← analysis

Apple at WWDC 2026: The Widest Gap on the Board

Apple holds the best on-device silicon, the deepest personal context, and the least capable assistant of any major provider. WWDC 2026 is where we find out if that changes — and what to watch for, dimension by dimension.

Myagi·June 8, 2026·7 min read

Update · June 8, 2026, after the keynote. Apple delivered. Siri is rebuilt, conversational, and context-aware — and its hard reasoning now runs on Google's Gemini. We re-graded six Apple dimensions live; the scores below auto-update, and every move is logged in the WWDC 2026 live feed. The piece below was written as a preview; the analysis held up, so we've left it standing with this note. The short version: intelligence jumped, the model-strategy weakness was confirmed, not fixed.

No company on this scorecard has a wider gap between what it owns and what its assistant can do than Apple. It scores a 9 on Apple On-Device AI 9 for the silicon and privacy architecture beneath it. It went into WWDC scoring a 4 on both Apple Assistant Intelligence 6 and Apple Agent Capability 5 (re-graded above after the keynote). That gap was not a rounding error. It was the entire story of Apple's AI position in one line.

The interesting question isn't whether Apple is behind. Everyone knows it's behind. The interesting question is whether the assets it holds — distribution, silicon, and the most intimate personal context on earth — are the things that actually matter once raw model intelligence commoditizes. WWDC 2026 is the first real test of that thesis.

The assets are real

Apple ships a frontier-class inference substrate to a billion pockets. Private Cloud Compute extends that to server-class models without surrendering the privacy story. The device already holds your messages, your calendar, your photos, your relationships, your routines — the context every other assistant has to beg, borrow, or reconstruct. That's why Apple Personal Context 8 ranked high even when the assistant could barely hold a conversation.

And then there's App Intents — the most underrated primitive in the entire agent race. It is, in effect, a system-wide registry of actions that apps already expose to Siri and Shortcuts. If that becomes the substrate an intelligent assistant orchestrates over, Apple doesn't need to win the model race to win the action race. App Intents entered the keynote as Apple's most underused asset — Apple Cross-App Actions 6 — and WWDC 2026 began, finally, to activate it.

The gaps are just as real

The assistant is the problem. Siri's conversational quality lags every major competitor, its reasoning is well behind the frontier, and its agency is essentially nonexistent — it suggests, it doesn't do. Apple's Apple Model Strategy 4 is a 4 because it has no frontier stack of its own and has been forced toward partner models for anything genuinely hard.

That's the uncomfortable trade. To make Siri smart quickly, Apple likely has to route hard reasoning to someone else's model — Gemini, Claude, or GPT. Doing so dents the model-strategy score and the narrative of self-reliance, but it's the fastest path to lifting assistant intelligence and, with it, the Apple MyAGI Alignment 5 north-star grade that began the day at a 4 — and is exactly the trade Apple made.

What we're grading at the keynote

The swing factors aren't the consumer features. They're four platform decisions.

1. The Siri rebuild. Does it close the reasoning gap, and what powers it? A credible answer moves assistant intelligence and conversational UX together.
2. App Intents. Any expansion that lets an assistant chain actions across third-party apps is the single highest-leverage change Apple can make. This is where cross-app-actions and agent capability live.
3. The Foundation Models framework. How much on-device and Private Cloud Compute inference opens to developers determines whether Apple Developer APIs 5 moves off a 5.
4. A model strategy that's actually a strategy. Owning small models, renting frontier reasoning, and saying so plainly would be a more honest 2026 position than pretending to do it all in-house.

Why it matters beyond Apple

If Apple's bet is right — that context, distribution, and a trustworthy action layer beat raw IQ once everyone has access to good-enough models — then the scorecard's center of gravity shifts away from benchmark intelligence and toward exactly the dimensions Apple is strong on. If it's wrong, Apple spends another year as the giant with the best hardware and the worst assistant.

Either way, we'll grade it live as it happens, and watch AAPL for the market's verdict in real time.