> WWDC 2026 Predictions Report Card: What Apple Actually Shipped

In late May I drafted a predictions post for WWDC 2026 and never hit publish. The keynote came and went, the coverage treadmill started, and the draft sat there quietly being wrong in some places and right in others. Now that iOS 27 has been out for a few weeks and the cycle is nearly complete, grading those predictions is more useful than publishing them ever would have been. It's a good calibration exercise: the misses say as much about how Apple operates as the hits do.

Here's what I predicted, what shipped, and the grade.

SwiftUI: Right Theme, Wrong Features

The prediction: refinements to Liquid Glass, more built-in components adopting the new design, better container and layout APIs, and navigation polish with deeper App Intents integration.

What shipped was better than what I asked for. The container story did get a major upgrade, just not the shape I imagined: drag-to-reorder now works in any container, swipe actions escaped List, and alerts picked up item: bindings. AsyncImage caching, the single most requested fix I can think of, wasn't on my list because I'd stopped believing it would happen. Meanwhile the navigation APIs I predicted would get attention were left alone, and the @State macro migration, the change most likely to actually touch your codebase, appeared in nobody's predictions anywhere.

Grade: B. Right that containers were the growth area, wrong about nearly every specific.

Swift: Closest to the Mark

The prediction: smoother strict concurrency adoption, better diagnostics, and ergonomics work rather than big new features.

This one mostly landed. Swift 6.4 is exactly an ergonomics release: async defer and task cancellation shields each delete a real workaround from production codebases, and neither introduces a migration. I'll also quietly count module selectors, which shipped in 6.3 before the conference, since "fixing awkward corners" was the spirit of the prediction. The ownership features I floated didn't materialize, and result builder error messages are still result builder error messages.

Grade: A-. Predicting "boring in a good way" was the safe bet, and it paid out.

On-Device AI and Xcode: Half Credit

The prediction: a push to make foundation models easier to run on-device, and Apple's own take on AI-assisted development in Xcode.

The first half was a year stale. The Foundation Models framework shipped at WWDC 2025, and this year's updates were iterative rather than a new framework. The second half, though, is my favorite hit of the whole card: Xcode 27 ships Apple's own agent skills and lets you export them to Claude Code, Codex, or Cursor. I predicted the category and still managed to be surprised by the execution. Letting Apple's guidance travel to other tools is more useful than any chat panel would have been.

Grade: B-. One hit, one prediction that had already happened.

Platforms: The Numbering Ambush

The prediction: visionOS 4 and watchOS 14.

Every platform is just 27 now. The unified numbering that started with the 26 releases apparently needed one more year to fully sink in for me. The actual platform work was real but unflashy, and the StoreKit side I didn't predict at all: commitment plans and a reworked offer code flow, plus multiseat subscriptions that quietly turned themselves on in App Store Connect this fall.

Grade: D for the version numbers, which I'd have gotten right by thinking about it for ten more seconds.

What Nobody Saw Coming

The predictions draft's biggest gap wasn't a feature, it was breakage. iOS 27 requires the UIKit scene-based life cycle, which stopped older apps from launching at all, and Xcode 27 raised the floor on supported deployment targets, which broke builds the morning people updated. If I write a predictions post next year, "what will Apple remove" gets its own section, because the removals defined more developer-hours this cycle than any addition.

And then there's the one nobody could have graded: the fall half of the cycle. iPhone Duo arrives October 23 running iOS 27.1, with fold-aware layouts, vertical toolbars, and a whole adaptation checklist that has dominated this site for a month. WWDC set up the APIs that made it possible, size classes and adaptive containers doing exactly what Apple spent years saying they were for, but the keynote gave no hint the payoff was four months away.

The Takeaway

Scoring it honestly: solid on direction, weak on specifics, blind to removals and hardware. Which is roughly how predicting Apple always goes. The practical lesson from this cycle is that the safe-sounding work, rebuilding with the new SDK early and taking the migration warnings seriously, was worth more than correctly guessing any feature. The teams that did that in June had a quiet September and an easy road to Duo readiness. The full list of what shipped, with links to everything we wrote about it, lives on the WWDC 2026 recap page.