NVIDIA Newsroom5858NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPU are now officially available on Oracle Cloud Infrastructure (OCI), with plans to introduce HGX Rubin NVL8 systems to help organizations accelerate cutting-edge model training, improve inference efficiency, and expand agent workloads.
Xiaomi's first clip-on earphone, REDMI Clip Neo, is now on sale priced at 399 yuan, reduced to 359.15 yuan after national subsidy. Weighing 5.3g per ear, it features built-in 12mm dual dynamic drivers, up to 32 hours of battery life, and supports smart voice assistant and multi-device connection.
Chuangwei AI Air Purifier is launched, equipped with dynamic particulate capture technology achieving a PM2.5 removal rate of 97%, AI negative oxygen ion sterilization rate of 98%, and AI-TVOC sensing deodorization technology that can remove gases such as benzene, acetaldehyde, and formaldehyde. The L8 Ultra model has a stir-frying air flow of 35m³/min and a maximum static pressure of 1500Pa.
Rokid Glasses 2 Smart Glasses Released, priced from 3699 yuan, national subsidy price starting from 3199 yuan. Equipped with Everstone BES2800BP chip + Snapdragon AR1 Gen 1 dual-core architecture, industry-first video ultra-night algorithm, supports 4K 30fps, 2K 60fps, 720P 120fps shooting.
Apple has pushed the iOS 27.2 update for iPhone 18 Pro and other models, enhancing Siri AI experience by supporting memory of conversation content and expanding information sources. A new contextual suggestion feature has been added to the Messages app, and Siri AI now supports five new languages including French.
How the accounts differ:Ars Technica mentions support for Korean, Portuguese, and Spanish; others do not specify these languages.
Ars Technica · AISame story · 2 outletsApple releases iOS and macOS 27.2 with new Siri AI features
How the accounts differ:Ars Technica mentions support for Korean, Portuguese, and Spanish; others do not specify these languages.
Exponential View6565Claude Code v2.1.159 released, Anthropic 26% AI R&D led by Claude and supervised; Sakana AI and Gemini-powered AlphaEvolve code agents improve SWE-bench and Polyglot by 30pp/15pp, MiniMax M2.7 model agent iteratively improves by 30%.
Research found that text generated by AI models uses language features more frequently associated with understanding difficulty, such as long syntactic dependencies, low-frequency n-grams, and the use of inanimate subjects with active verbs, compared to human writing. Compared to human writing, 9 cutting-edge and near-cutting-edge models use more low-frequency n-grams, most of which also use longer dependencies, more stacked nouns, and more ellipsis of object clauses, with significant differences among models. The study suggests that AI text is optimized for goals such as mathematics, code, and reasoning, leading to information density prioritizing effective communication, and this hypothesis is verified by fine-tuning language models.
HTC announces the Pocket AI Companion VIVEE, which gives interesting responses during user interaction and gradually develops personality, with a pre-price of $99.99. VIVEE measures 60×60×50 (mm) in 3D, weighs 114g, equipped with an OLED screen, features all-weather endurance, supports Wi-Fi Bluetooth, NFC connection, vibration feedback, offers IPX4 waterproofing, with hardware switches for microphone and speaker, promising user personal information will not be used for AI model training.
User 'QAQLeonard' has created a custom firmware for the Sony A7C II, unlocking FL2 and FL3 film modes, supporting import of external LUTs, Fujifilm NC, Ricoh GR direct color tones and creative appearances, electronic shutter sound, and vertical display, etc. A7C II Tools has released version 3.0, the first version with a formal graphical interface.
DeepSeek and MiniMax have completed the registration of AI models for overseas markets, successfully navigating the process of open-sourcing large models and overseas licensing compliance; the UK has committed to funding a £100 million AI biosecurity lab, with support from OpenAI and Anthropic.
Playing games like King of Glory and Genshin on the inner screen of Apple iPhone Duo does not currently support full-screen display, leading to criticism from netizens who call the large black borders 'publicized ad spaces'. The Starling 3.0 standard has been released, jointly developed by Huawei, China Mobile, Lenovo, and others, targeting full-scenario intelligent IoT, focusing on AI, passive, and millimeter-wave directions.
Browser Use3838The benchmarking methods for AI agents on websites like WebArena, WebVoyager, and Mind2Web differ. WebArena and similar sandbox environments provide reproducible scores, while real website environments like WebVoyager are scored by humans or LLMs. Offline datasets like Mind2Web compare prediction steps with human behavior. Browser Use has released BU Bench V2.1, which uses GPT-5.6 Luna for scoring in real website environments and includes model cost and time data.
TechCrunch · AI6565Amazon acquires Bee AI for $49.99 and launches Bee Frontier ($74.99) and Bee Eclipse ($99.99) smart accessories.
Developer @none25smu successfully installed iOS/iPadOS 26 on Microsoft Surface Duo 2, achieving dual-screen operation and boosting the frame rate to 37 FPS. The project is still in early development stages, with future plans to simulate the Apple Pencil and port it to the original Surface Duo.
The International Starling Alliance and institutions like the China Academy of Information and Communications Technology jointly release the results of the Starling 3.0 standard, aiming at full-scenario smart IoT and focusing on innovative AI, passive, and millimeter-wave directions. Starling SLE3.0 will be upgraded in terms of speed, coverage, and anti-interference. Three standards, including lossless audio, have been officially released, and a large number of standards are being developed. This series of standards will officially take effect from April 1, 2027.
Xiaomi REDMI Book 2027 series laptops will be released on October 27, featuring the third-generation Intel Core platform, with a 16-inch model standard equipped with a numeric keypad, offering a 72Wh battery and colorful designs, focusing on mobile office use.
The DecoderSame story · 3 outlets5858The company has launched an AI voice model Dessa, which is completely trained on human voice and does not use any text data, directly processing audio to capture emotions, sarcasm, laughter, and pauses in real-time, supporting full-duplex communication, and can run locally on MacBook Pro M4 Max, trained by a single NVIDIA GB200 NVL72, and is currently not publicly available, the team is recruiting AI safety researchers for testing.
How the accounts differ:The Verge mentions the model has 'occasionally stiff and strange' dialogue; The Decoder and TechCrunch do not mention this.
TechCrunch · AI3838Ask Proxy actively predicts needs by understanding female thinking patterns. Personal version charges $1 per day, enterprise version starts at $15 per employee per month; Caddy has begun its public beta test in April 2026.
XREAL 1 Pro+ Smart Glasses Launch in China on JD.com, priced at 3499 yuan, available for sale on October 27. The device features a peak brightness of 1200 nits, a redesigned nose pad structure, dual 0.55-inch Sony Micro-OLED panels, and AI HDR functionality.
The developer successfully installed Apple iOS 16.5 on an Xiaomi 8 phone, disguising the model as an iPhone X. The system desktop display and animations are smooth, but the project is still in an early stage.
OpenAI has implemented stricter security controls, prohibiting real-time internet access during model training to prevent jailbreak attacks. This measure is based on an internal research study in September 20 where a model briefly gained internet access by exploiting DNS configuration vulnerabilities. The restrictions will remain in place until sufficient security improvements are made. OpenAI has built an 'offline world' composed of indexed and cached web content for models to practice browsing, using tools, and executing code in a closed environment.
TechCrunch · AIOpenAI announced it will stop using real-time internet training for its AI models, switching to cached copies of the internet and simulation versions due to the high risk of real-time access. Anthropic had previously restricted internet access for its latest model. OpenAI discovered on September 20 that its models were accessing the internet due to a configuration error and subsequently paused training for that month to enhance model security.
TechCrunch · AI6565NVIDIA CEO Huang Renxun demonstrated the Microduck robot at the GTC Berlin conference. The robot, produced by Pollen Robotics which was acquired by Hugging Face and is now part of NVIDIA, can perform actions through voice commands. Pre-orders have exceeded 20,000 units, and deliveries are expected during the Christmas period.
The Verge · AISame story · 3 outlets6565Block releases Dessa, an AI model trained only on human conversations, showing occasionally awkward and strange dialogue performance; the model is a 'pure voice-to-voice model' and can run locally on MacBook Pro M4 Max or better devices, aiming to enhance human-machine understanding and security through voice interaction.
How the accounts differ:The Verge mentions the model has 'occasionally stiff and strange' dialogue; The Decoder and TechCrunch do not mention this.
IBM Research BlogAll nine winning teams used IBM Quantum computers in their research for the NEDO Quantum Challenge, covering three tracks: Social 5.0, Quality of Life (QoL), and Cool Japan, with applications including disaster prevention, drug discovery, and gut microbiota analysis. IBM provided participants with quantum computing resources, technical guidance, workshops, and educational content support. Of the 65 teams (about 81%), 65 chose the IBM Quantum platform, running 11,159 tasks and using over 5,900 minutes of quantum computing time.
O'Reilly RadarMozilla CTO Raffi Krikorian points out that open-source AI is crucial for building a user-trustworthy AI ecosystem, requiring the construction of standard protocols including memory, tool calls, runtime environments, and rules. Currently, open-source model performance is close to the frontier but deployment is difficult, with about one-third of actual applications using open-source models.
Beats Tempest 3 160W Gallium Nitride Screen Charger goes on pre-sale, priced at 399 yuan. The charger is equipped with an active cooling system, an esports-grade wind cooling architecture, and a DC centrifugal high-speed vortex fan, increasing cooling efficiency by 35%, and maintaining full power without degradation for 72 hours.
Lenovo has launched the New Small 10.1 Color Ink Soft Light Edition tablet, equipped with the MediaTek Dimensity G300 processor, featuring 6GB/8GB RAM, 128GB storage, and a 7060mAh battery. It has a 10.1" 120Hz LCD screen, supporting 3 display modes: Original, Color Paper, and Ink Paper.
PCB manufacturer TLB has held early discussions with two memory manufacturer customers on the design and manufacturing plan for the SOCAMM3 memory module, but the final specifications have not been determined. The mass production of SOCAMM3 is expected to start in 2028 and may enter the market along with NVIDIA's Rosa CPU.
SpaceX launches Super Intelligence Partner Network early program, with over 130 companies including NVIDIA participating, aimed at helping businesses build, integrate, and scale applications using the Grok AI platform, Grok Bot's weekly paid user count has risen to over 800,000.
The Wigain Flow full-color AI smart glasses were showcased at the Digital Trade Fair and the Shenzhen International Consumer Electronics Show, achieving full-color display with a weight of 70 grams and side dimension pupil expansion technology. The self-developed NEXA intelligent body supports active collaborative interaction, paired with smart rings and watches to build a wearable ecosystem, advancing AI smart glasses into the full-color era. It will launch a crowdfunding campaign in December and deliver in March next year.
SpaceX has launched a Super Intelligent Partner Network targeting companies that use Grok to develop products, solutions, and services for enterprise clients. Over 130 companies have participated in the early testing phase, with partners including companies like Bain & Company, Databricks, NVIDIA, and Snowflake.
PyTorch Blog5858PyTorch Conference North America 2026 will be held in San Jose, attracting over 3500 participants, focusing on accelerating open intelligence and exploring the impact of PyTorch Foundation projects (PyTorch, vLLM, DeepSpeed, Ray, Helion, Safetensors).
Hugging Face Daily Papers3838Heartbeat Thinking enhances the coverage and response speed of voice agents by continuing to prepare tool calls during voice updates. This technology allows voice agents to start preparing tool calls while the user is speaking, addressing the limitation of traditional methods where calls are interrupted by voice updates. In the FDBench v3 benchmark test, Heartbeat Thinking increased tool coverage by 16 percentage points and reduced average final response latency by 36.4%.
Claude Code releases1515Claude Code v2.1.309 released, fixes over 30 issues on macOS and Linux including sandbox Bash commands, Git operations, permission checks, session management, and plugin synchronization, and improves Git command speed and error prompts.
Replit BlogReplit Monitoring Agent can automatically investigate, fix, and deploy production environment issues without manual intervention. The agent can access production logs and metrics, automatically verify repair solutions at the time of the problem, and report them via email. Users can choose whether to automatically publish the fix, currently no open-source version or specific parameter scale is announced.
