Skip to content
Live
Nov 4Wed
  1. 36Kr · AI35

    Hongguo Short Drama has surpassed Doubao to become the second-largest application of Douyin Group.

    Douyin Group's internal app landscape has changed, with Hongguo Short Drama users surpassing Doubao to become the second-largest app, with unique total users of 1.029 billion, 430 million, and 419 million respectively; QuestMobile report shows that as of September 2026, domestic mobile internet monthly active users in China are 1.281 billion, a year-on-year increase of 0.9%, with an average monthly usage time per person of 189.8 hours, a year-on-year increase of 6.5%.

    Same story · 2 outlets
  2. ITHome · AI58

    After discovering 'listening' devices in OpenAI's executive meeting room, insiders said employees had been reporting security vulnerabilities for years

    After OpenAI discovered recording devices in its executive meeting room, insiders said employees had been reporting security vulnerabilities since the end of 2022 before ChatGPT was launched, believing management was more focused on catching up with competitors. Employees pointed out that OpenAI's security management was lax, making it easy for outsiders to access sensitive information, and that code access permissions could view details of many projects. OpenAI has adjusted its strategy, led by Théo Sotiropoulos, to strengthen defenses, but increased security measures may impact research and development speed.

  3. 36Kr · AI

    Apple warns that AI agents acting in 'unintended or unpredictable' ways pose risks

    In its annual financial filing, Apple stated that AI agents may act in 'unintended or unpredictable ways,' bringing risks of lawsuits or reputational damage. Such agents can take autonomous actions, including interacting with systems, executing code, or sending communications. The company emphasized that agents may take actions not instructed, highlighting the disruptive potential of AI agents.

  5. MarkTechPost71

    Anthropic open-sources FlashProve: using zkSNARK to prove responses from committed weights

    Anthropic's cutting-edge red team has released an open-source research prototype FlashProve, which uses zkSNARK to prove that a response is generated by committed model weights without revealing the weight information. On Llama 4 Maverick (400B MoE), a single H100 achieves a proof speed of approximately 110 tokens/s, with a proof size of 18.78 MB, and CPU verification takes 17 seconds.

    Same story · 5 outlets

    How the accounts differ:MarkTechPost states the proof volume is 18.78 MB; The Decoder states it is about 19 MB.

  7. ITHome · AI60

    OpenAI confirms: ChatGPT search volume has reached 20% of Google's

    Search volume on ChatGPT is about 20% of Google's search volume, confirmed by Adam Fries, OpenAI's head. 44% of American adults have used ChatGPT this year, far exceeding the 24% usage rate of Gemini.

    Same story · 3 outlets
  8. ITHome · AI52

    Microsoft's Win11 AI PC roadmap revealed: 4 tiers of hardware, 8GB memory also gets an AI ticket

    Microsoft announces four-tier hardware roadmap for Win11 AI PC, with the lowest tier Tier 0 covering PCs with 8GB memory, equipped with Intel Core 3/4 series, AMD Ryzen 100/200 series, Snapdragon X or X Plus processors; Tier 1 is for Copilot+ PCs with 16GB to 64GB memory, supporting configurations like Snapdragon X2 Plus to AMD Ryzen AI MAX+ 400.

  9. ITHome · AI58

    AMD Su Tseng Feng officially announced that the EPYC Venice processor has started to generate revenue and ship, with the first batch of OEM complete systems launching in the fourth quarter

    AMD CEO Su Tseng Feng announced during the earnings call that the sixth-generation EPYC 'Venice' server processor has started generating revenue and shipments in the third fiscal quarter of 2026, with the first OEM systems based on the processor expected to launch in the fourth fiscal quarter.

  10. TechCrunch · AI65

    xAI and OpenAI have already used modified jet engines to power AI data centers

    xAI has installed 8 modified jet engine turbines for its Memphis data center in Mississippi, and OpenAI will also use natural gas turbines at its first Stargate site in Abilene, Texas, in 2025. Due to the long grid connection and new turbine approval processes, some AI data centers have chosen to use jet engines modified from retired aircraft for power generation, but this is a short-term solution because their efficiency is lower than that of turbines designed for traditional gas power plants and they cause more pollution.

    Same story · 2 outlets
  11. TechCrunch · AI58

    Anthropic is developing a system called FlashProve to verify the identity of AI models

    Anthropic has released an open-source system called FlashProve, which uses cryptography to verify that specific outputs come from designated AI models, preventing model spoofing or replacement; the system requires AI labs to publish model fingerprints for verification without disclosing model weights, and the code is open-source.

    Same story · 5 outlets

    How the accounts differ:MarkTechPost states the proof volume is 18.78 MB; The Decoder states it is about 19 MB.

  12. ITHome · AI58

    Google adjusts K-12 education AI policies: Gemini, Search, and other services disabled by default

    Google will adjust its AI feature management policies for K-12 education users starting from 2027, requiring school administrators to proactively enable them. Services like Gemini, NotebookLM, Gemini in Classroom, and Google Search will only be available to students if enabled; if a school completely disables AI, students will not be able to use Google Search. This move aims to let schools decide how to use AI tools in the classroom, but the bundling of search with AI features has sparked new discussions.

    Same story · 6 outlets
  13. ITHome · AI65

    YouTube's promoted AI music composition tool 'Replace Songs' helps creators resolve audio copyright disputes

    YouTube's promoted AI music composition tool 'Replace Songs' allows creators to instantly generate four royalty-free instrumental background music tracks to replace copyrighted audio upon receiving an audio copyright claim, and automatically removes the relevant Content ID copyright notice. This feature currently only generates instrumental background music and does not support vocal or full song creation.

  14. MarkTechPost65

    Soniox STT v6 released: Adds intelligent agent text input and echo cancellation features

    Soniox STT v6 is its latest version of the speech-to-text API, which reads agent speech and analyzes its text content in real-time to achieve more accurate dialogue recognition and echo cancellation. The model supports over 60 languages, offering stt-rt-v6 (streaming) and stt-async-v6 (file) versions, running on the Soniox cloud API (only WebSocket streaming and asynchronous REST), with no self-hosting or device-side options.

  15. MarkTechPost55

    NVIDIA demonstrates 3 methods to accelerate MoE inference on DGX Spark, increasing Ling-3.0-flash throughput by 42.5%

    NVIDIA boosts the inference throughput of the 124B parameter Ling-3.0-flash model on DGX Spark by 42.5% and speed by 3.5x with a new MoE backend, FP8 output layer, and built-in speculative decoding, without compromising accuracy. This optimization is suitable for single-user desktop systems, improving throughput from 37.81 tokens/s to 53.89 tokens/s and reducing latency by 24.5% on 1,319 GSM8K problem tests.

    Same story · 2 outlets

    How the accounts differ:NVIDIA Technical Blog states decoding is the memory bottleneck due to MoE computation and output layer projection, while MarkTechPost highlights the 42.5% throughput increase and 3.5x speedup due to new MoE backend, FP8 output layer, and speculative decoding.

  16. Hugging Face Daily Papers65

    CrossSWE: Benchmark for testing large language models in cross-repository collaborative development

    CrossSWE benchmark tests the cross-repository consistency of large language models in cross-repository collaborative software development, including 171 instances, covering 165 repositories, 20 technical fields, and 13 programming languages. The evaluation shows that the strongest GPT-6.1 Sol only achieves 58.48% Pass@1, with cross-repository inconsistency being the main failure mode.

  18. Ars Technica · AI

    Ukrainian drone strikes prompt Russia to hastily plan underground data centers

    After Ukrainian drone strikes damaged facilities storing AI supercomputers and disrupted cloud services for millions of Russian users, Russian tech companies are rushing to reinforce and hide their data centers. The Kremlin has demanded that data center operators reserve computing power for emergency national needs and has even authorized the potential use of foreign data centers.

  21. The Verge · AI75

    Apple has opened up Siri AI capabilities to third-party apps in iOS 27.3

    Apple has opened up new Siri AI capabilities to developers with iOS 27.3, allowing third-party apps to access personal context, natural language prompts, and screen awareness to complete tasks. This update was released in Developer Preview form and adds app skills that can be created by developers, combining multiple tasks into a single action.

    Same story · 2 outlets

    Why it matters: The original text listed the specific interaction methods for Siri AI to access third-party applications, and readers can judge which cross-application operations the voice assistant can perform based on this.

  24. ITHome · AI62

    Apple opens Siri AI 'App Skills': enabling cross-app tasks

    Apple has opened up Siri AI's 'App Skills' to third-party developers, allowing Siri to perform multi-step cross-app tasks. Related features will be released with iOS 27.3, macOS 27.3, and simultaneous system updates, currently in the developer testing phase. Users can ask Siri to do something like 'through a chat app, send my favorite sneakers to my partner', and Siri will link multiple app operations to complete the task.

    Same story · 2 outlets
  26. The Decoder74

    OpenAI claims GPT-6 Astra designed a hardware device from scratch in less than a week.

    OpenAI states its hardware team used GPT-6 Astra to design a device that can communicate with Codex encoding agents as a side project, completed in less than a week. Astra handled all aspects including casing design, PCB layout with component selection and routing, firmware, and user interface. Contract manufacturers produced small batches of prototypes in less than a month, but due to Astra not always considering manufacturing constraints, tolerances, and assembly details, several rounds of modifications were required.

    Same story · 2 outlets

    Why it matters: OpenAI reveals the end-to-end hardware design process for GPT-6 Astra, also pointing out its shortcomings in manufacturing constraints.

  27. ITHome · AI65

    Apple iOS 27.3 expands Siri AI: Supports Traditional Chinese, covers operations of 15 apps, can manage files / schedule sending messages, etc.

    Apple expands Siri AI functionality in iOS 27.3 Beta, adding support for Traditional Chinese, with a total of 10 supported languages; covers 15 apps such as Calendar, Files, etc., can manage files, schedule sending messages, and perform complex tasks; Safari supports clearing history, managing tabs via Siri.

  28. OpenAI News37

    OpenAI appoints Jill Hazelbaker as Chief Corporate Officer

    OpenAI appoints Jill Hazelbaker as Chief Corporate Officer, leading the company's global communications and corporate positioning.

    Same story · 3 outlets
  30. OpenHands Blog71

    OpenHands and Snowflake introduce a workflow for training models using their own Agent harness

    OpenHands partnered with Snowflake AI Research, allowing developers to directly use the existing OpenHands Agent harness for reinforcement learning and training without rebuilding the harness in the training framework.

    Why it matters: The original text provides specific configuration methods and code locations for directly using the existing Agent harness for RL training, allowing readers to judge whether training and deployment can share a single harness.

  31. Hugging Face Daily Papers65

    Order-Only On-Policy Distillation released

    Order-Only On-Policy Distillation (O^{3}PD) trains student models to match the token ordering of the teacher model rather than probabilities, averaging 2.9 points better than OPD on HMMT25, AIME24, and AIME25, with a 5.6 point improvement on AIME25. This technique requires no temperature tuning and is unaffected by the teacher model's temperature.

  33. ITHome · AI50

    Apple Safari 27.3 Beta Released: Adapts to iPhone Duo and introduces vertical tabs, adaptive dark mode, and new Siri AI features

    Apple Safari 27.3 Beta Release Pushes Beta Version, Adapts to iPhone Duo and Introduces Vertical Tabs, Adaptive Dark Mode, and Expanded Siri AI Support. Vertical tabs make it easier to view full page titles, adaptive dark mode adjusts display for websites without dark mode, Siri AI can group tabs by voice, close all tabs for a website, or clear browsing history for specified websites.

Showing up to 50 pages. Use search or topics to find older stories.