Five changes from WWDC26 for iOS developers

On June 8, 2026, at WWDC26, Apple announced iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27, and tvOS 27 (Apple Newsroom). According to the same press release, the developer beta was available from the day of the announcement, and the release comes this fall as a free software update.

This post sorts what reaches people who build iOS apps into five areas. The announced facts are what I rechecked on October 4, 2026 on six official Apple pages, and what comes later in each section is my own view. Results of implementing or measuring anything on the beta are not in this post.

Exposing app content and actions to the system

According to the WWDC26 iOS guide, Siri reaches the content and actions of an app in natural language through the App Intents framework. Entity schemas contribute the content of the app to the Spotlight semantic index. With intent schemas, people can ask for actions on that content in natural language, with no specific phrases to define. The new View Annotations API maps views to entities, so that what is on screen can be referred to and acted on in conversation. The App Intents Testing framework validates the integration through real system pathways, without UI automation.

From here on it is my view. The ways into a feature from Siri and Spotlight, outside the app, are getting wider, so I see App Intents support as something to put into the first design, not an extra convenience. I expect it to be harder to add later. I recommend deciding first which entities and actions to expose. For a sports app the candidates are games, teams, leagues, and schedules, and for a commerce app they are products, orders, the cart, and delivery status. These examples are mine and are not in Apple’s guide.

Where the model runs

The same iOS guide introduces the Foundation Models framework as a Swift API for the on-device model that Apple Intelligence uses. It says the framework now works with Apple Foundation Models, with cloud models like Claude and Gemini, and with any other provider that conforms to the Language Model protocol. Multimodal prompts pass images alongside text, and Dynamic Profiles swap models, tools, and instructions within a continuous session. The Evaluations framework is a tool for verifying that AI features behave correctly across conditions.

Core AI is a new framework for running your own models on-device. It is built into the OS and made for Apple Silicon, and models are loaded and run through a Swift API. The guide names ahead-of-time compilation, zero-copy data paths, and stateful execution as its performance elements. For Private Cloud Compute, there is a condition. The guide says that if you are enrolled in the App Store Small Business Program and your app has fewer than 2 million total first-time downloads, you can access the next generation of Apple Foundation Models running on Private Cloud Compute at no cloud API cost.

As I read it, the two have different roles. Foundation Models connects language and multimodal models to an app, and Core AI runs a model you already have on Apple Silicon. I recommend looking first at handling summarizing, classifying, recommending, and input correction on the device, and at Private Cloud Compute or an outside provider for large inference and generation. For sensitive information, check whether it can be handled on the device before it is sent out. The things to compare are cost, latency, and privacy. Also look at whether testing and evaluation can be automated.

Xcode 27 and agents

According to the summary of the Platforms State of the Union, Xcode 27 is Apple silicon only. Projects load faster, settings sync via iCloud, and the toolbar is customizable. The new Device Hub replaces Simulator and brings virtual and physical devices together in one place. The same article says agents in Xcode can run tests, try things in Playground, run apps in the Simulator, fix issues, and localize apps. Plugins add skills and MCP tools, and other agents connect through the Agent Client Protocol. The WWDC26 Xcode guide covers Device Hub, improvements to Instruments, and localization with agents.

The sentence that Device Hub replaces Simulator and the sentence that agents run apps in the Simulator are in the same article. From that article alone I could not tell whether only the management screen changed or the runtime changed as well.

If you develop on an Intel Mac or run CI builds on one, I think a plan to move is needed. I expect agents to be useful for validating small changes, analyzing test failures, and tidying string catalogs. Even so, I recommend treating generated code as a draft and always reviewing the diff. A team does well to agree first on how far agents are used and on the review rules. Changes that touch architecture, authentication, payments, storage, or sync deserve a stricter look. I think decisions about structure and quality have to stay with the developer.

The material and the size of the screen

According to the iOS 27 page, Liquid Glass has more uniform refraction and improved contrast, and app icons are sharper. A slider lets users adjust Liquid Glass from ultraclear to fully tinted. The page also gives performance figures. App launches are up to 30% faster, new photos load in Photos up to 70% faster, and AirDrop transfers are up to 80% faster. These numbers are upper bounds that Apple measured on specific devices under specific conditions, and the footnotes give the test conditions. They do not mean every app gets that much faster.

The State of the Union summary says iOS apps are now resizable. People can use larger displays on iPad, and in iPhone Mirroring on Mac.

I see custom navigation bars, tab bars, toolbars, blur backgrounds, translucent overlays, and floating buttons that an app implements itself as things to check again. Confirm that they stay readable with Dynamic Type, dark mode, Reduce Transparency, and increased contrast. A layout that assumes the size of an iPhone can break when the size changes, so I recommend checking touch areas, scrolling, and transitions on a real device.

Subscription products and review submission

According to the WWDC26 App Store guide, multiple In-App Purchases, including subscriptions, can be grouped into a single submission in App Store Connect. They can also go together with other review items such as In-App Events, custom product pages, and product page optimization tests.

Retention Messaging shows subscribers a message about the value of the subscription. Apple describes it as not adding friction to the cancellation flow, and says a special offer can be shown with it. It is listed as coming this fall. Subscription Bundles let people buy several auto-renewable subscriptions as a single subscription, and Suites offer a set of subscriptions that are not sold standalone. The guide says more on how to request these two will come later in the summer. Group Purchases let one subscriber buy multiple seats in a single purchase and invite others. Volume Purchasing makes a subscription available to enterprise and education buyers through Apple School Manager and Apple Business.

If you run subscriptions, I do not think the StoreKit code is the only thing to look at. The product model, operations in App Store Connect, review, entitlements after purchase, and the cancellation process need to be reviewed together. Pricing and bundling, the response to churn, and sales to organizations are questions to settle together with policy.

What this post relies on is Apple’s press release and guide pages as they were on October 4, 2026. API names and availability conditions can change in the release version. Availability by region and device, and beta numbers, are not covered here. This post includes no implementation results for the parts written as my view.