Xcode goes agentic and Core AI arrives as the new developer model framework

At WWDC 2026, Apple introduced agentic capabilities in Xcode — including full app simulation — and unveiled Core AI as the successor to Core ML, with image input, custom skills, and server models in the Foundation Models API.

Xcode simulates apps and generates code autonomously

Craig Federighi presented a significantly more capable Xcode at WWDC 2026. MacRumors and Engadget confirm that the built-in coding assistant has been expanded to handle app localization and can now interact with device simulators, simulating entire app flows for testing and building features. Federighi described Xcode as "the best place" for agentic coding — a claim that did not go unnoticed among attending developers, well aware of how much ground tools like Cursor, Claude Code, and Copilot have gained.

On the models front, Apple announced an updated Foundation Models framework with three concrete additions: image input (developers can now pass images alongside text to on-device models), custom skills (to extend model capabilities with custom logic), and server models (to execute models server-side rather than exclusively on-device). Apple also introduced Core AI as a new framework that complements — and will eventually replace — Core ML.

The developer picture

Taken together, these announcements mark a shift in Apple's posture toward AI developers: 2025 introduced Foundation Models APIs; 2026 makes them multimodal and opens them to the cloud. The App Intents Framework, demonstrated during the Siri AI section, now enables third-party apps such as Line to integrate with Siri to perform actions on the user's behalf — a step toward an ecosystem where Siri AI acts as an orchestrator across applications.

← Back to home