple’s accelerated progress depends on more than one feature launch. The keynote and a follow-up “tech talk” attributed the sweeping capability improvements to an under-the-hood upgrade across Apple’s platforms. During the keynote video and tech talk, Craig Federighi—Apple software chief—acknowledged Gemini’s role in powering this update. But his framing matters: Gemini’s large language model is said to support Apple’s own models—known as Apple Foundation Models—rather than supplanting them.

Apple’s approach is described as isolated rather than intermingled with Google infrastructure. Federighi’s explanation says Apple’s new AI stack runs on “its own Nvidia-based servers in Google data centers.” In Google’s own products. Gemini is said to run on machines powered by Google’s custom TPU chips.

That distinction is offered as an intentional effort to dispel a fear that Apple Intelligence is merely relabeled Google technology. The author compares it to a licensing analogy involving Colonel Sanders’s “11 secret herbs and spices”: the idea is that even if some ingredients are the same. the dish isn’t KFC.

There is also a more structural reason the story reads like a turnaround. Siri’s long history is cited as a major factor in why reimagining it for the post-ChatGPT era proved so difficult—by 2026 AI standards. Siri is described as having “the engine of a Model T.” The new platform Federighi outlined is presented as a “much more modern” foundation. positioned to enable not only the WWDC AI features. but “many more it hasn’t even thought of yet.”.

Apple’s broader goal is clear in the way it ties AI into a wide range of hardware and operating system features. rather than treating it as a standalone novelty. Other operating system and device makers are still “figuring that out. ” leaving Apple “less behind than it might have been if they’d made more progress in the past two years.” The author points to Microsoft’s ongoing push for the “age of the AI PC. ” but says it doesn’t yet feel like it’s truly arrived.

Competition, though, is not standing still. On the AI-everywhere front, Apple’s biggest rival is its partner: Google. Less than three weeks before WWDC. Google announced “100 things” at its I/O conference. with Gemini Spark—an agent described as having access to data stored in Google products like Gmail and Drive—framed as a standout. Spark is positioned as a personal AI agent that can work more independently than what Apple previewed at WWDC.

Still, WWDC 2026 did not come off as purely defensive. Among its most intriguing announcements were new abilities to “vibe code” Shortcuts and Safari extensions for iPhones, iPads, and Macs. The idea is to give relatively nontechnical people the power to create new features for their devices. In the first developer beta. these tools are described as “a tad rudimentary. ” but also “useful. fun. and agentic by nature.” The author also highlights another Safari AI addition. Notify Me. which can autonomously monitor webpages for changes—such as an out-of-stock product becoming available again. They are not Spark, but are described as having “a spark of life.”.

For now, early developer betas can only show so much. The author notes that Apple’s OS updates are likely to evolve before shipping in the fall. And with AI moving at an aggressive pace. Apple could fall behind again if WWDCs in 2027 and beyond don’t deliver. Even so. the piece’s core point is that the “mojo” Apple demonstrated this week is more significant than any single feature. The real test is whether Apple can convert a two-year recovery effort into lasting momentum—this time with the software actually arriving when it’s promised.

Apple Siri AI WWDC 2026 vaporware Siri John Giannandrea Amar Subramanya Mike Rockwell Google Gemini Apple Intelligence Private Cloud Compute Spotlight Shortcuts Safari Notify Me