NAll news

All news

Tuesday792 items

  1. SiliconANGLEModel releases
    AI score58

    Cogent Security releases VR-1.1 vulnerability-fixing model, scoring 57.9% on LiveCVEBench, surpassing…

    Cogent Security has released VR-1.1, a frontier reasoning model designed to write fix patches for newly disclosed software vulnerabilities. On the 190 tasks in LiveCVEBench, VR-1.1 produced usable fixes for 57.9% of tasks, while Claude Code and Codex both scored 45.3%. The benchmark was built by Harbin Institute of Technology and Kuaishou, covers 14 programming languages, and requires patches to pass both security tests and functional tests; Cogent cut off external network access during fixing and hid the scoring tests until each agent had finished.

    Same story · 2 reports

    #Model releases#Safety / alignment#English text is being prepared.

  2. SiliconANGLE
    AI score38

    Novee Launches External Surface Testing to Verify Whether Exposed Assets Can Actually Be Exploited Using AI Penetration Testing

    Novee Cyber Security has added External Surface Testing to its platform, extending AI-driven attack testing to cover all of a customer's internet-exposed surface and producing a short, evidence-verified list of prioritized issues. The feature runs on the same platform as Application Pentesting, and discovered hosts can be directly escalated to gray-box or white-box penetration testing; testing scope requires prior customer approval and does not use destructive payloads. Novee was founded in May 2025 and launched in January with $51.5 million in funding; the new feature is now available.

    #Product updates#Safety / alignment#Agents

  3. SiliconANGLE
    AI score39

    groundcover Turns Wand Acquisition Into Kubernetes Resource Auto-Tuning Beta

    The observability company groundcover has launched the beta of Autonomous Infrastructure, which automatically adjusts the CPU and memory requests of Kubernetes workloads based on actual behavior in production environments; the technology comes from Wand Cloud, which it acquired in September. The feature reuses production historical data already collected by groundcover sensors, so customers can view comparisons between requested values and actual usage as well as estimated monthly savings without additional installation. During the beta, only requests are adjusted, not limits, and upper and lower bounds and audit logs are in place. The beta is available for free to all groundcover customers running its sensors on Kubernetes.

    #Product updates#English text is being prepared.

  4. SiliconANGLE
    AI score48

    Snowflake Expands Agentic AI Platform, Launches Business Ontology and Cortex AI…

    At Expedition 2026, Snowflake announced the Business Ontology framework, expanded governance capabilities for Cortex AI Gateway, and autonomous data engineering capabilities, enabling AI agents to gain business context and governance within a unified environment. Business Ontology is currently in private preview, while Cortex AI Gateway adds agent abuse detection, observability and tracing, Cortex sandbox, AI cost governance, and dynamic model routing. Cortex Training enters public preview, enabling customers to post-train open-weight models with proprietary data without managing underlying GPU infrastructure.

    Same story · 2 reports

    #Product updates#Agents#English text is being prepared.

  5. IEEE Spectrum
    AI score41

    How Kenyan startup Fastagger is redesigning AI for small businesses

    Kenyan startup Fastagger has built an AI model that can run offline on low-end phones. Its business intelligence assistant Auni was launched in 2024 and integrated into the business app of M-Pesa, Kenya's largest mobile payment service; about 6,000 businesses have registered so far. Auni helps barbershops, restaurants and small manufacturers analyze transaction data and offer marketing suggestions. Some functions, such as reading sales records, data analysis and visualization, can be completed offline, while more complex analysis requires an internet connection. Fastagger still has only three co-founders and is adding to Auni an agent that can automatically execute marketing.

    #Industry news#English text is being prepared.#Agents

  6. The Guardian Technology
    AI score49

    Britain's most polluting Drax power plant plans to build a 1GW AI data center, with investment exceeding £20 billion

    Britain's most polluting Drax biomass power plant plans to build the country's largest AI data center at its site, with final capacity exceeding 1GW and investment potentially exceeding £20 billion; the first phase of 100MW of computing power is expected to come into operation in the first quarter of 2028. Drax will first provide land and existing grid connections, and hopes to achieve direct power supply in the early 2030s. After the announcement, Drax's market value rose nearly 9% to nearly £3 billion.

    Same story · 2 reports

    #Industry news#English text is being prepared.

  7. 9to5Mac
    AI score15

    Apple CEO John Ternus's first earnings call criticized for "reading from a script"

    Apple's new CEO John Ternus's first earnings call was criticized by Bloomberg as a case of "stage fright," saying he hid behind a carefully written script when answering analysts' questions and missed the chance to lay out the "Ternus roadmap." Fortune also noted that Ternus handed some questions to CFO Kevan Parekh and COO Sabih Khan, who will become regular participants in earnings calls going forward. Apple warned this quarter that memory costs continue to rise, and that it will increase RAM spending to build Siri AI data centers.

    #Industry news

  8. Numerama
    AI score34

    OpenAI offers French students €50 in ChatGPT credits

    OpenAI is distributing €50 in ChatGPT credits to French students on UNiDAYS. Students can claim them after verifying their identity with a school email address, with no subscription required. These credits are mainly used for agent features such as Codex, ChatGPT Work and ChatGPT for Excel, and continue to be usable after the plan's quota is exhausted; ordinary conversations are unaffected. By comparison, OpenAI offers $100 in credits in the United States, while Google offers students in France one year of Google AI Plus.

    #Industry news#OpenAI

  9. MacRumors Apple Intelligence
    AI score17

    Hands-on: The OLED iPad mini is almost an iPad Pro mini

    MacRumors has gone hands-on with the OLED iPad mini launching this Friday (November 6): it is the second iPad to use an OLED display after the iPad Pro, with an 8.5-inch screen and a 60Hz refresh rate, lacking ProMotion. The body weighs 255 grams, making it Apple's lightest iPad, and it is the first to support IP67 water and dust resistance and nano-texture anti-glare. It is powered by the A20 Pro chip with 12GB RAM, and the 128GB model starts at $699, $100 more than the A17 Pro version.

    #English text is being prepared.

  10. MacRumors Apple Intelligence
    AI score48

    Apple's fiscal 2026 R&D spending rises to $45.77 billion, up 32% year-over-year

    Apple's fiscal 2026 R&D spending reached $45.77 billion, an increase of $11.22 billion from $34.55 billion in fiscal 2025, a 32% increase, equivalent to about $125 million per day. Revenue grew 15% over the same period, with R&D growth significantly outpacing sales, and R&D investment now approaching $1 for every $10 of revenue. Apple CFO Kevan Parekh previously linked the R&D budget increase to AI investment.

    Same story · 9 reports

    #Industry news#English text is being prepared.

  11. Deutsche Welle Chinese
    AI score51

    Li Jiaqi Sells 100,000 Qwen Office Membership Monthly Cards in 30 Minutes During Double 11 Livestream

    On the evening of November 2, streamer Li Jiaqi and Qwen Office CEO Chen Yusen promoted Qwen Office membership monthly cards in a livestream. Originally priced at 28 yuan per card, they were sold at a promotional price of 0.01 yuan, with 100,000 cards sold in 30 minutes, corresponding to a computing power quota of about 2 trillion AI Tokens. This is the first time AI computing power packages have entered Double 11 promotions as an important consumer category. Both Alibaba's Tmall and JD.com have set up AI Token sections. Alibaba said that from October 1 to 20, sales of AI computing power packages increased by 450% compared with the same period last month.

    Same story · 12 reports

    #English text is being prepared.#Industry news

  12. Deutsche Welle Chinese
    AI score42

    Merz: Germany wants to become an "AI country" by 2030

    German Chancellor Merz said at the digital summit that Germany must and aspires to become a leader in the AI era, and the government AI working group's final report sets the goal of building Germany into an "AI country" by 2030. The plan requires that by then at least 80% of large enterprises and 55% of small and medium-sized enterprises use AI, the number of AI startups triples, AI's share of economic output exceeds 10%, and 90% of federal agencies use at least one AI application to handle core tasks. The German government expects that the next 12 to 18 months will determine whether Germany can maintain a leading position in the international AI race.

    #English text is being prepared.#Industry news

  13. 404 Media
    AI score54

    《World of Warcraft: Forever》Flooded by AI-Generated Addons Before Launch, Warcraft Logs Announces API Changes

    Before the launch of 《World of Warcraft: Forever》, the related subreddit had already been taken over by a large number of AI-generated addons, some of which plagiarized the work of veteran developers. Warcraft Logs said it will adjust its API, restricting the practice of scraping data to display basic statistics, and introduce paid commercial keys; its team said that two years ago about 200 developers used the API and a small number abused it, whereas today about 1,000 of roughly 10,000 developers are abusing it.

    #English text is being prepared.#Agents#English text is being prepared.

  14. Shao MengAnalysis
    AI score57

    Google researcher Alex Krentsel reproduces the Weight-Space Programming experiment…

    Google researcher Alex Krentsel reproduced Exp. 5 of the Weight-Space Programming experiment from the MIT blog post: in a single response, each model, without training or using tools, directly writes out all of the roughly 20,000 weights of a playable Breakout network, with a network trained via PPO reinforcement learning as the 1× baseline. A cross-generation sweep shows that GPT-5 through 5.5 and Opus 4.5 through 4.8 are basically near or below the baseline, GPT-5.6 Sol and Opus 5 slightly exceed 1×, and only GPT-6.1 Sol, GPT-6 Astra and Opus 5.5 jump to about 18×–22×; he believes this is a step-change new capability that has just emerged in the latest batch of models. He emphasizes that the network whose weights are written has only about 20,486 parameters, several orders of magnitude away from trillion-parameter frontier models, and this cannot be extrapolated to mean that large models can already modify themselves; at present it is still an outer model modifying an inner small network, and he says he is reproducing it on open-weight models.

    Quoted postAlex Krentsel · @AlexKrentsel

    i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models. https://t.co/EhMsQaaOkL

    #Papers/Research#Reasoning#Open-source ecosystem

  15. Shao Meng
    AI score16

    Anthropic is hiring a Head of Communications for the Claude Developer Platform

    Anthropic is hiring a Head of Communications for the Claude Developer Platform (i.e., Anthropic's API), responsible end-to-end for external communications around flagship launches, capability updates, and developer news. The role pays an annual salary of $265,000–$295,000, is based in San Francisco, New York, or Seattle, follows a hybrid work model, and requires at least 25% of time in the office each week. Hard requirements include experience in communications, PR, or product marketing, as well as hands-on experience with platforms, APIs, developer, or enterprise products; nice-to-haves include 8–10 years of relevant experience and relationships with AI platform journalists.

    Quoted postedwin · @edwinarbus

    Looking for a developer (who happens to be great at comms) to lead comms for the Claude developer platform: https://t.co/UnVvvABP3d

    #Industry news#Anthropic

  17. X.PIN
    AI score42

    Unitree Taishan Robot Race: Robot dog wins championship, G1 humanoid robot overheats and withdraws

    At the Taishan Robot Race, Unitree's quadruped robot won the robot dog sprint championship with a time of 43 minutes and 41 seconds, but the G1 humanoid robot overheated after climbing all 3,100 stone steps and was carried off the course on a stretcher. The humanoid robot's compact joint housings accumulated heat during 45 minutes of continuous uphill walking, triggering the overheat protection, after which it squatted down and shut off; by contrast, performing a controlled backflip on a mat requires only 3 seconds of burst power. This shows that humanoid robots are not yet ready for commercialization.

    Same story · 4 reports

    #English text is being prepared.#Embodied AI

  18. The Beijing News
    AI score23

    DeepSeek Website and API Partially Outage, Now Fixed

    From 16:58 to 18:48 on November 3, DeepSeek's website and API experienced a partial outage lasting 1 hour and 49 minutes, during which the chat interface displayed the message "Server is busy, please try again later." As of press time, DeepSeek has resumed operation, and its status page shows "All systems operational."

    Same story · 5 reports

    #Industry news#DeepSeek

  19. The Beijing News
    AI score40

    Inside Beijing's OPC communities: Are AI entrepreneurs' one-person companies reliable?

    Beijing has established more than 20 OPC (one-person company) startup communities. Since it began operating in July 2025, the Zhongguancun AI Beiwai Community has gathered 57 OPC enterprises, 41 of which are pure one-person companies. Entrepreneurs say AI boosts product development efficiency by more than 10 times, but they generally believe that the last few miles from technology to implementation still require human control, with AI output about 70% correct and 30% requiring manual correction. Haidian District provides 100,000 yuan in startup capital and up to 2 million yuan in model voucher subsidies for OPCs founded by tech talent, while Yizhuang distributes up to 300 million yuan annually in computing power vouchers and other resource vouchers.

    #English text is being prepared.#Agents#Embodied AI

  20. Data Center Knowledge
    AI score21

    Liquid cooling is changing the metrics data centers need to measure

    Liquid cooling expands data center operations from single air-side measurement to a dual domain of liquid cooling and air. Direct-to-chip cooling delivers liquid to processors through CDUs and cold plates. On the liquid side, liquid temperature, pressure, flow rate, liquid level, leak detection, as well as coolant health indicators such as pH, turbidity, conductivity, and refractive index all require specialized instrumentation. Air-side management remains indispensable; the coolant supply temperature typically must be kept at least 2°C above the measured ambient dew point to prevent condensation. AI training workloads are intensive and inference workloads are intermittent, and clusters can go from idle to full load within minutes. Only real-time, accurate measurement can support dynamic CDU control.

    #Industry news#English text is being prepared.