The SpaceX SI team completely rewrote Grok reinforcement learning and reasoning software from scratch, primarily developed in C with a small amount of C++, replacing JAX and SGLang. This underlying engineering accumulation is the core foundation of Grok's new AI stack, reusing Starlink's engineering expertise to support Grok's iteration.
Walt Disney collaborates with Field AI to test AI parking robots, helping visitors find parking spots in the park; the robots draw design inspiration from 'Cars,' equipped with screens and mechanical arms to provide guidance, but Disney has not announced deployment time.
The Honor 700 series is previewing the back screen arbitrary stick function, supporting expressions, GIFs, etc., which can be pressed and dragged onto the back screen, and also supports rear selfie, AI interaction, heart rate and blood oxygen detection. The series will be equipped with a back screen in full, with a body thickness of 7.96mm, internal code Flora, aims to inherit the flagship imaging capabilities and explore new AI play methods, the spokesperson is temporarily set as Ju Jingyi.
Claude Code exposed database password, API Token and other sensitive credentials to GitHub repository due to improper authorization handling, identifying 2539 real credentials, affecting 63 organizations. Security company Apiiro's research found that over 100,000 local configuration files were uploaded, of which about 1,100 contained confidential data, and the team is studying preventive measures.
KFC Japan used generative AI to create promotional images for some trial products, with ice cubes in the cups showing the same deep green as the soda and a jelly-like sheen, prompting netizens to question whether the images matched the actual products and whether the brand image was damaged. KFC admitted that the promotional images used AI but refused to respond to whether it would be applied to other products.
Huawei announces a launch event for the new storage system trial on November 12, marking the technology's official entry into the real-world joint verification stage. With the advent of the Agentic AI era, AI corpora and other hot and cold data are experiencing explosive growth. Huawei's new storage architecture aims to break through the traditional hot and cold storage bottleneck, providing a new path for the long-term preservation of massive data.
RayNeo upgrades its RayNeo iO AI Smart Glasses with Gaode Navigation feature, allowing users to wake up RayNeo AI via voice and select walking or cycling mode, displaying navigation information in HUD form in real-time in the field of vision, supporting automatic route recalculation on deviation and intelligent pause and resume in weak network environments.
Samsung Electronics plans to complete the development of advanced packaging technology '3D Cube-H' by early next year. This technology stacks multiple chips vertically through hybrid bonding, with terminal spacing of about 6 μm, much smaller than the 25 μm of traditional TC bonding, aiming to increase connection density, reduce energy consumption, and optimize heat dissipation, primarily targeting ultra-high-performance AI chips.
The GAC Aion Ray 7 will be launched on November 12, equipped with Huawei's 192-line LiDAR, and powered by the Weiyuan Intelligent L4-level Robotaxi-derived large model. The car is nearly 5 meters long, with a drag coefficient of 0.22Cd, and a CLTC electricity consumption of 9.7 degrees per 100 kilometers, offering a CLTC pure electric range of 700km.
Microsoft has announced a new patent that allows users to contribute idle GPU computing power to participate in AI training and get paid. The system breaks down training tasks into micro-tasks and distributes them to idle devices for execution, with users able to earn points to exchange for free software, cash, etc. Microsoft plans to open this solution to third-party organizations to alleviate the issue of AI training computing power shortages, but users need to actively authorize and privacy protection issues remain to be resolved.
Qwen Office 2.0 introduces the industry's first 'Collaboration' feature supporting multi-person and Agent collaboration, editorial team tests show that by establishing collaboration spaces linking DingTalk group chats and materials, cross-group information organization, task adoption, and automatic reminders are achieved, transforming scattered discussions into work arrangements, with significant efficiency improvement
Google has introduced memory summary features for Gemini, allowing users to view, edit, and delete personal information remembered by AI. The new feature organizes personal information by category, including background, relationships, work, and interests, and supports deletion and modification. The access point is located at 'Settings' - 'Personal AI' - 'Memory Summary'. It is currently mainly targeted at American Google AI subscribers aged 18 and above, requiring English, Keep activity logging, and memory features to be enabled.
Apple introduced the Skimmer module in the first developer preview version Beta of iOS/macOS 27.3, with about 2% of English requests routed to local model processing, mainly covering high-frequency commands such as timers and playing music, to reduce latency, save computing power, and protect privacy. Skimmer matches user requests with predefined tool workflows using regular expressions, and after a rule is hit, it can directly execute the operation. The relevant rule file is Enigma..
The all-new Shenlan G318 All-Wheel Drive Free Ride Qiankun version will go on sale on November 11, equipped with Huawei Qiankun ADS 5 Pro and Hongmeng Cockpit Harmony Space 5. The car is equipped with 27 high-perception sensors, including 1 cabin LiDAR, supporting natural language interaction with mixed 10 dialects and fuzzy voice command route planning.
Google Gemini violates its 2023 commitment by routing personal Drive files and emails to human annotators. Tests show that with Connected Apps enabled, Gemini autonomously grabs personal topic content (such as schedules, bills, names, reservations) and injects it into session context, including complete unredacted files or emails, without the need for @Drive or @Gmail tags to trigger the process.
Mechanical Revolution announces the release date for the Master Road X10 product launch on November 19, featuring an AI SSD developed in-house that can offload ultra-large local models from DRAM to NAND, while also saving and reusing processed context to optimize model loading methods, reducing redundant computation and loading overhead
Claude Code releases1515The Claude Code v2.1.322 version has added a fast mode in the global, US, and European regions on the Vertex platform. If Vertex rejects this mode, it will fall back to standard speed.
Nubia NaviX Ultra Star Collection Version priced at 6999 yuan starts pre-sale, offering a 16GB+512GB version, the phone is claimed to be the 'world's first AI smart body phone', equipped with the 5th generation Snapdragon 8 Ultra platform and Seed full-duplex voice large model.
OpenAI Deploys 'Control Room' System to Stop High-Risk AI Tasks in 35 Seconds. The system determines task output via a central service, proactively preventing suspicious behavior rather than just flagging. Researchers need to request a task ID, and the system authorizes reasoning after confirmation; the stop command has a maximum delay of 35 seconds, with the company working to reduce it to 0.
vLLM releases1515vLLM v0.34.0rc0 released, migrating Gemma and StableLM to the Transformers backend.
Tencent's 'Eggy Dimension' AI game companion agent has launched its internal test today, supporting real-time voice interaction and game screen information exchange; for PC players, it offers anthropomorphic AI companions with selectable characters like 'Xi Xia' and 'Shen Luba', skilled in games such as 'Valorant' and 'League of Legends'.
AWS Machine Learning BlogAWS Quick processes natural language requests through agents, which call skills and use connectors to operate data sources to complete tasks. Quick comes with default agents, and users can configure custom agents to limit their permissions and functions. System skills support basic operations such as document generation and web browsing. Users can build and share skills to meet team needs.
LangChain Deep Agents releases1515LangChain Deep Agents release deepagents-code==0.2.6 version, including new features such as making idle compression settings traceable, adding trigger tags for uninstall tracking, and fixing bugs like dcode startup under Windows and local server traffic bypassing the proxy.
Apple is introducing a paid 'Smart Usage Pack' for some Apple Intelligence features. The first product to adopt this is Logic Pro, priced at $49.99, valid for 12 months, with unused additional amounts that can be carried over monthly. This usage pack does not support 'Family Sharing' and is only usable with the iCloud Apple account used at the time of purchase.
LangChain Deep Agents releases1515LangChain Deep Agents 0.7.35 released, fixing bugs such as preserving the parent agent's prompt cache prefix for child agents, retaining symbolic link names in virtual file system results, marking `read_file` media and Nemotron nudge as harness notifications, etc.
Codex releases00GitHub Copilot rust-v0.169.0-alpha.8 released, this version includes the latest feature improvements and fixes.
Ollama releases1515Ripple Preview is a 65B dense model designed for smart coding. Recursive Intelligence released version 0.40.16, while Qwen3.8 Flash Next reduced memory usage by about 27 GiB on Apple Silicon chips. Claude always routes to the selected model, and /v1/responses support truncation, no longer defaulting to silent truncation.
Google announces the full opening of Pathways on Google Cloud Platform, the large-scale distributed machine learning runtime system that was used to train models such as Gemini and GPT, now open to Google Cloud customers.
Meta updates iOS version of Muse app with Apple Watch support, allowing users to use their personal AI assistant on their wrists to complete tasks, supporting typing, voice input, voice calls, and watch face interaction. Muse is Meta's personal AI assistant, integrated into Meta apps such as Instagram and Facebook, with its snowman image becoming the core visual element of the Apple Watch app.
OpenAI NewsOpenAI's Control Room can track model workloads in research infrastructure and quickly stop them in the event of security or safety issues.
OpenAI NewsCodeRabbit uses OpenAI models to govern and trust AI-generated code, combining Codex to increase the software delivery volume of engineering teams by 20 times.
OpenAI NewsOpenAI Control Room can track model workloads in research facilities and quickly stop them in case of security or safety issues
The Verge · AISame story · 2 outlets6565Apple offers a one-time purchase option of $49.99 to extend the use of AI features in Logic Pro, which allows users to continue using Apple Intelligence features limited by reliance on OpenAI or Google Gemini cloud AI models.
Codex releases00GitHub Copilot has released version 0.169.0-alpha.7, offering code generation, review, security, and collaboration features, supporting custom integrations and continuous updates.
AWS Machine Learning Blog5858NVIDIA Nemotron 3.5 Super VL is now available on Amazon SageMaker JumpStart, supporting text generation from image, video, and text inputs.
AWS Machine Learning BlogAnthropic Launches Claude Fable 5.2 and Claude Mythos 5.2 on AWS, Fable 5.2 is its most powerful general model for coding and scientific fields, supporting unsupervised operation for days; Mythos 5.2 is provided through a validation program and performs more reliably in cybersecurity and life science fields.
AWS Machine Learning BlogAmazon Bedrock Managed Knowledge Base now supports single-retrieval enhanced generation (RAG) and retrieval-only requests, implemented through AgenticRetrieveStream.
