Apple Is Still Behind on AI. Its WWDC Plan Might Still Be the One That Matters.

Apple did not catch up on AI this week. Let's start there.

WWDC 2026 was a plan, not a product. A promising plan - a genuinely smarter Siri, an assistant that reaches across your apps, on-device models developers can build on. But nothing shipped. It's a roadmap with a demo reel attached, and we have been here before.

Remember 2024? Apple unveiled Apple Intelligence and a reimagined Siri with a gorgeous keynote and marketing to match. It looked incredible. Then the headline features slipped, and slipped again, and the Siri they sold us never really arrived. So when Apple stands on stage at WWDC 2026 and promises an assistant that can actually do things, the right posture is the one Linus Tech Tips landed on in front of half a million viewers: "Apple's going to have to forgive me for being a little skeptical that they're going to actually ship this on time." They have earned that skepticism. Twice now the marketing has run ahead of the product.

So why am I still paying close attention? Because of what the plan is aimed at - and that is the part the scorekeepers miss. It's the same question I keep coming back to in the last post: not whether the AI is clever, but whether it can act, and whether ordinary people will ever actually use it that way.

Apple doesn't need to be first. It needs to reach everyone - eventually.

Here is what the benchmark crowd keeps missing. Agentic AI - AI that does multi-step things on your behalf - has one enormous unsolved problem, and it isn't intelligence. It's distribution. The people building the most capable agents reach developers and early adopters. They do not reach your mum. Your mum has an iPhone.

The hard part of agentic AI was never the agent. It's distribution - and Apple is the only one holding a billion devices. The open question is whether they can actually deliver.

Apple spent fifteen years teaching every company on earth to become an app. If it delivers what it described, that groundwork becomes the moat: the future of an app on Apple's platforms stops being "can I get this approved" and becomes "can an assistant reach into it and do the thing." At WWDC, Apple described opening on-device model access through a native Swift framework and wiring Siri to act across apps. Described. Not shipped - and that distinction is the whole story right now.

Why I'm watching the plan, not the scoreboard

The capability already exists out in the world. I have spent months watching agentic AI do genuinely remarkable things in developer tools and my own small scripts. What is missing everywhere is the mainstream on-ramp - the part where a normal person, who has never heard the word "agentic" and never will, taps their phone and an errand just gets done. Apple is the only company positioned to build that on-ramp, because it would arrive as an update to a thing people already hold for four hours a day. WWDC 2026 was Apple finally pointing at that on-ramp. Pointing is not arriving.

Apple is behind, unproven, and out of free passes on AI. But it is still holding the one door the rest of us have been trying to walk a billion people through. I'm not betting that they deliver. I'm watching whether they do - because if anyone can make agentic AI ordinary, it's them, and this was the first time they pointed at a credible way to do it.