WWDC is a week away, so I've been doing what every iOS engineer does this time of year: half excited, half bracing myself. After fifteen years of these, the keynote stopped being a surprise and started being a checklist. Here's where I think the platforms actually are right now, from someone who ships on them for a living.
Swift finally feels finished
I'll say the controversial thing first: Swift concurrency is good now. For a couple of years, async/await and actors felt like a tax — you adopted them because you had to, fought the compiler, and quietly missed Grand Central Dispatch. That's mostly over. The strict concurrency checking that everyone complained about has settled into something you stop noticing, which is the highest compliment you can pay a language feature.
The flip side is that the language is big now. Onboarding a junior engineer into a modern Swift codebase means explaining actors, sendability, macros, and result builders before they've written anything useful. I don't think that's a crisis, but it's worth being honest that "approachable" isn't the word for Swift anymore.
SwiftUI crossed the line
When I shipped the first SwiftUI views in the adidas Training app years ago, it was a bet. You wrote the easy 80% in a day and then spent a week fighting the last 20% with UIViewRepresentable and weird layout bugs. I told people it wasn't ready for the hard screens.
It's ready now. I'd start a new app in SwiftUI without thinking twice, and I'd only drop to UIKit for very specific things. The tooling caught up too — previews are no longer a coin flip. What's still missing is honest debugging. When a SwiftUI view re-renders too often or a layout does something strange, you're still mostly guessing. That's the thing I'd most like Apple to fix, and it never makes a keynote because it doesn't demo well.
visionOS is still looking for its job
I want to be careful here, because I remember when people said the same thing about the Watch and it found its purpose. But three years in, visionOS still feels like a platform in search of the app that makes it necessary. The technology is genuinely impressive. The reason to use it every day is not.
I'm not building for it, and most teams I talk to aren't either. That can change fast — it changed for the Watch the moment fitness clicked — but I'm watching, not investing.
The actual question is intelligence
Everything else is a footnote next to this one. Apple spent the last two years catching up on AI in public, and the developer story is still thinner than it should be. I don't need another chat assistant. I need real on-device model APIs I can build features on, with predictable behaviour and no surprise latency.
The companies shipping the most interesting AI features right now mostly aren't doing it through Apple's frameworks — they're rolling their own. That's the gap I'll be watching the keynote for.
What I actually want next week
Less of a wishlist, more of a short, realistic ask:
- On-device model APIs that a normal app team can adopt without a research budget.
- SwiftUI debugging tools that tell me why a view updated, not just that it did.
- Audio frameworks that admit the Mac exists. I've spent months in Core Audio and the HAL building Spades, and the documentation in that corner of the system is from another era. (More on that another time.)
- Boring concurrency improvements — fewer footguns, better diagnostics.
None of that will get applause in the hall. All of it would make my week-to-week work better, which is the only metric I really care about after this many years.
Ask me again in two weeks how much of it actually shipped.