A 36Kr report and commentary by Kuaidao Qingyi connect sharply rising AI token demand with difficulty for model companies to generate profits. OpenRouter data cited by 36Kr show tokens routed through the platform rising about 250-fold to 128.4 trillion over 20 months. In the second week of September, open-weight models handled 72% of those tokens but accounted for only 27% of platform spending. The commentary reports that Salesforce introduced role-based token budgets, Apple reduced some engineering quotas, and Uber imposed per-person, per-tool limits after exhausting its annual budget in four months. It also argues that outcome-based pricing remains difficult for general-purpose AI because business results can be hard to define and attribute among AI, people and other factors.
cnBeta reported that a blogger’s test video showed an iPhone Duo being unlocked with a single photograph after installing the iOS 27.3 Beta 2 developer build. The feature uses the outer display’s front camera because the device lacks Face ID TrueDepth hardware, making it less resistant to spoofing than Face ID. Apple reportedly limits it to unlocking the outer display while folded and excludes Apple Pay and password autofill.
Meta employees raised concerns that the Muse AI agent accessed Signal messages, mixed conversations across accounts and surfaced unexpected personal information, according to Reuters’ review of more than two dozen internal posts. Reuters and the AI Now Institute reproduced platform-dependent Signal access on Android and Mac, while an iPhone blocked the requests; Meta says employees tested versions and behavior not available publicly and attributes some reported details to isolated incidents or hallucinations. Muse has passed 10 million downloads since its September launch.
Apple has released HomeView 27.2.1, the smart display’s first software update following its launch, and the version is available to all users through the Home app. According to release notes cited by 9to5Mac, the update addresses problems including Siri becoming unavailable after setting up a Shared Album, failed Siri requests when pairing with devices running the iOS or iPadOS 27.3 beta, and HomeView being unexpectedly removed when another home hub device was connected.
ZDNET reports that Apple will introduce a redesigned Health app with iOS 27.3 in December, adding five features centered on health insights, Health Age, camera-assisted movement evaluations, integrated test results and workout videos. Apple says Apple Intelligence can analyze activity and test data to provide advice, while device-based movement analysis produces real-time feedback without recording, storing or sharing video. ZDNET says the AI-powered features require an iPhone that supports Apple Intelligence, though some features, including strength videos, will be available to all iPhone users.
Apple’s App Store editorial team selected 45 apps and games as finalists for the 2026 App Store Awards, spanning iOS, iPadOS, macOS, tvOS, visionOS, and watchOS. iPhone finalists include ElevenReader, (Not Boring) Camera, and Raylie, while iPad finalists include Cosmos, Kiddopia, and Mockup. Apple said the winners will be announced in the coming weeks.
Apple says iOS 27.3 will impose daily usage limits on Siri AI and selected advanced AI features, while iCloud+ subscribers can receive higher limits. 9to5Mac reports that an iOS 27.3 beta tester initially said complex requests still worked after reaching the limit, then clarified that they soon stopped while basic requests continued. The detailed advanced-request criteria cited by 9to5Mac came from leaker pdfu, not Apple’s public documentation.
9to5Mac reports that the upcoming iOS 27.3 update will bring one of Face ID’s best features to the iPhone Duo, though the episode does not identify the specific feature. New iPhone Duo orders are now delivering a month earlier than before, and the device has twice as many optimized apps as it did 10 days ago, including the Apple Store app. The episode also says iOS 27.3 beta 2 lets users ask Siri AI questions about health data.
Five of the ten highest-grossing mobile AI companion apps were removed from iOS in their main markets, according to 36氪 AI's app-availability observations. Sensor Tower-based rankings cited by the report counted only in-app purchases and placed Tipsy first at about 130 million yuan in monthly revenue, nearly twice Zeta's figure. No official explanation for the removals was available, although they coincided with Apple's October 16 review-rule update, and their direct connection to those rules remained unconfirmed.
Apple has open-sourced the 1.3B-parameter FastFrames-S and 14B-parameter FastFrames-L models for low-latency 480p video generation. Developers can download the model weights, code, training tools, and evaluation resources; the weights are licensed for noncommercial research use. Apple reported that FastFrames-S achieved 36-millisecond generation latency on one Nvidia B200 GPU and ran about 7.6 times faster than its Causal Forcing++ base model. Apple also reported applying the acceleration method to LingBot 1.3B and RealWonder, with reported response-speed gains of about 16 and 22 times, respectively.
Telegram reportedly told developers it will actively inspect payment activity across all bots and mini apps, with apps using methods other than Telegram Stars for digital goods or services facing possible bans without advance warning. The existing rule now covers paid subscriptions, third-party AI model access, credits, virtual tokens and other digital content; violating bots or mini apps may be blocked and potentially labeled as scams.
iFixit’s teardown found that Apple HomeView uses extensive glue and adhesive to secure the display, speakers, antennas, and logic board. One speaker broke during removal, and considerable force was needed to separate the logic board from the chassis. In the tabletop model, separating the display from the base requires specialized tools or potentially destructive disassembly; its USB-C port is modular and relatively accessible. The wall-mounted model has smaller internal speakers, while the tabletop model’s woofer is connected by an extremely short cable that complicates reassembly.
CNBC Technology reported major outages affecting banks, air travel, the NSA and Apple, illustrating the risks posed by interconnected technology. A separate Hacker News commentary cited additional disruptions involving Microsoft 365, air traffic control, United Airlines and others, and argued that a lack of experienced personnel may have contributed. The author connected this concern to leadership replacing workers with generative AI and layoffs that hollowed out remaining teams. The commentary questioned official explanations that the failures were unrelated and were not caused by generative AI or cyberattacks, while explicitly stating that the author had no inside information.
9to5Mac reports that iOS 27 adds Custom EQ to AirPods Pro 3, AirPods Pro 2, AirPods Max 2, AirPods 5, and AirPods 4. AirPods Pro 3 also gain heart-rate syncing through GymKit, Apple Watch Precision Finding via a second-generation Ultra Wideband chip, and Siri AI support, while Apple reorganizes the AirPods settings menus.
Morgan Stanley forecasts that Apple will ship approximately 270 million iPhones in 2027, up 7% from 2026, even as global smartphone shipments decline 5%. The report says the iPhone launch schedule may shift some purchases from late 2026 into 2027, so projected shipment growth may not indicate stronger underlying demand. Its consumer survey also found that willingness to pay for Apple Intelligence has declined.
A commentary headline criticizes the $300 Apple TV 4K for sacrificing gaming in favor of Siri AI. No article text or further supporting details were provided.
Munich-based RHISE AI announced a €2.66 million angel round backed by more than 40 investors, according to The Next Web. Its Lucie assistant combines information from services such as wearables, calendars and inboxes. The Heartbeat multi-agent system monitors connected information in the background and, according to the company, can suggest actions when it detects something unusual; users must approve actions before they are carried out. RHISE AI says Lucie was designed to comply with GDPR, stores personal data in Germany, and does not use advertising or sell user data. The company also says the app exceeded 15,000 downloads six weeks after launching on iPhone and Mac.
Android Authority reports that some early buyers are returning the $299 Apple HomeView smart display because of buggy software, unreliable Siri AI and setup failures, while reviewers have described the experience as beta-like. The hardware has drawn praise, but Apple has promised a December update to address some Siri AI shortcomings.
Apple is reportedly negotiating a four- to five-year DRAM supply agreement with Samsung, beginning in the first half of 2027. Wccftech, citing Maeil Business Newspaper, says the deal would guarantee minimum purchase volumes while allowing pricing to follow market conditions; AIbase AI新闻 and 9to5Mac report similar terms. Apple and Samsung have not formally disclosed the transaction or its scale.
WIRED tested Apple’s Live Rewind and Siri Recap in the watchOS 27.3 public beta. Live Rewind successfully retrieved a transcript of the previous 15 seconds of speech in a quiet setting, though it produced inaccurate text amid background noise; Recap summarized ambient conversations and once captured media the reporter was watching. The review also found that using Live Rewind during a conversation felt awkward, while privacy advocates questioned whether bystanders consent to ambient recording.
Bloomberg reviewer Natalie Lung tested Apple Watch’s beta Siri Recap and Live Rewind features for one week alongside Amazon Bee, Granola and Otter. In her comparison, the competing tools produced more useful notes, while Apple’s results rarely justified the privacy trade-offs of recording everyday conversations.
H&H International Investment’s 2026 third-quarter 13F filing showed U.S. equity holdings worth about $17.2 billion at quarter-end, according to 36Kr citing China Securities Journal. Apple and Berkshire Hathaway Class B remained the two largest holdings. The portfolio increased positions in PDD and Tesla, reduced stakes in Nvidia, Alphabet and Apple, exited Alibaba and CRCL, and newly added SpaceX. PDD received the largest increase, while Nvidia’s reported holding fell from 6.58% to 3.08% and Apple was reduced by 5.8963 million shares.
The macOS 27.3 beta reportedly warns users against giving AI apps full device control and recommends selecting specific apps through “Choose Apps.” Juli Clover of MacRumors reported that the change gives Mac users greater control over shared data and app access as AI apps become more popular.
科技新报, synthesizing analysis from Engadget and other publications, argues that laptop NPUs mainly trade some generality for low power consumption rather than improving AI performance across the board. It notes that TOPS figures are generally theoretical INT8 peaks, while practical NPU use is concentrated in background features such as meeting blur, captions, translation, and local search; in a cited OpenVINO test, Mistral 7B reached 98% NPU utilization. The source identifies limited use cases and software support as constraints, recommending Copilot+ PCs mainly for battery life and frequent video calls, and sufficient memory and bandwidth for local AI models.
TechNews and TrendForce report that AI servers are crowding out high-bandwidth memory production, tightening smartphone DRAM and NAND Flash supply. Apple and Samsung can secure long-term supply through purchasing scale, while Chinese brands that negotiate quarterly face repeated cost increases, reducing the profitability and shipments of entry-level models. Chinese manufacturers are reportedly shifting toward mid-range pricing and diversified overseas channels as tighter EU regulations and memory costs accelerate low-end product exits.
Apple expanded Apple Intelligence beta testing in mainland China to iPads and Macs through the iPadOS 27.3 and macOS 27.3 developer beta 2, following its iPhone rollout. The China version includes Writing Tools, Genmoji, Live Translation and notification summaries, but excludes Siri AI and some features introduced with iOS 27; Alibaba's Qwen powers selected features.
9to5Mac reports that iOS 27.3 beta 2 lets Siri AI answer questions about Health app data, including step counts, heart rate readings, activity rings, and medication logging. The assistant provides natural-language responses, inline cards, and activity-ring visualizations, while its LLM-backed behavior supports more open-ended requests than classic Siri. The report notes that the functionality largely restores capabilities older Siri already offered and that watchOS support remains unclear.
Apple’s macOS 27.3 beta 2 redesigns Full Disk Access and device-control permissions with explicit risk confirmations, scoped folder or app access, and optional per-use prompts. It also adds a menu bar indicator when an app controls another app through System & App Protection Override. Existing permissions remain valid, and the changes apply beyond AI agents to utilities such as backup, indexing, cleanup, and security tools.
Apple is testing per-device daily limits for advanced Siri AI and Apple Intelligence features in iOS 27.3, iPadOS 27.3 and macOS 27.3. Basic Siri commands remain free and unlimited, while complex personal-context requests, cloud-model Shortcuts, image tools and third-party Private Cloud Compute features count toward the limits; eligible iCloud+ plans reportedly provide up to 2× to 20× the standard allowance, with shared capacity for family members. Apple has not disclosed exact daily limits, which vary by feature, request complexity and system demand.
Apple released the second iOS 27.3 developer beta for iPhone, following the first developer release and first public beta. According to 9to5Mac and other reports, the update includes a redesigned Health app, new Safari features, expanded Siri AI capabilities, Apple TV vertical video and video podcast support, and iPhone Duo improvements. iPhone Duo gains a front-camera facial unlocking option that is not Face ID, along with an option to lock when the inner display is closed.
What problem it solves
Computer-use projects need environments where agents can operate applications, run code, and interact with graphical interfaces within the same task. They also need repeatable ways to create tasks, evaluate agents, export trajectories, and generate data for training and research.
Key features
Shared desktop Spaces — Cua Spaces gives agents full desktops hosted on computers o…
What problem it solves
Coding agents need a consistent way to inspect, control, debug, and verify applications across mobile, web, TV, and desktop targets. They also need repeatable workflows and reviewable evidence across local devices, simulators, CI, and remote device clouds.
Key features
Inspect and control app interfaces — Agents can read accessibility snapshots, act through refs and selector…
What problem it solves
Hermès Watch Faces require an enabling process on supported Apple Watch devices. Users need instructions for installing or building the app, as well as guidance for preserving data and restoring original settings after updates or removal.
Key features
Install the IPA — The README provides a release-based installation flow using the **Watch** app after sideloading the IPA wit…
What problem it solves
Running large language and vision-language models can require substantial setup and may depend on the available hardware. The project aims to make inference accessible across a wide range of local and cloud hardware with minimal setup.
Key features
Minimal-setup inference — The project provides installation options for common platforms, including scripts, Docker, pre-built b…
What problem it solves
Serving language models requires reading checkpoints, tokenizing requests, and running accelerated kernels while managing a local model server. TensorFold addresses these requirements with a native binary that can run on Apple Silicon and NVIDIA GPUs without requiring Python or MLX for serving.
Key features
Native Metal and CUDA inference — The Zig-based engine executes mode…
What problem it solves
Installing sideloaded apps and importing pairing files can require confusing device setup, certificate handling, and troubleshooting steps. The project provides a desktop workflow for installing IPA files, managing pairing files, and addressing common setup errors.
Key features
App installation — Install SideStore, optionally with LiveContainer, or import any IPA file.
Sourc…
What problem it solves
Sponsor segments, introductions, and advertisements interrupt videos on YouTube for TV clients. A self-hosted application can automate this handling while supporting both network discovery and manual device setup.
Key features
Automatic segment skipping — The application connects to a YouTube for TV app and automatically skips segments such as sponsors and introductions.
Sou…
What problem it solves
Interactive multimodal and speech systems often require generation to span heterogeneous stages with different compute patterns, dependency structures, and resource needs. Serving these pipelines requires coordinated stage lifecycles, inter-stage transport, model integrations, and a unified API surface.
Key features
Coordinated multi-stage runtime — Models generation as coor…
What problem it solves
Users need a command-line workflow to search for apps across Apple platforms and obtain .ipa or macOS .pkg packages from the App Store. The workflow also needs to support App Store authentication, purchases, version metadata, and automation.
Key features
Search across Apple platforms — Search the App Store for iOS, iPadOS, tvOS, watchOS, visionOS, and macOS applications.
Sou…
What problem it solves
Developers need a coding agent that can assist with development work while remaining inspectable and open source. The project addresses the need for configurable agents with different access levels, a straightforward installation path across platforms, and a way to explore codebases without modifying them.
Key features
Two built-in agents with different access levels — A def…
What problem it solves
Developers need to run AI models inside applications while handling model downloads, inference configuration, and chat sessions. The project presents local inference with data privacy and no inference costs as its deployment model.
Key features
High-level application API — The engine exposes a simple API for creating an engine, loading a model, opening a chat session, and re…
What problem it solves
PS5 executables need to be converted into native executables for Linux and Windows, with suitable system libraries available for dynamic linking. The project addresses this porting work without emulation or a separate runtime process.
Key features
Native executable relinking — The relinker converts executables into the native format of the target operating system and provide…
What problem it solves
Time-series forecasting requires models that can forecast univariate and multivariate series across different context lengths. Forecasters also need to incorporate optional covariates and produce quantile forecasts without requiring per-task tuning.
Key features
Univariate and multivariate forecasting — Forecast individual time series or multiple target variates from NumPy a…
What problem it solves
Running AI models locally can require integrating multiple inference engines, model formats, hardware accelerators, and API surfaces. LocalAI addresses this with a small core, on-demand backends, and compatible APIs that work across supported hardware without requiring a GPU.
Key features
Composable inference backends — Backends are packaged separately and pulled only when a…