Could you also clarify if Apple Intelligence Foundation Models are supposed to run in a VM? When running macOS Tahoe with Xcode 26 in VirtualBuddy, I get FoundationModels.LanguageModelSession.GenerationError.Context(debugDescription: "Model is unavailable", underlyingErrors: [])..

Actually, now the problem is back without any changes. Might be related to iOS 18.1 update, as I am currently getting a notice in battery settings about update related activity, battery drain, and quite a warm iPhone... I have attached the sample project to the bug report now.

I have changed the Timeline policy from .atEnd to .after(tomorrow) (with tomorrow being start of next day) - while keeping the timeline of 10 entries, and I am no longer seeing this issue. Switching back to .atEnd also worked. Also, I was not successful in creating a sample project to reproduce the issue with .atEnd.

It looks like AVCam has been updated with an actor. Still I found this example very interesting. Would it be possible to use @preconcurrency or would this change the semantics? actor VarnishCentral: @preconcurrency Varnisher.Delegate { func varnisherShouldUseGloss(_ varnisher: Varnisher) -> Bool { shouldUseGloss() } }

Awesome, can't believe I have overlooked this 🙈. Thanks!

Thanks, good to know! You mean Multipeer Connectivity rather than RealityKit's MulitipeerConnectivityService, right? This is still good news, because MC is required by MulitipeerConnectivityService. So apparently you can network across simulators? Nice.