All news

LLeiphone21:07

AI score53

Alipay AHA Multi-Terminal Agent Interconnection Solution Wins World Internet Conference "New Glory" Scenario Award

On November 2, the 2026 World Internet Conference Wuzhen Summit opened, and Alipay's AHA (Agent Hub Access) multi-terminal agent interconnection solution won the "New Glory" Scenario Award. AHA uses the Alipay App as the anchor for accounts and capabilities, and builds a three-layer interconnection protocol system based on the open-source A2A, MCP, and A2UI protocols, exporting Alipay's government services, transportation, healthcare, life services, and other capabilities to terminals such as phones, car infotainment systems, smart glasses, and large-model apps. The solution was officially released to the public in August this year. Currently, Abao's service capabilities have been exported to 6 types of terminals and deployed in more than a thousand service scenarios, and it has built an ecosystem with more than 20 enterprises, achieving cross-terminal collaboration with 5 major smartphone manufacturers and covering 16 automobile brands on the car side.

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

LLeiphone21:07

AI score27

XGIMI Jide's full lineup of new AI products unveiled at the INNO100 first "hard" gift event: MemoMind One, G1 and L1

On November 1, the full lineup of XGIMI Jide's new AI products was unveiled at the INNO100 first "hard" gift event, including the AI display glasses MemoMind One, the AI memory glasses G1 and the AI memory band L1. MemoMind One uses Zeiss Lingxi lenses with a custom prescription lens solution, is equipped with open dual speakers jointly tuned by Harman, adopts a camera-free design, and supports AI long-term memory capabilities.

#Product updates#English text is being prepared.

LLeiphone21:07Product release

AI score69

Zhiyuan Unveils Worken, a Metric for Effective Intelligent Work Hours, Along with GO-3 and Other Embodied Technologies

At its Shanghai A.I. DAY on October 29, Zhiyuan proposed Worken (effective intelligent work hours) as a metric, with the formula being the number of active robots multiplied by autonomous working hours multiplied by the output efficiency ratio, and on site demonstrated real-time work-hour dashboards for robots at the Longcheer production line, Shangri-La Hotel, and Chimelong Park. At the same event, it released the third-generation embodied foundation model GO-3 Preview, the four-layer architecture AGILE motion engine, the omni-modal interaction large model WITA-Omni, the third-generation hardware architecture AHA3 and the heavy-load body G2 Max, and announced that the embodied-native operating system Lingqu OS 1.0 will be fully open-sourced by the end of the year. Zhiyuan said it shipped 9,700 units in the first half of the year, its 20,000th robot rolled off the line in September, and first-half revenue was close to 2 billion yuan.

#Industry news#Embodied AI#Model releases

LLeiphone21:07

AI score46

RoboScience Machine Science completes A2 round of hundreds of millions of yuan in financing, as industrial capital continues to bet on embodied intelligence

RoboScience Machine Science completed its A2 round of financing on November 2, led jointly by Amber Capital, Hongruida Investment, and Shanghai Semiconductor Industry Investment, with 360 Strategic Investment, Shuang Capital, Guolian Minsheng Capital, Jinding Capital, and several listed company platforms participating, amounting to hundreds of millions of yuan. This company, which began operations in March 2025, has cumulatively completed five rounds of financing. It has previously self-developed the RoboMirage simulation engine, the Visics general embodied large model, and the REX G1 wheeled humanoid robot, and has co-created the cloud-based embodied intelligence service EaaS with Tencent Cloud.

Same story · 3 reports

#Industry news#Embodied AI

LLeiphone21:07

AI score42

Post-90s Peking University associate professor Dong Hao: How Qiyuan Robot uses HIL-UMI and the horizontal Scaling Law to tackle home embodied intelligence

Peking University associate professor Dong Hao has taken on the role of Chief Scientist at Qiyuan Robot, leading the team to tackle general-purpose embodied intelligence for the home. He proposes using HIL-UMI bodyless data collection for post-training, and using a world model to amplify one real-robot trajectory into 50 equivalent samples, magnifying the value of real-robot data by 50 times. He also proposes a horizontal Scaling Law with dual axes of data volume and task volume, enabling new tasks to have a higher initial success rate and require fewer samples.

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

LLeitech21:01

AI score34

WeChat tests "Voice Clarity Enhancement" in limited rollout: one-tap local AI noise reduction for voice messages

WeChat is testing a "Voice Clarity Enhancement" feature in a limited rollout. The entry point is under "Me — Settings — General — Accessibility." Once enabled, it automatically extracts the human voice and suppresses background noise such as wind and traffic when voice messages are played. The feature only affects playback, and noise reduction is performed in real time entirely by a local AI algorithm on the phone; voice content is not uploaded. "Dialect-to-text" and other features are also being tested in the same limited rollout and are still being pushed out in batches.

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

AAlibaba Cloud Developer Community AI20:54

AI score27

Multi-Agent Data Governance: Who Proves That the Agents Changed It Correctly?

Yang Chaoran, founder of Shuwo Technology, proposes that the core increment of agents in data governance is to expand the space of investigable, explainable, and verifiable candidates, rather than automatically making all candidates take effect. The value of multi-agent collaboration comes from complementary evidence rather than majority agreement; work should be divided according to uncertainty and dependencies, and automation should be undertaken through evidence-bearing changes, with promotion of feedback after verification. The article presents an artifact-driven process, review and submission boundaries, counterexamples, controlled evaluations, and a framework verification checklist.

#Papers/Research#Agents#English text is being prepared.

AAlibaba Cloud Developer Community AI20:54

AI score38

Why Stronger Models Still Make Organizational AI Answer Wrong: Data Semantic Governance and the Implementation Boundaries of Enterprise Agents

Yang Chaoran, founder of Shuwo Technology, proposes that the reliability of organizational agents depends not only on the model, but also on whether business meanings can be correctly selected and continuously used, and offers six judgments centered on legitimate definitions, unknown conditions, applicable context, authorization boundaries, industry variables, and consumption relationships. The article discusses architecture and evaluation using synthetic counterexamples and public specifications, emphasizing that semantic governance is one of the operating conditions and does not replace the separate validation of data, models, security, and responsibility.

#Industry news#Agents#RAG

AAlibaba Cloud Developer Community AI20:54

AI score14

Alibaba Cloud Wanwang Double 11 Special Offers: 9.9 yuan AI website building, 99 yuan GEO package, .COM domains as low as 60 yuan for the first year

The Alibaba Cloud Wanwang Double 11 event starts on November 2. Wan Xiaozhi is launching a 9.9 yuan website building trial (including a 1-year .CN domain, ICP filing, and 2,000 inspiration points) and a 99 yuan GEO package (limited to 500 units, including monitoring, copywriting, and distribution, covering 4 major domestic mainstream AI platforms). In bundle purchases, .COM is as low as 60 yuan for the first year, and standalone new registrations/transfers are as low as 76 yuan.

#English text is being prepared.

YYicai20:45

AI score34

The Token Economy: A Compute Ledger Rooted in the Physical World

Tokens are transforming from a natural language processing concept into the basic economic unit that drives model operation and settles cloud computing power, with costs built up from the bottom by electricity, chip depreciation, infrastructure, and model amortization. The unit price of output tokens is usually several times that of input, because the decoding stage is constrained by memory bandwidth bottlenecks; some regions in China have partnered with financial institutions to try incorporating enterprises' monthly token consumption and the value of computing power service contracts into credit assessments. The article also emphasizes that computational tokens are subject to physical limits such as chip yield, memory capacity, and grid carrying capacity, and are fundamentally different from blockchain tokens.

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

KKling AI20:38First-party

AI score15

Kling AI Partners with Chroma Awards Season 2, with Over $450,000 in Prize Money

Kling AI has partnered with Chroma Awards to launch Season 2, offering over $1 million in promotional opportunities, over $450,000 in cash prizes, and student internship opportunities. The competition features five tracks: Film & TV, Advertising, Music Video, Games, and Behind the Scenes. Submissions using Kling AI and sharing on social media can unlock additional rewards. Submissions are open until January 7.

#Industry news#Video#Image generation

PPixVerse20:37

AI score10

PixVerse and WaveSpeed Host V6 Max Creation Campaign

PixVerse announced that it is partnering with WaveSpeedAI to host the PixVerse V6 Max Creator Campaign, running from November 1 to 14. WaveSpeedAI said the campaign offers creation credits, and details can be found on its Discord.

Quoted postWaveSpeedAI · @wavespeed_ai

Got a wild idea? We might just help you make it happen. WaveSpeed × @PixVerse are teaming up for a special PixVerse V6 Max Creator Campaign! 🎬 💡 Your vision. Your creativity. No holding back. 🎁 Creation credits to help bring your ideas to life! Ready to take your vision to the MAX? 👇 🔗Join our Discord for the details

#Industry news#Video#OpenAI

SSmart City Industry Analysis20:27

AI score33

Xinhua News Agency's China Economic Information Service and CAICT jointly released the "2026 Embodied Intelligence Industry Innovation Report," proposing the XYZ curve and Worken

Xinhua News Agency's China Economic Information Service, together with the Shanghai Industrial Innovation Center of CAICT, released the "2026 Embodied Intelligence Industry Innovation Report," which uses the "XYZ curve" proposed by AgiBot to divide development into three stages: early-adopter development, deployment growth, and deployment popularization, and judges 2026 to be the inaugural year of the deployment stage. The report proposes a "three intelligences in one" technology path, integrating the three major intelligences of motion, operation, and interaction with the embodied agent in an end-to-end manner, and introduces "effective intelligent work hours (Worken)" to measure physical AI productivity through three factors: the number of active robots, autonomous working hours, and the output efficiency ratio.

#Papers/Research#Embodied AI

PPhoenix Technology20:27Research

AI score75

ACM Web Conference 2026 paper deconstructs ChatGPT memory: 96% is written unilaterally by the system

The paper "The Algorithmic Self-Portrait: Deconstructing Memory in ChatGPT," accepted by ACM Web Conference 2026, used GDPR data access rights to obtain 1,058 conversations, 22,971 interaction turns, and 2,050 memory entries from 80 users, finding that only 4.1% of memories were explicitly requested by users to be remembered, while 95.9% were written at ChatGPT's own discretion. The study also shows that 28% of memories contain personal data as defined by GDPR, 7% contain special category data (of which health information accounts for 61%), and 52% contain inferences about users' psychological states. The researchers replicated the memory extraction process using Qwen2.5-32B, Gemma3-27B, and GPT-OSS-20B, and built Attribution Shield to warn about sensitive memories before questions are asked and provide rewrites that reduce personal attribution by 94%–100%; the article also includes a prompt for checking the memories and user profile currently accessible to ChatGPT.

#Papers/Research#Safety / alignment#OpenAI

PPhoenix Technology20:27Opinion

AI score61

AI "Kills the Crypto World"? The $6.3 Billion Theft Case and the Crypto Security Predicament

The article reviews several recent security incidents in the crypto world, one of which involved the theft of about $6.3 billion from a hot wallet. Chainalysis said the attackers completed high-frequency, large-scale transactions in an extremely short time, suspected to have been carried out by AI. The article cites a Blockworks report stating that more than 300 crypto attacks have occurred this year, a year-on-year increase of 111%, and mentions the growing capability of models such as Anthropic's Mythos in vulnerability discovery. The author argues that attackers mostly bypass the underlying algorithms of Bitcoin and Ethereum and turn to "backdoors" such as third-party components and unreleased branches, while the crypto world also faces new dependency relationships when seeking security tools from model companies.

#English text is being prepared.#Safety / alignment#Agents

PPhoenix Technology20:27

AI score46

The AI-coded "Bed Bug Map" mini program goes viral: over 600 hotels marked, developer Wang Chuang responds to controversy

Individual developer Wang Chuang used AI programming to develop the "Bed Bug Map" mini program, which since launching on September 21 has attracted over 200,000 users, with the marked area expanding from coastal major cities to the whole country, covering more than 600 hotels. Because some marks lacked objective evidence, hotels filed appeals, and the mini program was suspended twice; the new version was restored on November 1, changed so that users must first submit evidence and only after manual review can a mark be made. Wang Chuang's main job is in manufacturing; he had no prior programming experience and only began writing code with AI in July this year.

#Product updates#Agents

WWired AI20:27

AI score47

When to use and when not to use AI web search

WIRED has outlined the applicable boundaries of web-connected search in AI chatbots such as Claude, ChatGPT, and Gemini: AI search performs well when querying widely known simple facts or handling complex needs that require integrating multiple sources; traditional Google search is more reliable when querying obscure details, breaking news, and high-risk issues such as health, legal, and financial matters. Stanford University research points out that AI is very capable of extracting information when it finds accurate sources; the problem is that when it cannot find accurate sources, it still gives confident answers; UC San Diego research found that AI sometimes chooses sources based on relevance rather than authority.

#Tutorials / practice#Search#Evaluations / benchmarks

SSina Tech20:27

AI score71

Marvell Announces All-Stock Acquisition of Lumentum, Implied Enterprise Value of $115 Billion

Marvell Technology and Lumentum have signed a definitive agreement under which Marvell will acquire Lumentum in an all-stock transaction, with each Lumentum share exchanged for 4.500 shares of Marvell common stock, implying an enterprise value of $115 billion and a per-share price of approximately $1,132, a premium of about 23.9% over the closing price on October 28, 2026. Upon closing, former Marvell shareholders will hold approximately 67% of the combined company and Lumentum shareholders approximately 33%; the combined company's total addressable market is expected to expand to $550 billion, with annualized cost synergies expected to exceed $1 billion. The transaction is expected to close in the first half of 2028, subject to shareholder votes by both companies and regulatory approvals in various countries.

#Industry news#English text is being prepared.

CCailian Press20:26Product release

AI score53

DeepSeek Harness Team Unveils Super-agent Architecture

The DeepSeek Harness team published a technical article on its official WeChat account, unveiling a new architecture for autonomous agents called Super-agent. Users only need to communicate with an "AI leader" responsible for coordination, and the tasks are dispatched by it to multiple AIs to advance separately. The architecture adopts ideas such as a dual-layer architecture, event-driven design, and separation of storage and compute, with the system taking over session management and context synchronization. An internal instance built on this approach has been running continuously, able to autonomously identify software defects in group chats, reproduce problems in the background, and seek fix solutions.

Same story · 21 reports

#Product updates#Agents#DeepSeek

TTrail of Bits Blog20:25First-partyAnalysis

AI score71

Trail of Bits uses Codex to backport security patches for stable Linux, covering 149 CVEs

Trail of Bits used Codex together with GPT-5.6 to backport security patches for the stable Linux kernel and distributions. As of October 29, it had delivered fixes for 149 CVEs to stable kernel branches, plus several Fedora and Ubuntu package fixes. In its workflow, the agent identifies missing fixes, adapts patches, assembles verification evidence, runs tests, and prepares submission materials, while engineers review and submit them to maintainers; excluding initial project setup, a simple backport requires about 30 minutes of human effort, and complex cases about 3 hours. The authors also provide the number of CVEs fixed in newer versions but not backported in each stable branch of the Linux kernel and in Debian 13 and Ubuntu 26.04, and note that regression testing and maintainer acceptance remain necessary steps.

#Safety / alignment#Agents#English text is being prepared.