Apple’s AI promises are finally, almost, sort of, here

The buried qualifier here is in the rollout itself: Apple has a documented pattern of announcing AI capabilities at WWDC and shipping them months later, in degraded form, or quietly shelving them entirely. 'Positioned as centerpiece' and 'actually shipping' are not the same thing, and the current announcement does not appear to confirm general availability dates for the full feature set.

This is largely disconnected from recent activity in our archive, as we have no prior coverage to anchor it to. It belongs, however, to a broader pattern visible across the industry: major platform vendors announcing on-device inference capabilities as a competitive response to cloud-native AI products, then struggling to close the gap between demo and deployment. Apple's specific challenge is that its privacy-first, local-inference framing is a genuine architectural constraint, not just a marketing position, and that constraint limits what the model can actually do without cloud offload.

Watch whether the full Siri AI feature set ships in the iOS 20 point-one update before the end of Q3 2026. If core LLM features slip to a later release or arrive with significant capability carve-outs, that confirms the announcement was pacing ahead of the actual product.