- Apple introduced a rebuilt Foundation Models framework developed in collaboration with Google's Gemini family.
- The new Liquid Glass design system for version 27 mandates resizable apps, hinting at future foldable hardware.
- Xcode 27 arrives with a 30% smaller footprint and integrated support for AI coding agents from Google and Anthropic.
- The Device Hub replaces the traditional Simulator, offering enhanced control over virtual and physical hardware testing.
Following the hype of the morning keynote, the Platforms State of the Union offers a much more technical look at how the ecosystem is evolving for those who actually build the apps. It’s essentially a second keynote specifically for the developer community, where the company moves past the marketing slides and dives into the nitty-gritty of APIs, frameworks, and low-level system changes that define the user experience.
This year’s session was particularly heavy on how intelligence is being woven into the core of the OS. Rather than just treating AI as a separate feature, the focus has shifted toward providing tools that allow third-party software to leverage on-device models and cloud-based computation seamlessly, while maintaining the privacy standards users have come to expect from the brand.
The New Architecture for Apple Intelligence and Core AI
A significant reveal during the session was the complete overhaul of the Foundation Models, which were surprisingly developed in tandem with Google using tech from the Gemini family. This partnership aims to balance local processing with high-end cloud capabilities, allowing developers to prompt models with images and text. To keep things open, the Foundation models framework is slated to become open-source later this year, which is quite a departure from the usual walled-garden approach.
For those building privacy-first features, the introduction of the Core AI framework is a game-changer. It is built to maximize the performance of Apple Silicon when running models locally on the device. Developers can also take advantage of Dynamic Profiles, which make it easy to swap models or instructions mid-session, ensuring that an app’s behavior stays relevant to what the user is doing at that exact moment.
Siri is also getting a major brain transplant via updated App Intents. By adopting new schemas and view annotations, apps can allow Siri to discover and interact with their internal features through natural language. This means the assistant can actually see what is on the screen and perform actions on the user’s behalf, such as moving files or editing content, without the user ever lifting a finger.
Refining the Visual Language with Liquid Glass
The design language for the version 27 releases has been polished under the Liquid Glass moniker. Legibility has been improved across the board, with new diffusion effects applied to backgrounds beneath toolbars to keep text crisp. A nice touch for users is a new configurable slider to adjust opacity, letting people decide exactly how much transparency they want in their interface elements.
Mac users will notice that the sidebar is now completely separated rather than inset, featuring vibrant, colorful icons that pop more than before. Additionally, app icons now feature a refraction effect that gives them a more tactile, glassy appearance. Developers can utilize the updated Icon Composer app to ensure their branding looks sharp across all 27-series updates, which now force all mobile apps to be freely resizable.
This move to mandatory resizability is officially for iPhone mirroring on the Mac, but it’s hard not to notice how it perfectly prepares the ecosystem for a rumored foldable device. By ensuring apps can handle any aspect ratio gracefully, the transition to new hardware form factors becomes much less of a headache for the people writing the code.
Productivity Gains in Swift and SwiftUI
SwiftUI continues to be the primary focus for cross-platform development, and the latest updates emphasize performance and flexibility. A new API now allows for easy list reordering and custom swipe actions, while text selection has been significantly improved for complex layouts like vertical text. Performance-wise, loading images and resizing elements on the Mac should feel much snappier than in previous versions.
On the language side, Swift 6.4 brings several quality-of-life upgrades that make daily coding a bit smoother. You can now hide specific warnings in certain code blocks if you aren’t ready to address them, or conversely, promote them to errors for mission-critical sections. Interestingly, the company is moving even more of its low-level kernel and WebKit code over to Swift, moving away from older, less memory-safe languages.
Developers will also appreciate the simplified availability checks. Instead of listing every OS version individually, a single AnyAppleOS 27.0 tag can be used to set feature availability across the entire ecosystem. The compiler has also been optimized to provide more descriptive error messages, which should help track down bugs without having to decipher cryptic logs for hours.
A Smaller, Smarter Xcode 27 and Device Hub
Xcode 27 has seen a massive diet, coming in at 30% smaller than its predecessor by dropping support for older architectures. It now features iCloud syncing for developer preferences, so you don’t have to spend all day setting up a new machine. The toolbar is now fully customizable, and for those who like a bit of flair, new themes like Neon Noir and Coral Reef have been added to the editor environment.
The most notable change in the toolset is the retirement of the old Simulator in favor of the Device Hub. This new app consolidates virtual and physical hardware management into one place, allowing for live resizing and instant toggles for dark mode or text size. It even supports multitouch gestures directly from the Mac, making it much easier to test how an app feels to the touch.
AI integration has also reached the IDE, with Xcode 27 supporting the Agent-Client Protocol (ACP). This allows coding agents from Google Gemini, OpenAI, and Anthropic to help with generating code, running tests, and even localizing text based on the visual context of the app. New plugins for Figma and GitHub integrate these agents directly into the workflow, potentially speeding up the jump from a design mockup to a working prototype.
The evolution of these platforms shows a clear trend toward making complex AI and design tasks more accessible to the average developer while tightening the integration between different devices. Between the rebuilt intelligence frameworks and the refined visuals of Liquid Glass, version 27 appears to be less about reinventing the wheel and more about polishing the existing infrastructure to handle the next era of computing. As these tools roll out, the focus remains on ensuring that apps remain performant and private, regardless of how much machine learning is happening behind the scenes.
