WWDC 2026: Apple's AI Reset — Core AI Framework, New Siri, and Opening the Platform

Apple’s developer conference has always been the most closely watched week in the consumer technology calendar, but WWDC 2026, which opens June 8 at Apple Park in Cupertino, arrives with a peculiar mixture of excitement and reckoning. The company that invented the smartphone assistant is walking into this keynote having watched rivals ship generation after generation of genuinely capable AI while Siri continued to fumble basic requests. The pressure to respond is immense — and the leaks suggest Apple is finally ready to do so at scale.

Core AI: Retiring a Decade-Old Framework

The most significant developer-facing announcement expected at WWDC 2026 is the replacement of Core ML with a new framework called Core AI. Apple introduced Core ML in 2017 as its developer API for on-device machine learning, and it served its purpose in an era when ML meant image classification models and natural language pipelines measured in tens of megabytes.

The generative AI era has made Core ML a mismatch. Modern on-device inference involves models measured in gigabytes, new numerical formats like FP16 and INT4, and workloads that demand tight integration between the Neural Engine, GPU, and CPU memory pools. Core AI is Apple’s attempt to build a framework that fits 2026 rather than 2017.

According to developer reports and 9to5Mac sources, Core AI will expose new APIs for local LLM inference, multimodal inputs (text, image, audio simultaneously), and streaming token generation — the fundamental building block of responsive chat-style applications. Critically, Core AI is expected to offer deeper access to Apple Silicon’s Neural Engine than Core ML ever provided, which should unlock substantially better performance for third-party apps running on-device models.

For developers, this is significant. The current path to on-device LLMs on iOS involves working around Core ML’s constraints or using alternative frameworks like llama.cpp via bridging headers — solutions that are functional but fragile. Core AI, if it delivers on expectations, provides a first-class native path.

Siri’s Second Act

Siri is getting its most consequential redesign since its 2011 launch. Based on extensive leaks — including code found in developer betas and design mockups shared with 9to5Mac and MacRumors — iOS 27 will ship with a fundamentally different Siri that is simultaneously more capable and more honest about its limits.

The most visible change is structural: Siri is becoming a standalone app. The new dedicated Siri application allows users to review conversation history, start new chat sessions, upload images and documents for analysis, and switch between voice and text input without leaving a persistent context. This persistent conversation model — standard on ChatGPT, Claude, and Gemini — has been conspicuously absent from Siri, which treated every query as a stateless event.

The interface design is equally striking. The new Siri uses an all-dark color scheme, with no light mode currently in development for the redesigned UI. When activated, Siri displays a pill-shaped animation in the Dynamic Island — a visual treatment that transforms the hardware cutout from a notification surface into an always-accessible AI entry point. Query responses appear as transparent cards that float over the current app, and swiping expands into a full chat-like conversation mode with the persistent history.

Apple is also reportedly building onscreen awareness into the new Siri — the ability to understand and act on what is currently displayed on screen. This has been a marquee feature of Gemini Live on Android for over a year, and its absence from Apple Intelligence has been a consistent criticism. With iOS 27, Siri would be able to answer questions about a webpage you’re reading, summarize a document open in the foreground, or take actions in third-party apps with full visual context.

Opening the Garden (A Little)

Perhaps the most surprising development expected at WWDC 2026 is Apple’s apparent willingness to introduce genuine third-party AI competition within its own platform. The proposed Extensions system — previewed in regulatory filings related to the EU Digital Markets Act and confirmed by multiple developer sources — would allow users to designate alternative AI assistants as default handlers for certain query types.

In practice, this means that a user could configure Siri to route complex reasoning tasks to Claude, delegate coding questions to a GitHub Copilot integration, or hand math and science questions to Wolfram. The routing happens within the Apple Intelligence layer, maintaining the privacy guarantees and on-device processing for simpler queries while leveraging external models for tasks that require them.

The Extension architecture uses a standardized API that third-party AI providers must implement, with Apple retaining control over which extensions are approved for distribution through the App Store. This is not an open ecosystem — it is a carefully bounded aperture through which Apple can claim platform openness while preserving its gatekeeping role.

Anthropic, Google, and Microsoft have reportedly been in discussions with Apple about integration, though xAI’s Grok integration is reportedly complicated by Elon Musk’s public statements about Apple’s AI partnerships.

The Developer Platform

Beyond Siri and Core AI, WWDC 2026 is expected to preview Swift 7 with native AI task types, updated Xcode AI assistance powered by stronger on-device models, and new APIs for Vision Pro that accelerate spatial AI use cases. Apple has confirmed that more than 100 video sessions will be available during the conference week, with a heavy concentration on AI tooling for developers.

The developer trajectory is important context. Apple’s on-device AI strategy has always been built around privacy as a differentiator — processing data locally, with Private Cloud Compute handling overflow workloads in Apple-controlled data centers. This architecture requires developers to build within Apple’s frameworks, which has historically limited what they can do compared to the cloud-native AI tools available on Android or desktop platforms. Core AI is Apple’s attempt to close that gap.

Why This Moment Is Different

Apple has previewed AI capabilities at previous WWDCs that subsequently shipped later or weaker than announced. The original Apple Intelligence reveal in 2024 generated enormous expectations; the reality of Siri’s actual capabilities at launch disappointed many developers and users. The “delayed ambitions” narrative has been a recurring theme.

What may be different in 2026 is competitive pressure from a direction Apple cannot ignore: the enterprise. Microsoft’s AI Copilot, Google’s Gemini Workspace, and Anthropic’s Claude for Work are all competing aggressively for enterprise software spend. If Apple cannot offer IT decision-makers a credible AI-native platform story in iOS 27 and macOS 27, it risks ceding corporate deployments to Android and web-based alternatives in ways that compound over time.

The Core AI framework, the Siri redesign, and the Extensions system are all pieces of the same strategic answer: Apple is repositioning itself not as the company that builds one AI assistant, but as the platform that integrates AI everywhere — with the user in control of which models do what work.

Whether the keynote on June 8 delivers the goods remains to be seen. But the architectural pieces are in place for WWDC 2026 to be Apple’s most consequential developer conference since the introduction of Swift in 2014.