At WWDC25 we launched a new type of Lab event for the developer community - Group Labs. A Group Lab is a panel Q&A designed for a large audience of developers. Group Labs are a unique opportunity for the community to submit questions directly to a panel of Apple engineers and designers. Here are the highlights from the WWDC25 Group Lab for Developer Tools.
Will my project codebase be used for training when I use Xcode's intelligent assistant powered by cloud-based models?
When using ChatGPT without logging in, your data will not be used to improve any models. If you log in to a ChatGPT account, this is based on your ChatGPT account settings, which allows you to opt-out (it defaults to on). When using Xcode with accounts for other model providers, you should check with the policies of your provider. And finally, at no point will any portion of your codebase be used to train or improve any Apple models.
We'd love to make our SwiftUI Previews (and soon, Playgrounds) as snappy as possible. Is there any way to skip certain build steps, such as running linters? It seems the build environment is exactly the same (compared to a debug build), but maybe there's a trick.
Starting with Xcode 16, SwiftUI previews use the exact same build artifacts as the regular build. The new Playgrounds support in Xcode 26 uses these build artifacts too. Shell script build phases are the most common thing that introduces extra build time, so as a first step, try turning off all shell script build phases (like linters) to get an idea if that’s the issue. If those build phases add significant time to your build, consider moving some of those phases into asynchronous steps, such as running linters before committing instead of on every build. If you do need a shell script build phase to run during your build, make sure to explicitly define the input and output files, as that is a huge way to improve your build performance.
Are we able to provide additional context for the models, like coding standards? Documentation for third party dependencies? Documentation on your own codebase that explains things like architecture and more?
In general, Xcode will automatically search for the right context based on the question and the evolving answer, as the model can interact multiple times with your project as it develops an answer. This will automatically pick up the coding style of the code it sees, and can include files that contain architecture comments, etc. Beyond automatic context, you can manually attach other documents, even if they aren't in your project. For example, you could make a file with rules and ideas and attach it, and it will influence the response. We are very aware of other kinds of automatic context like rule files, etc, though Xcode does not support these at this time.
Once ChatGPT is enabled for Coding Intelligence in Xcode 26, and I sign into my existing ChatGPT account, will the ChatGPT Coding Intelligence model in Xcode know about chat conversations on Xcode development done previously in the ChatGPT Mac app?
Xcode does not use information from other conversations, and conversations started in Xcode are not accessible in the web UI or ChatGPT app.
Is there a plan to make SwiftUI views easier to locate and understand in the view hierarchy like UIKit views?
SwiftUI uses a declarative paradigm to define your user interface. That allows you to specify what you want, with the system translating that into an efficient representation at runtime. Unlike traditional AppKit and UIKit, seeing the runtime representation of SwiftUI views isn't sufficient in order to understand why it's not doing what you want. This year, we introduced a SwiftUI Instrument that shows why things are happening, like view re-rendering.
Is it possible to use the AI chat with ChatGPT Enterprise? My company doesn't allow us to use the general ChatGPT, only the enterprise version they have setup that prevents data from being leaked
Yes, Xcode 26 supports logging into any existing ChatGPT account, including enterprise accounts. If that does not meet your needs, you can also setup a local server that implements the popular chat completions REST API to talk to your enterprise account how you need.
Now that Icon Composer is here, how does it complement or replace existing vector design tools such as Sketch for icon design?
Icon Composer complements your existing vector design tools. You should continue to create your shapes, gradients, and layers in another tool like Sketch, and compose the exported SVG layers in Icon Composer. Once you bring your layers into Icon Composer, you can then use it to influence the translucency, blur, and specular highlights for your icon.
What’s one feature or improvement in the new Xcode that you personally think developers will love, but might not immediately discover? Maybe something tucked away or quietly powerful that’s flown under the radar so far?
One feature we're particularly excited about is the new power profiler for iOS, which gives you further insights into the energy consumption of your app beyond what was possible with the energy instrument previously. You can learn more about how to use this instrument and how it can help you greatly reduce your apps battery usage in the documentation, as well as the session Profile and optimize power usage in your app. There were also improvements in accessibility this year with Voice Control, where you can naturally speak your Swift code to Xcode, and it understands the Swift syntax as you speak. To see it in action, take a look at the demonstration in What’s new in Xcode 26.
We have a software advisory council that is very sensitive to having our private information going to the cloud in any form. What information do you have to help me guide Xcode and Apple Intelligence through the acceptance process?
One thing you can do is configure a proxy for your enterprise that implementing the popular Chat Completions API endpoint protocol. When using a model provider via URL, you can use your proxy endpoint to inspect the network traffic for anything that you do not want sent outside of your enterprise, and then forward the traffic through the proxy to your chosen model provider.
Are there list of recommended LLMs to use with Xcode via Intelligence/Local? I've tried Gemma3-12B, but.. I hope there are better options?
Apple doesn't have a published list of recommended local models. This is a fast-moving space, and so a recommendation would become out of date very quickly as new models are released. We encourage you to try out the local model support in Xcode 26 with models that you find meet your needs, and let us and the community know!
(continued below)
