Overview

Hey everyone, I'm hitting an absolute brick wall with Sign in with Apple on an Expo/React Native project distributed via TestFlight. The native API (isAvailableAsync() via expo-apple-authentication) is returning false globally across multiple physical devices, different user Apple IDs, and different network environments. We have completely ruled out a local code crash—the framework initializes, but the local iOS kernel is explicitly returning a hard false on the availability check before even attempting to present the modal sheet. Here is exactly what we have isolated and verified based on DTS documentation: Identifiers & App ID Setup: Team ID: NJ5JQ25956 (Luxor Interactive LLC) iOS Bundle Identifier: com.myoseph.carswipeapp App ID Capabilities: Checked and verified that "Sign in with Apple" is explicitly enabled on the Apple Developer Portal for this explicit bundle ID. Credentials & Provisioning Purge: We suspecting a cached provisioning profile or identifier cross-talk (our Android layout previously used a different package string, though we just unified it to com.myoseph.carswipeapp across the configuration scripts to rule out cryptographic hash collisions). We completely revoked and purged all old Distribution Certificates and App Store Provisioning Profiles from both the build engine cache and the Apple Developer Portal. Regenerated a completely fresh Distribution Certificate (Serial: 7383AE36E5D21E93F3F0B907D3A87DF3) and a brand-new, active Provisioning Profile (ID: L328ZDW2A4). Entitlements Inspection: Inspecting the compiled .ipa container structure shows the explicit com.apple.developer.applesignin entitlement key array is correctly present and injected. Despite having pristine, newly-minted cryptographic certificates tightly bound to a verified App ID, Build 62 on TestFlight still rejects the authentication framework instantly at the OS level. Is there a known propagation lag on Apple's identity database servers for newly updated Team ID capabilities, or is there an internal backend entitlement lock on our corporate Team ID container that isn't reflecting on the developer portal UI? Any insights or a sanity check from DTS would be an absolute lifesaver. Thanks in advance.

Hi everyone, I’ve spent the last 2 months pulling off a wild solo experiment: building a complete mobile puzzle game from scratch using Unreal Engine, where over 100k lines of code (C++ & pipelines) and thousands of 3D/UI assets were entirely generated by various AI models. Before anyone rolls their eyes—no, AI didn't replace the game designer here. AI provided the brute-force production muscle, but I gave it its soul. Every single bit of core design—the underlying math models, difficulty pacing, and counter-mechanics for the dynamic gimmick bosses—was designed and balanced by me. The game itself is a tactical, gimmick-boss driven match-3. I got tired of the usual candy-coated reskins and wanted to make something with rigid logic, irregular board topologies, and tight move economies for people who actually like strategic puzzle-solving. I’m currently running a closed beta via TestFlight on iOS to stress-test the build, and I'd love to get some brutal feedback from fellow devs. (Just a heads up: save progress won't carry over to final release). You can grab the build directly here: https://testflight.apple.com/join/VnUWvHFC I’m specifically looking for insights on: Performance & Thermals: How does the UE rendering pipeline handle on your device during heavy chain reactions? Aesthetic Cohesion: Does an all-AI asset pipeline manage to feel like a unified universe, or do you feel that "AI uncanny valley"? Balance: Do the non-Euclidean layouts and boss disruption mechanics feel strategic or just plain frustrating? I'll be hanging out in the comments. Drop any questions you have about prompt-engineering a 100k-line codebase, solo pipeline management, or UE integration. Thanks for testing!

I'm building a multi-box PIN/OTP input in SwiftUI. In UIKit, I used UITextFieldDelegate to detect backspace presses on an empty field to move focus backward. SwiftUI’s .onChange(of: text) only triggers when text is actually deleted, completely missing backspaces on an already empty field. Is there a pure SwiftUI way to handle this now, or are we still forced to wrap UITextField via UIViewRepresentable?

I have a custom Layout that places a number of labels for a cell footer in a certain way based on the available width that needs to conditionally hide those views that do not entirely fit anymore (based on some priorities I specify). Currently I simply move the subviews that do not fit anymore off-screen and use clipping to hide them outside the layout, as I did not find an "official" way to hide / exclude a subview from a Layout. Does anyone know a better / nicer way to do this in SwiftUI?

Overview When calling onChange(of:initial:_:) with initial as true, closure called even when the value assigned is the same as previous value (not just the first time, subsequently too). However when initial value is false it is called only when value changes. Questions Is this a bug? Am I missing something?

Overview I have a view B contains @Query for cars, now this @Query predicate depends on an input which is passed from view A. Current approach I am creating @Query in the init of view B by using _cars. Questions Now how can I compose @Query based on input from view A? Is my approach correct? In my approach Query will be created every time init gets called Or is there a better approach?

Overview I have a navigation split view. The detail view contains a model now this model depends on id from the parent view. Questions How can I pass data from the parent view and yet create the view in the detail view? Or should I be pass the model from the parent view, but the problem is the parent view needs to persist model. Or is there a better approach?

Previously, we used NSUserActivity to enable invoking specific in-app features from the Shortcuts app. This worked as expected up to iOS 17, but it no longer works on iOS 18 and later. We’ve confirmed that the same functionality can be achieved by using App Intents as an alternative. Question. Is the change in iOS 18—where NSUserActivity-based Shortcuts integration no longer works (or is restricted)—an intentional behavior change? If so, could you point us to any relevant documentation or release notes describing this change? We’ve tried to find information ourselves but haven’t found any clear references. Request. In cases like this where an OS update impacts existing implementations, it would be very helpful if such changes were clearly documented in release notes or API change logs in advance.

I am encountering a consistent, reproducible crash in our app when presenting a UIDatePicker configured with the calendar style. The crash triggers every single time the picker is invoked and points directly to the modern date picker's internal UICollectionView. The Exception: *** Terminating app due to uncaught exception 'NSInternalInconsistencyException', reason: 'UICollectionView internal inconsistency: attempted to set layout with the collection view requiring a reload' let datePicker = UIDatePicker() datePicker.datePickerMode = .date datePicker.preferredDatePickerStyle = .compact This crash is occuring in inline style too when I try to open the calendar. I tried this in other apps. It works fine. I didn't override any collectionView layouts

Do apps like Files, Safari, and Finder provide on screen data on request? I've got an entity that implements transferable, but Siri is consistently telling me that it is unable to export data from Files, Safari and photos, and does not seem to want to construct my entity. (it's an @PhotoAsset entity, and I've implemented @createAssets as well.) Is that true at the moment that Siri cannot read data from those apps? I'm not sure if this is a bug or a feature.

我今年18岁。没有机器学习背景，没有上过大学，高中都没去上，没有导师。 几天前我盯着一张纸发呆。突然想：为什么计算机神经网络一定要是2D的？可以模拟生物吗？为什么一定要在平面上算？如果多个平面，岂不是翻倍？如果把六张纸想象成一个魔方，六个面各自承载神经元，八条体对角线变成新的通信通道会怎么样？ 我真的很喜欢折腾这些，然后我立刻制定了详细计划，使用AI工具辅助写下了第一个 kernel。跑崩了。我又重新想了一下，和qq群友分享了我的目标，又写。又崩。连续几十次。没有 PyTorch，没有 TensorFlow，没有 CUDA。只有Swift和Metal。因为我的电脑显卡是AMD Vega 64，没装任何框架辅助，因为我想明白最底层的运行方式是什么原理。 这就是CubeNN。 ##以下为AI的详细解答，内容与架构改动太多，我在这里一次讲不清楚 它是什么 一个用魔方几何作为计算架构的神经网络引擎。 标准 Transformer: 把数据排成一行，O(n²) 地互相看 CubeNN: 把数据分布在 14 个面上，只在该看的地方看 6 个标准面 → 块稀疏注意力（粗看全局 + 细看局部） 8 个 X 面对角线 → 跨面信息桥（不做 Attention，只负责传递） 每轮：6 面算 → 投影到 8 X 面 → 上采样精炼 → 融合回 6 面 最关键的是 Cube Cascade——一个树+链级联推理： 树阶段: 1 个魔方 spawn 8 个 → 8 个 spawn 64 个 → 73 个并行探索 GPU 上同时跑，选最优路径 链阶段: 最优叶子无限深度精炼 3-5 步收敛，方差提升 ~7% 怎么实现的 纯 Swift + Metal。零依赖。零框架。 // 大致代码就是这些 import Metal import Foundation let device = MTLCreateSystemDefaultDevice()! let library = try! device.makeLibrary(filepath: "cube_nn.metallib") // ...12 个 GPU kernel，12,000 次 dispatch 关键技术决策： 单 Command Buffer：整个树阶段 73 个魔方的全部 kernel dispatch 打包进一个 CB，0 次 CPU-GPU 同步 Pipeline State 缓存：编码从 1022ms 降到 42ms Buffer 偏移：所有 73 个魔方的 14 个面存进一个连续 buffer，kernel 通过 buffer(15) 传偏移量 FP16：N≥64 时半精度提速 21% 性能 ##经过测试，但是因设备差异可能不准确，仅参考 AMD Radeon RX Vega 64 (2017 年显卡, 14nm, 295W): 规模 神经元 魔方数 耗时 N=32 6,144 73 (树) 435ms N=64 24,576 21 (树) 817ms N=128 98,304 1 116ms N=32 全连接 Attention 每层 201M FLOP → CubeNN 块稀疏 370K FLOP (544× 减少) N=128 全连接需要 32GB 显存（物理上不存在）→ CubeNN 用 192KB N=256 全连接需要 2.2T FLOP → CubeNN 52M FLOP (42,300× 减少) 代码体积：161KB。 对比 PyTorch 的 800MB。 我经历了什么 这个项目最困难的不是写 kernel，是在没有任何人告诉我"能不能做"的情况下，靠反复试错找到路。 第一次试图跑 73 个魔方，GPU 直接 hang 了。花了 3 天定位到是 Command Buffer 堆叠过多。 改了 single encoder 方案，又碰上 SIGILL——Metal 不允许 makeBuffer(length: 0)，B=0 时创建了零长度 buffer。 想用 threadgroup memory 做 kernel fusion，结果跨 threadgroup 读不到数据，才明白 LDS 是 per-group 的。 N=64 的 FP16 要手动写 float↔half 转换函数，因为 macOS 11 上 Float16 类型被标为 unavailable。 每一次崩溃都教会我一个 Metal 的底层细节。没有人教我，但 Metal 的报错信息就是最好的老师。 为什么发在 Apple 开发者论坛 因为这是为苹果生态而生的项目。CubeNN 从头到尾只用了两个东西：Swift 和 Metal。它不需要移植就能跑在任何 Apple Silicon Mac 上（API兼容）。如果未来能把部分 kernel 映射到 Neural Engine，效率会再翻几倍。 我想问 Apple 的 Metal 工程师和 Core ML 团队： ** 有没有更好的 GPU 任务调度方式？**目前表现仍然欠佳（对于我这个完美主义者来说），可能改得有点乱了 有没有兴趣评估这个架构在 M4 上的表现？ 我手里只有 Vega 64。M4 GPU + ANE方法 跑 CubeNN 会是什么效果？ 源代码 ├── run.swift # 统一 CLI，参数化 N/B/depth ├── src/ │ ├── cube_nn.metal # FP16 kernel │ └── cube_nn_fp32.metal # FP32 kernel └── benchmarks/ # 实测数据 如果你读到了这里——谢谢你。一个门外汉靠痴狂的，纯粹到几乎是妄想的主意和Metal走到了这里。我懂的不是很多，如果这个架构有任何价值，我想让它变得更好。任何建议、批评、或者指教，都非常欢迎。

I can't figure out what I'm doing wrong, maybe it's just "in development." I am running on Tahoe, but running the 27 beta. I created a simulator and asked it to run and it did, but claude had to jump through hoops to get it to actually display. The hub only shows my appleTV device that I test on. Simulators do not show at all. Even when running. Hints on what I am doing wrong? Is it because I'm on Tahoe? Known bug? Dumb user?

Can the new Siri correctly handle SiriKit INIntent? My app uses INIntent like INSendMessageIntent. It is used not only for Siri action but also used Communication notifications and CarPlay and more OS features. Is any migration required?

It appears that recording video of a device in Device Hub does not record the associated audio of the device. Are there any plans to do this? (Or a workaround that's easier than rerouting the audio output to an audio editor app and recording it separately?) It would be incredibly useful for making app demonstrations and tutorials.

In visionOS 26 (with Website Environments enabled), there was a two step action to enable the environment: click button on the left of the address bar (1 in image), then click "Open Website Environment" (2 in image). This example was shown again this year in Build next-generation experiences with visionOS 27. But I do no see this working in visionOS 27. Instead, I've needed to add a button for the user to open (and optionally close) the website environment. Is adding a button/control to the web page the new way to activate website environments?

I'm trying to build a Mac Catalyst version of our iOS (iPad) app but the build always fails with 'Internal error, please file a bug at ...' when compiling MainWindow.xib file. (I have already filed a bug.) The error in the log.txt file is: "Failed to compile nib consisting of IBProxyObject (x2), IBUICustomObject (x1), IBUINavigationBar (x1), IBUINavigationController (x1), IBUINavigationItem (x1), IBUIToolbar (x1), IBUIViewController (x1), and IBUIWindow (x1). Exception name: IBPlatformToolFailureException Exception reason: The tool closed the connection (IBAgent-macOS-UIKit)" The same .xib file works fine when building an iOS version of the app and it worked for MAC catalyst until Xcode v26.x. I've already tried cleaning the build folder, deleted project DerivedData folder, replaced all custom classes with base classes and tried other suggestions found in the internet but nothing seems to help.

iCloud KeySyc is a privacy-focused feature that securely syncs a user’s personalized keyboard learning across all Apple devices through iCloud. Today, when users upgrade to a new iPhone or begin using a new iPad or Mac, the keyboard often loses much of the personalization it has developed over time. Users must re-teach autocorrect, rebuild custom vocabulary, and wait for the keyboard to relearn their typing habits. iCloud KeySync would allow users to move seamlessly between devices while keeping the keyboard experience familiar from day one. Problem Statement Apple’s keyboard becomes more useful over time as it learns: Frequently used words and phrases Personal slang and abbreviations Emoji preferences Autocorrect adjustments Writing patterns and typing habits When users switch devices, much of this personalization is lost or takes time to rebuild. This creates friction during device upgrades and reduces the feeling of continuity within the Apple ecosystem. Proposed Solution iCloud KeySync securely synchronizes keyboard personalization data across Apple devices. The feature would: Learn and store personalized keyboard behavior on-device. Encrypt and sync personalization data through iCloud. Restore keyboard preferences automatically when signing into a new device. Keep keyboard learning consistent across iPhone, iPad, and Mac. Key Features Personalized Vocabulary Sync Synchronizes user-added words, nicknames, slang, and frequently used phrases. Emoji Preference Sync Maintains commonly used emoji patterns across devices. Autocorrect Learning Sync Preserves learned corrections and accepted custom spellings. Typing Behavior Continuity Allows devices to share keyboard learning so users receive familiar suggestions regardless of which Apple device they use. User Controls Enable or disable iCloud KeySync. Reset keyboard learning at any time. View privacy information and storage usage. Privacy and Security Privacy should remain the foundation of the feature. All keyboard learning occurs on-device. Personalization data is end-to-end encrypted. Raw message content is never uploaded. Users retain full control over synchronization settings. Security protections match the standards used by iCloud Keychain. User Benefits New devices feel familiar immediately. Less time correcting autocorrect mistakes. Faster onboarding after device upgrades. Consistent typing experience across the Apple ecosystem. Improved customer satisfaction and retention. Example User Scenario A user upgrades from an older iPhone to a new iPhone. After signing into iCloud and restoring their device, the keyboard immediately recognizes their commonly used phrases, preferred emoji, custom words, and learned typing patterns. The new device feels personalized from the first message typed, without requiring weeks of retraining Business Value Improves the upgrade experience. Strengthens ecosystem continuity. Reinforces Apple’s privacy-first approach to AI and machine learning. Creates a subtle but meaningful quality-of-life improvement for millions of users. Switch phones. Keep your vibe.

Are evaluations just for Text-text, or is there an efficient ways to evaluate image-text, like for MobileClip2, or YOLOE? At first blush, it seems like that would just be redoing training, testing, and validation of the dataset, or am I missing something? (newbe)

For an app review, the reviewer rejected my app review submission and posted their comments about why it was rejected. I replied back with my message clarifying some things but I never got any response back. Do I need to resubmit to app review for my message to be checked out? I have been experiencing lots of delays.

With the recent introduction of the unified asset pipeline and Icon Composer, managing cross-platform icon designs has become incredibly efficient. However, there is still a significant platform disparity when it comes to maintaining visual consistency and brand uniformity across paired devices. On iOS, we can dynamically change the app icon at runtime using the setAlternateIconName API. Currently, watchOS completely lacks an equivalent mechanism. If a user selects an alternative icon inside an iOS companion app, the paired Apple Watch app icon remains locked to the default primary asset. This creates a disjointed experience and directly impacts user recognition. The watchOS home screen relies entirely on instant visual cues. If a user changes their iPhone icon to a custom colorway, they instinctively look for that same colorway on their wrist. Leaving the watch icon unchanged makes it harder to quickly locate the app. I have submitted an enhancement request via Feedback Assistant to bring alternate app icon support to watchOS, ideally through an automatic system-level sync within the unified .icon pipeline, or via a native watchOS runtime API. If your app utilizes alternate icons and you would like to see this cross-device continuity brought to watchOS, please consider filing your own duplicate request to help bring visibility to this gap. Filed Feedback: FB23080617