Archive

KKling AI19:59First-party

AI score22

Kling 4.0 Generates New Emil Heliot Campaign Film

Copenhagen fashion brand Emil Heliot has released the campaign film for its Core Uniform collection, directed by UNVEIL with Julius Juul as creative director, produced using Kling 4.0. UNVEIL co-founder Tom-Jacques Perret said that Kling 4.0 balances chance and controllability—first building a world, then letting the film unfold naturally within it.

#Industry news#Video#Multimodal

MMIT Tech Review19:36

AI score43

How "Reverse Game Theory" Constrains AI Agents

MIT economist Stephen Morris and others proposed in a preprint paper using "reverse game theory" (mechanism design) to constrain AI agents: instead of rewarding only test scores, they reward the accuracy with which a model's answer predicts the output of a second model, making deliberate "letting things slide" a losing choice. Princeton professor Elad Hazan, meanwhile, envisions having models take turns acting as solver and auditor, with cheating solvers and negligent auditors both penalized, and disagreements referred to human experts for adjudication. This approach assumes agents act rationally, but human performance under mechanism design does not always match the predictions of economic models.

#Safety / alignment#Agents#Papers/Research

MMistral AI · X19:25

AI score50

Mistral Large 4 Builds Satellite Tracking App Orbital Trace

Mistral AI had Mistral Large 4 build a real-time 3D tracker for tracking roughly 19,000 satellites and debris. The model built a working app, tested it like a user, optimized speed, and passed rigorous validation; the finished product is called Orbital Trace.

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

TThe Decoder19:08

AI score53

OpenAI raises bug bounties to a maximum of about $172,000

OpenAI announced it is quadrupling its bug bounties: critical vulnerabilities in core products such as ChatGPT and Codex can earn up to $30,000, exceptional findings can reach $100,000, and with quality, loyalty, and event bonuses stacked on top, the total comes to about $172,000. It added a five-tier loyalty program and time-limited bonus campaigns tied to new product launches, and those who submitted vulnerability reports in the past 30 days will receive the difference between the old and new rates. OpenAI is also preparing a security sprint in which the security firm Cathedral and its own engineers will stress-test its systems together; the company says 25% of its product engineers took part in vulnerability hunting during the last sprint.

Same story · 2 reports

Why read thisThe original article lists the specific tiers and bonus structure of the bug bounty, which can be compared with similar programs at Google, Apple, and Microsoft to see where OpenAI's security investment stands.

#Safety / alignment#OpenAI#Industry news

AAyi AI Notes18:57

AI score17

Author uses Claude Opus5.5 + Blender to review OKX's five-year comeback

Ayi AI Notes spent 16 hours using Claude Opus5.5 plus Blender to render a video, reviewing OKX's three contrarian product lines over the past five years, from the trough of the withdrawal crisis to a strategic reversal: the industry's first unified trading account, a fully open-source self-custody Web3 wallet, and routine Merkle tree proof of reserves exceeding 100%. The post opens with the incident in which Binance was drained of $6.3 billion by hackers in 20 minutes, and claims that OKX wallet traffic reaches more than ten times that of traditional exchanges.

Quoted postStar_OKX · @star_okx

English text is being prepared.

#Industry news#Open-source ecosystem

SSmart City Industry Analysis18:52

AI score23

Summary of Results Released at the 2026 Artificial Intelligence Industry Conference: AIIA Releases 20 Results in a Concentrated Manner

The 2026 Artificial Intelligence Industry Conference was held in Jinan, where the Artificial Intelligence Industry Alliance (AIIA) released 20 research results in a concentrated manner, covering industry empowerment, technology evaluation, and standards systems. Among them, the "Atlas of High-Value Scenarios for AI Empowering New-Type Industrialization" covers 20 sub-sectors and more than 1,000 high-value scenarios, and the CSI BOC Shanghai-Hong Kong-Shenzhen Artificial Intelligence Industry Index selected 314 target enterprises. The conference also released the On-Device Intelligence Testing Platform 1.0 and the "Embodied Intelligence Development Report (2026)" and other reports.

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

PPhoenix Technology18:51

AI score37

DeepSeek Service Restored; At Least 10 Major Outages Have Occurred So Far This Year

On the afternoon of November 3, DeepSeek's chat feature was unavailable. The official status page showed that since 16:58, the web/API had been partially interrupted, affecting the chat service, DeepSeek V4 Pro, and the DeepSeek V4.1 Flash API. The official website subsequently said that the interruption had been fixed and was being monitored, but the two APIs and the chat service were still experiencing degraded performance. Official service status records show that there have been at least 10 major outages so far this year, including service unavailability, service anomalies, and more frequent reports of degraded performance.

Same story · 2 reports

#Industry news#DeepSeek

PPhoenix Technology18:51

AI score49

arXiv paper A Heart Means Recursion: Show the model a red heart, and it switches to writing code recursively

The arXiv paper A Heart Means Recursion found that when a red heart ❤️ is attached next to a binary tree traversal problem, the model shifts from an iterative approach to a recursive one, but when asked about the meaning of the heart, it calls it irrelevant. The paper records that GPT-5.4 showed "unfaithful attribution" in as many as 74.7% of samples, meaning its reasoning explanations are decoupled from its actual behavior. The researchers therefore argue for replacing explanation tests with behavioral tests: change only that "element that should be irrelevant" and see whether the output shifts systematically.

#Papers/Research#Reasoning#Agents

SShanghai Observer18:42

AI score26

Zhiyuan Receives 2026 World Internet Conference Outstanding Contribution Award; Embodied Intelligence Deployed in Factory, Cultural Tourism and Retail Scenarios

At the 2026 World Internet Conference Wuzhen Summit, Zhiyuan received the Outstanding Contribution Award for its systematic technological breakthroughs and industrialization practices in the field of embodied intelligence. Zhiyuan has built a "three-intelligence-in-one" embodied intelligence technology system and released Worken, the first metric that defines the value of physical AI. Models such as the Jingling G2, Lingxi X2 and Yuanzheng A3 have been deployed in real scenarios including Longcheer's manufacturing lines, Aishida stores and Shangri-La hotels, respectively.

#Industry news#Embodied AI

SShanghai Observer18:42

AI score28

Zhiwang Weilai Appears at CeMAT, Showcasing Warehouse Picking Robot Armstrong Pro in Live Demonstration

Embodied intelligence company Zhiwang Weilai exhibited two generations of warehouse mobile picking robots, Armstrong and Armstrong Pro, at the CeMAT Asia International Exhibition for Logistics Technology and Transportation Systems, and recreated a real warehouse environment with dense shelving and narrow aisles at its booth for live operation demonstrations. The total degrees of freedom of the Armstrong Pro have increased from 17 in the first generation to 22, with an effective working height of 0–2.4 meters; it adopts a four-wheel omnidirectional chassis that supports forward and backward movement, lateral movement, diagonal movement, and in-place rotation. Through its self-developed WISE-RL world model architecture, Zhiwang Weilai distills transferable general manipulation capabilities from real-world operational scenarios.

#Industry news#Embodied AI

SShanghai Observer18:42

AI score28

Bosideng's Changshu smart factory uses humanoid robots, with AI covering core down jacket production processes

At its approximately 18,600-square-meter AI Leading Science and Innovation Center in Changshu, Jiangsu, Bosideng is applying AI technology to core down jacket processes including cutting, sewing, down filling, and warehousing. Humanoid robots are responsible for grabbing cut pieces and loading materials, while X-ray intelligent inspection equipment checks down-filling uniformity and skipped stitches. The factory's intelligent hanging system comes from Jack Technology of Taizhou, Zhejiang. Bosideng is also working with a Jiangnan University team to advance the construction of an AI and embodied robotics laboratory, conducting research and development around scenarios such as sorting, packaging, and quality inspection.

#Industry news#Embodied AI

Tt3n18:38

AI score60

Goldman Sachs Restricts Claude Code and GitHub Access to Developers Only

Starting in October, Goldman Sachs will restrict access to Anthropic's Claude Code and Microsoft GitHub to developers only; previously, traders had used these tools to monitor client trading flows and model market changes and positions. According to Bloomberg, citing anonymous sources, the adjustment has caused dissatisfaction among some traders; a Goldman Sachs spokesperson said the reason is more precise monitoring of productivity and risk management, and emphasized that the tools should be given to those best suited for long-term use. The restriction does not apply to Strats employees in roles at the intersection of trading, quantitative research and software development, and Goldman Sachs' in-house assistant, which integrates models from OpenAI, Anthropic and Alphabet, remains open to all employees.

Same story · 3 reports

Why read thisGoldman Sachs has pulled Claude Code and GitHub back to developers only, reflecting Wall Street's redrawing of the boundaries of AI tool usage under cost pressure.

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

Tt3n18:38

AI score60

Anthropic modifies Claude system prompt to prohibit recreating existing games

Anthropic added a clause to the system prompts of Claude Code and Claude Opus 5.5 prohibiting the recreation of specific existing games, even if developers use a different name; the company has confirmed this change and said it is to avoid intellectual property infringement. The original text states that Claude can still assist in developing games using the user's own designs and assets, but users on GitHub and Reddit have reported that legitimate requests such as analyzing game mechanics, patching game code, and even troubleshooting Windows crashes were also refused. Anthropic employee Addy Osmani said the rules will be adjusted to find a balance and asked users to send him requests that were mistakenly refused.

Same story · 14 reports

Why read thisThe original text presents specific feedback that after Anthropic tightened its system prompt, legitimate development projects were also mistakenly blocked, allowing readers to judge the tension between copyright protection and usability.

#Product updates#Safety / alignment#Anthropic

LLeiphone18:23

AI score26

China Mobile 2026 Mobile Network Conference Renamed: From Global Developer Conference to MobileLink 2026

China Mobile's annual industry event has officially been renamed the "2026 Mobile Network Conference (MobileLink 2026)," with the theme set as "Co-creating a New Form of the Intelligent Economy," shifting from the Global Developer Conference and Global Partner Conference toward building a foundation for the intelligent economy. The conference emphasizes that networks are upgrading from transmitting bits to a foundation for computing power scheduling, intelligent task distribution, and multi-agent collaboration, with the operator's role shifting from "empowerer" to "enabler." China Mobile has already amassed over a thousand 6G patents, and the MoMA mobile model service platform has gathered more than 300 mainstream AI models.

#Industry news#English text is being prepared.

LLeiphone18:23

AI score41

Unitree Robot Withdraws from Mount Tai Climbing Race Due to Overheating, Carried Down on a Stretcher

The Unitree Robot G1 Developer Edition withdrew from the inaugural CMG World Robot Mount Tai Climbing Race due to the machine overheating and was carried down the mountain by staff on a stretcher. The course is 3.9 kilometers long with more than 3,100 steps in total, and the operator chose to withdraw for safety reasons. Other robots at the same event also wobbled and staggered, and some needed to be pulled with ropes.

#Industry news#Embodied AI#Agents

LLeiphone18:23

AI score34

Ant Digital Technologies Brings Agent Super Factory to Hong Kong Fintech Week, Releases ZOLOZ KYA Solution

At Hong Kong Fintech Week, Ant Digital Technologies released its "AI Agent Super Factory" solution and launched ZOLOZ's trusted identity governance solution for the agent era, KYA (Know Your Agent), extending its capabilities from KYC and KYB to identifying agents. Its core platform Agentar already has a general version and three industry versions for finance, energy, and healthcare, with the solution covering industries such as finance, energy, healthcare, and consumer. An IDC report shows that Ant Digital Technologies ranks first among non-cloud vendors in China's private deployment market for agent development platforms, and holds the top spot in the financial industry large model platform market with a 13.3% share.

#Product updates#Agents#Safety / alignment

LLeiphone18:23

AI score17

Qi An Xin upgrades Security Maker Summit to AI Maker Summit

Qi An Xin announced during the 2026 World Internet Conference Wuzhen Summit that it is officially upgrading the Security Maker Summit, which has been held for eleven editions, into the "AI Maker Summit," expanding its track from cybersecurity to AI innovation and application fields. Of the first ten editions, all 130 teams that reached the finals received multiple rounds of financing after the events, with total financing exceeding 6.5 billion yuan. The new platform will continue the original selection mechanism and venture capital ecosystem, focusing on cross-sector innovation between AI and various industries.

#Industry news#Safety / alignment

LLeiphone18:23

AI score25

Ant Group's self-developed AI data platform OceanBase wins World Internet Conference Outstanding Contribution Award

Ant Group's self-developed AI data platform OceanBase received the "World Internet Conference Outstanding Contribution Award" at the 2026 World Internet Conference. The conference's award citation stated that it has built AI data infrastructure and served the development of the global digital economy. This is another major World Internet Conference award for OceanBase, following its receipt of the World Internet Leading Scientific and Technological Achievement Award in 2022.

Same story · 3 reports

#Industry news#English text is being prepared.

LLeiphone18:23

AI score21

Baidu Wins World Internet Conference Distinguished Contribution Award for Endangered Species AI Guardian Project

Baidu has won the Social Contribution category of the 2026 World Internet Conference Distinguished Contribution Award for its "Endangered Species AI Guardian" project, developed in collaboration with the International Fund for Animal Welfare (IFAW). The project uses PaddlePaddle and the ERNIE large model to identify images of endangered wildlife products, achieving an accuracy rate of 86% and covering 34 categories of endangered species and their products. As of April 2026, the project has assisted in screening nearly 500,000 images, helped remove nearly 18,000 illegal trade advertisements, and open-sourced the recognition model through PaddlePaddle.

#Industry news#Multimodal#Open-source ecosystem

LLeiphone18:23

AI score28

Three-Body Universe and Infinigence AI Launch Sophon Agent 1.0, Partner with Tripo to Launch 3D Warship Co-Creation Plan

At the 2026 Three-Body Day · Dehydration Soak Carnival, Three-Body Universe, together with Infinigence AI, launched Sophon Agent 1.0, positioned as a "Three-Body worldview expert" that provides conversational interaction grounded in knowledge from the original novels. The two parties use ultra-long text memory, multi-Agent interactive calibration, and positive and negative samples from the works to reduce characters going "out of character." Three-Body Universe also partnered with Tripo, the AI 3D generation platform under VAST, to launch the "Three-Body · Global 3D Warship Co-Creation Plan," enabling users without modeling skills to participate in warship design and explore 3D printing.

#Industry news#Agents#Multimodal

LLeiphone18:23

AI score58

Xiaomi Wins Two Awards at World Internet Conference, Discloses Progress on Humanoid Robot Deployment in Factories and Apartments

Xiaomi Group received the "Outstanding Contribution Award" at the 2026 World Internet Conference, and its "Key Technologies for a General-Purpose Robot Manipulation Foundation Model Oriented Toward Large-Scale Deployment" received the "Leading Science and Technology Award." Xiaomi also disclosed that its humanoid robot completed an 8-hour operation for the first time at the center console side cover panel sequencing workstation in an automobile factory, and achieved autonomous anomaly correction and system-level coordination with the production line; on the home front, an Agent system built on the Xiaomi-Robotics-1 pre-trained model is undergoing trial operation at Xiaomi Youth Apartments, completing long-horizon task validation lasting more than 30 consecutive minutes, covering 3 rooms, and including 50+ tasks. Xiaomi-Robotics-1 is pre-trained on 100,000 hours of real-world manipulation data, and Xiaomi claims it is the first in China to systematically verify the existence of the Scaling Law in the embodied intelligence field.

Same story · 2 reports

Why read thisXiaomi disclosed deployment data for its embodied foundation model in two types of scenarios—factories and apartments—offering a view of the progress in migrating the same technology foundation across scenarios.

#Industry news#Embodied AI#Agents

TThe Decoder18:08

AI score59

Anthropic restricts cloning existing games in Claude Code

Anthropic is preventing its coding assistant Claude Code from replicating existing video games. Employee Addy Osmani confirmed on X that two sentences have been added to the Opus 5.5 system prompt to prevent cloning others' games, and that it triggers even if the clone is renamed or disguised as original work. The rule has been in effect for several days; users can still build their own games, works in the same genre, and projects using custom assets, but some have reported that original projects also ran into problems. The change comes after a surge of AI cloning projects, and Osmani said Anthropic plans to continue adjusting the relevant rules.

Same story · 14 reports

Why read thisAnthropic has added restrictive clauses to the Claude Code system prompt, giving readers a look at how AI coding tools are tightening their boundaries around intellectual property.

#Product updates#English text is being prepared.#Safety / alignment

DData Center Knowledge18:08

AI score58

AI Expansion Hits Local Power Bottleneck: Data Center Grid Connection Waits Can Reach Up to Ten Years

Data center projects are being delayed by power access and local opposition; for projects starting construction between Q3 2025 and Q3 2026, the average wait from announcement to groundbreaking is slightly over two years, compared with about 18 months for projects from early 2020 to mid-2025. GlobalData analyst Martina Raveni said new grid infrastructure takes 5 to 15 years, while data centers take only 1 to 3 years, and developers in major European hubs can face grid connection waits of up to ten years. GlobalData expects annual data center electricity consumption to rise from 590 TWh in 2025 to 1520 TWh in 2030; IDC's Luis Fernandes believes there is currently no hard evidence that compute capacity has constrained AI development.

Why read thisThe article uses the contrast between construction timelines and power grid buildout timelines to illustrate that power access is becoming the main bottleneck for data center expansion.

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

DData Center Knowledge18:08

AI score44

Synergy: Global cloud infrastructure spending reached $168 billion in Q3 2026, up 55% year-over-year

A Synergy Research Group report shows that global cloud infrastructure services spending reached $168 billion in the third quarter of 2026, up 55% year-over-year, with AI-related services accounting for about one-third of revenue. Amazon led with a 26% share, followed by Microsoft at 19% and Google at 16%; Microsoft and Google grew faster than Amazon, while Oracle and neoclouds such as CoreWeave, Crusoe, and Nebius were the fastest-growing vendors outside the top three. JLL data shows that hyperscalers accounted for 59% of North American data center tenant demand, neoclouds for 11%, and AI companies for 7%.

#Industry news#English text is being prepared.

MMeshy17:49First-party

AI score28

Meshy Releases N1 Technical Report

Meshy shared the full-text link to the Meshy N1 technical report on X.

#Papers/Research#Image generation

PPhoenix Tech17:49

AI score65

Anthropic Warns of Recursive Self-Improvement Risk in S-1/A Filing

In the risk disclosures of its S-1/A prospectus supplement updated on Monday, Anthropic stated that the company may soon reach the state of recursive self-improvement as defined in its Responsible Scaling Policy. The filing said the company has already used AI to develop new models and achieve some capability jumps, but the team did not notice at the time, and may in the future only notice after an acceleration in R&D has already occurred. Anthropic also stated that none of its existing models have crossed this threshold, and committed to retaining human involvement in frontier model R&D for the foreseeable future, and is considering revising its Responsible Scaling Policy.

Same story · 4 reports

Why read thisAnthropic lists recursive self-improvement among risk disclosures in its prospectus supplement, giving readers a view of how a leading lab defines this threshold.

#Safety / alignment#Anthropic#Industry news

PPhoenix Tech17:49

AI score48

Goldman Sachs Tightens Access to AI Coding Tools Such as Claude Code, Some Traders Unhappy

In internal guidance updated in October, Goldman Sachs tightened access to Claude Code and Microsoft GitHub-related tools, mainly restricting them to use by engineers, which has drawn complaints from some traders. Previously, some traders relied on these tools to monitor client trading flows and simulate market and position changes; "strats" employees whose work spans trading, quantitative research and software engineering are not subject to this restriction. Goldman Sachs' own AI assistant, GS AI Assistant, integrates models from OpenAI, Anthropic and Alphabet.

Same story · 3 reports

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

PPhoenix Technology17:49

AI score62

Oracle Raises $20 Billion for AI Expansion, Shares Fall Nearly 9%

Oracle announced that it would raise approximately $20 billion through bond issuance and stock sales to support AI and cloud computing infrastructure construction. After the news was announced, the stock price fell 8.8% to $119.22, and the cumulative decline for the year widened to about 39%. The offering consisted of 86.2 million shares priced at $116 per share, a discount of about 11.2% to last Friday's closing price; the bond portion consists of five tranches of fixed-rate notes with maturities ranging from 2029 to 2046 and interest rates between 7% and 8.65%. As of August 31, Oracle's notes payable and other borrowings reached $125.34 billion, and the company expects revenue to exceed $225 billion by fiscal year 2030.

Same story · 2 reports

Why read thisThe scale of the financing and the magnitude of the stock price decline are presented side by side, allowing readers to observe the market's immediate reaction to the pace of AI infrastructure investment.

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

PPhoenix Tech17:49Research

AI score73

MIT CSAIL uses GPT-6 Astra to directly rewrite the weights of Qwen3.5-0.8B

The MIT CSAIL team had OpenAI's GPT-6 Astra directly read and modify the weights of Qwen3.5-0.8B. By changing only about 0.016% of the parameters, they raised the model's score on MATH-500 from 7.9% to 48.1% and improved IFEval by about 3x, all without gradient descent or backpropagation. The team calls this method Weight-Space Programming (WSP) and points out that it removes obstacles in the process of the model exercising capabilities it already has, rather than creating new capabilities; applying the same method to Gemma 3 270M only pushed MATH-500 from 6.8% to 7.7%. The experiment has so far been released only as a blog post and has not yet produced a peer-reviewed paper.

Same story · 11 reports

Why read thisThe original article reviews MIT CSAIL's early experiments and distinguishes the boundary between "removing obstacles" and "creating capability," so readers can judge the practical scope of applying such methods to model optimization.

#Papers/Research#Reasoning#Open-source ecosystem

SShanghai Observer17:42

AI score37

Li Jiaqi's livestream sells Tokens: 100,000 Qwen Office monthly cards sold in half an hour

On November 2, during Tmall's "Double 11" shopping festival, Li Jiaqi's livestream sold AI computing power products for the first time, selling 100,000 Qwen Office individual membership monthly cards in half an hour, totaling about 2 trillion Tokens, at a unit price of just 1 cent. A report by the China Academy of Information and Communications Technology shows that in June this year, domestic daily average token consumption was about 500 trillion, nearly four times that of December last year, and the normalized average price rose from 6.8 yuan to 13.46 yuan. The article argues that what the livestream is selling is low-cost customer acquisition and the psychological comfort of "already being on board," rather than scarce computing power.

Same story · 7 reports

#English text is being prepared.#Industry news

SShanghai Observer17:42

AI score22

China Eastern Airlines APP's "Yanbao Intelligent Agent" gets a refreshed upgrade, with digital and intelligent services covering the entire travel process

The intelligent travel assistant "Yanbao Intelligent Agent" on the China Eastern Airlines APP has completed a refreshed upgrade. Travelers can handle the full chain of needs—including itinerary planning, flight booking, airport waiting services, in-journey inquiries, and benefits usage—in a one-stop manner within the same conversational interface. After the upgrade, hotel and attraction ticket booking have been added, allowing free combination of flight tickets, accommodation, and sightseeing products. Services such as VIP lounges, expedited security screening, and airport parking can also be booked directly through intelligent conversation. From now until December 31, booking domestic all-route flight tickets through the "Yanbao Intelligent Agent" offers a direct discount of up to 66 yuan.

#Product updates#Agents

SShanghai Observer17:42

AI score42

Li Jiaqi's livestream sells about 2 trillion Tokens in half an hour; Qwen Office membership card becomes a Double 11 hit

On the evening of November 2, Li Jiaqi's livestream began selling the Qwen Office personal membership monthly card, selling 100,000 units in half an hour, totaling about 2 trillion Tokens, marking the first time an e-commerce livestream has sold AI computing power products. This Double 11, Tmall for the first time made AI Tokens an independent category, and manufacturers such as Zhipu, Kimi, MiniMax, and DeepSeek have already joined the Token recharge center, allowing consumers to purchase monthly card packages just like topping up phone credit. As of July this year, China's average daily Token calls exceeded 500 trillion, an increase of more than 5,000 times compared with 100 billion at the beginning of 2024.

Same story · 7 reports

#English text is being prepared.#Industry news

4404 Media17:42

AI score53

AI scrapes AO3 fanworks to build a site for sale, authors send 2,000 takedown requests in one day

A website called ao3.life, which used AI agents to scrape AO3 fanworks and convert them into paid audiobooks, shut down one day after launching; the creator said it received more than 2,000 takedown requests from authors that day. The site claimed it used a group of AI agents controlling 40 virtual machines to scrape works from the internet, offered the first 3 minutes free, then charged for AI-generated audiobooks, and claimed authors could receive 91% of the revenue. The creator described himself as a 17-year-old developer from India and said the site had 126,592 unique users within 24 hours, but actually sold only about 11 audiobooks.

Why read thisThe incident presents the full chain of AI scraping fanworks and reselling audio versions, and also exposes the failure of authors' rights-protection methods during a platform outage.

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

Tt3n17:37

AI score67

OpenAI claims AI solved nearly 400 math problems; mathematicians criticize it for prioritizing results over process

OpenAI claims that with the help of an as-yet-unreleased AI model, it published more than 700 manuscripts in October, covering more than 370 math problems, with each taking an average of three hours, including work related to the Riemann Hypothesis. Institutions such as the German Mathematicians Association criticized this approach as a results-oriented "power display," arguing that the value of mathematical research lies in the proof path and new methods rather than the answers themselves, and expressing concern about the use of unverified knowledge and issues of academic norms. Stuttgart professor Christian Hesse, on the one hand, called OpenAI's demonstration "stunning," and on the other hand, believes that doctoral training and the definition of mathematical understanding need to be readjusted.

Why read thisRegarding OpenAI's use of AI to solve hundreds of math problems, this presents the math community's doubts about results-oriented research, authorship norms, and academic inheritance.

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

Tt3n17:37

AI score60

Anthropic's internal AI usage spending exceeds labor costs for the first time

Anthropic has published data for the third quarter of 2026, stating that since January 2026 its internal AI usage has grown 250-fold, and that at list prices, internal AI usage spending has exceeded all labor costs—though this comparison excludes stock-based compensation. The cost is calculated at external list prices rather than the actual cost of self-used infrastructure, and Anthropic has not separately disclosed the total spending on internal AI usage. Anthropic also says that in September 2026, 90% of contributions to its codebase were completed by AI, per-developer code output grew 45-fold compared with 2024, and headcount over the same period grew more than fivefold compared with the end of 2024.

Same story · 5 reports

Why read thisThe original article breaks down how Anthropic calculates its internal AI costs, so readers can judge the actual meaning of this comparison figure.

#Industry news#Anthropic

336Kr · AI17:29

AI score20

Anthropic Fable 5.5 rumor: 14 trillion parameters, animation generated purely from code, to replace video and image models within 6 months?

A rumored Anthropic model, Fable 5.5, has had real-world tests leak on X: by autonomously writing Python code to generate a single HTML file, it played an animated short in the browser, and some users say it autonomously completed a 66-second short film in 2 hours and 40 minutes. Rumors say it belongs to the Mythos tier, has a parameter scale of 14 trillion, and uses a hybrid architecture combining GDN with MLA, but the weight leak on Hugging Face has not yet been independently verified. The article also mentions that on October 13, LiteLLM briefly showed a claude-fable-5-5 entry in the Azure catalog, which was withdrawn the next day, and that the words Claude Mythos 5.5 appeared in the source code of Anthropic's official website.

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

336Kr · AI17:29

AI score22

SheNicest Women's Hackathon: Projects such as the toilet-finding app TOILER and the cognitive observation tool Jianwei showcased

At the "SheNicest Fission" hackathon, where women made up 70% of participants, teams presented products including the toilet-finding app TOILER, "Jianwei," a home cognitive health observation tool for adult children, and "Hemian," a smart mattress for late pregnancy. TOILER can filter toilets by conditions such as stall doors, hooks, and hand dryers; it currently includes only some toilets in areas such as Phase II of the China National Convention Center, 798, Guomao, and Sanlitun, and does not yet update in real time. Most projects are still in the prototype stage; for example, the "Taxian" input method cannot display the corresponding Chinese characters when more than four characters are entered.

#English text is being prepared.#Agents#Speech

336Kr · AI17:29Research

AI score63

MIT CSAIL proposes weight-space programming: GPT-6 Astra directly rewrites neural network weights

The MIT CSAIL team proposes a weight-space programming method that lets GPT-6 Astra directly read and modify neural network weights without training or computing gradients. On Qwen3.5-0.8B, Astra modifies only about 0.016% of the weights, raising MATH-500 from 7.9% to 48.1%, GSM8K from 13.8% to 67.1%, and IFEval from 21.0% to 62.3%, mainly because it fixed the model's failure of getting stuck in reasoning loops. The research was led by doctoral students Rikiya Takehi and Jayden Teoh, with participation from Omar Khattab and others, and the results were published on the research blog "Sparks of Weight-Space Programming"; additional experiments on Gemma 3 270M yielded improvements of no more than about 1 percentage point, and the team says more independent replications are still needed.

Same story · 11 reports

Why read thisThe original text provides concrete experimental data for weight-space programming on small networks, and also explains its boundary case where it failed to stably improve capability on Gemma 3 270M.

#Papers/Research#Agents#Reasoning

336Kr · AI17:29Opinion

AI score63

AI data companies are fierce: UniPat reportedly backed by Alibaba and Tencent, valued at $2.5 billion

According to a report by 36Kr's Silicon Star Pro, UniPat, a data company founded less than a year ago, is reportedly in talks to raise about $300 million, led by Alibaba with participation from Tencent and Sequoia China, at a post-money valuation of $2.5 billion. The article says a single piece of data with user behavior trajectories can be quoted at 1,000 to 3,000 yuan, while work-scenario data with a complete operation process can sell for around 10,000 yuan, and Mercor estimates that only 2% to 5% of the data frontier labs need has been completed so far. As model vendors tighten procurement, calculate ROI and run ablations on vendor data, the industry has already seen problems such as blacklisting data vendors and data quality being hard to verify.

Why read thisFrom funding valuations and procurement quotes to quality chaos, a look at how data vendors have become a key link in model post-training and the risks involved.

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

KKling AI17:21First-party

AI score38

Kling 4.0 Hands-On Test: Eight Creators Discuss Dynamic Motions and Native 30-Second Generation

Kling 4.0 invited eight creators to share their hands-on creative experiences. Key capabilities include the physical performance of Dynamic Motions in highly difficult action sequences, more stable camera scheduling, and Native 30s Generation, which supports native 30-second generation. Omni Reference can fuse multiple visual references, and character consistency and lip-sync have both improved in long shots and Cantonese dialogue scenes.

#Industry news#Video#Multimodal

SSenseTime17:16

AI score48

SenseTime Releases SenseNova 6.8 Flash, Focusing on Office Intelligence and Dynamic Design

SenseTime Releases SenseNova 6.8 Flash, Focusing on Reliable Workflows and Design Output: Office intelligence supports long-term memory and proactive interaction, while dynamic design can generate animated PPTs, dynamic posters, interactive websites, and editable Blender 3D projects. The model is available for free on the SenseNova Token Plan, offering 1,500 free API requests every 5 hours, with the model ID SenseNova 6.8 Flash.

Same story · 4 reports

#Model releases#Agents#Multimodal

CCailian Press19:27

AI score32

DeepSeek Web and APP chat services fail again, only half a month after the last outage

The chat functions on DeepSeek's web and APP platforms began malfunctioning at 16:58 on the afternoon of the 3rd, with replies showing "server busy"; after a brief recovery, the outage recurred. The official website shows a partial web/API interruption, affecting the chat service, the DeepSeek V4 Pro API, and the DeepSeek V4.1 Flash API. As of press time, the web version has recovered, while the APP still shows a malfunction. This is the second occurrence of the same type of problem within half a month, following the outage on October 15 that lasted more than five hours.

Same story · 2 reports

#Industry news#DeepSeek

LLeitech19:21

AI score18

Lei Jun gifts Zhou Yunjie a SU7 Ultra car model; Xiaomi and Haier discuss ecosystem, AI, and car-home connectivity cooperation

On November 2, Lei Jun visited Haier headquarters and presented Zhou Yunjie with a Xiaomi SU7 Ultra car model, fulfilling the mutual visit agreement the two made during this year's Two Sessions. The two sides held in-depth discussions on ecosystem, AI innovation, and globalization. Zhou Yunjie mentioned that Haier is leading the construction of a national AI application pilot base for home and home appliance sectors, with the platform open to the entire industry.

#Industry news

CChina.com Technology18:56

AI score20

Aippy Appears at 2026 Tencent Global Digital Ecosystem Summit, NADA AI and Tencent Cloud Reach Strategic Cooperation

Aippy, an AI gaming community developed by the NADA AI team, demonstrated its capability to create interactive games using natural language at the 2026 Tencent Global Digital Ecosystem Summit, and completed a strategic cooperation signing with Tencent Cloud. The two parties will cooperate deeply in three major directions: access to large language models and multimodal large models, sandbox Runtime, and cloud deployment of overseas projects. Aippy launched in 2025, with total global downloads exceeding 5 million, and the number of games within the app has surpassed 8 million.

#Industry news#Agents#Multimodal

AAlibaba Cloud Developer Community AI18:33

AI score28

Data Readiness Assessment for Enterprise Agent Scenarios: Four Questions and Four Answers

Before deploying an enterprise agent, four questions can be used to judge a scenario's data readiness: can the required fields be listed; which system each field resides in and whether the definitions conflict; whether historical data covers enough months; and whether the data source can be sustained and has a data contract. The article provides a scoring model: 80 points or above means the project can be initiated directly, 50-79 points means data must be supplemented first, and below 50 points means the scenario is not viable. Data readiness must be confirmed jointly by the business side and IT; asking only one side yields an incomplete answer.

#Tutorials / practice#Agents#English text is being prepared.

AAlibaba Cloud Developer Community AI18:33

AI score15

2026 Alibaba Cloud Double 11: New customers' first order from 38 yuan, up to 1,728 yuan in vouchers

The 2026 Alibaba Cloud Cloud Double 11 event runs until November 30. New customers can grab a lightweight application server with 2 cores, 2GB of RAM, a 40GB ESSD cloud disk, and 200M peak bandwidth for as low as 38 yuan on their first order. Stock is released in limited quantities at 10:00 and 15:00 every day, and this flash-sale item cannot be combined with vouchers. Enterprise-verified users can receive up to 1,728 yuan in vouchers, and students can receive a 300 yuan no-threshold voucher after verification. Bailian Token Plan subscriptions are as low as 50% off, the Night Plan is 40% off at night, and the Wanxiang video generation model is 30% off for a limited time.

Same story · 10 reports

#English text is being prepared.

AAlibaba Cloud Developer Community AI18:33

AI score20

How to Choose Enterprise-Level Data Governance Tools: 2026 Practical Evaluation Framework and the Lingyang Dataphin Case Study

The 2026 version of Alibaba Cloud's Lingyang Dataphin shifts large model capabilities from external plug-ins to native embedding; AI assistance can reduce repetitive data development workload by more than 70% and boost R&D efficiency by more than 40%. The article presents five selection dimensions: functional completeness, technical architecture, degree of AI integration, industry implementation capability, and total cost of ownership, and states that Dataphin supports access to more than 50 types of heterogeneous data sources. After a leading retail group completed a governance closed loop, its data asset reuse rate rose from 31% to 79%, and precision marketing ROI increased by 53%.

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

AAlibaba Cloud Developer Community AI18:33Opinion

AI score59

GitHub urgently scales up by 4 million cores: AI Agents turn CI into a trial-and-error playground—how should testing reclaim the quality gate

At the GitHub Universe conference, it was disclosed that single-month commits in September reached 7.4 billion, about 5 times the same period last year; this year a cumulative 4 million CPU cores and 160PB of storage were added, and 90-day availability once dropped to about 84%. The author believes AI Agents treat the Actions pipeline as a debugger, repeatedly committing by trial and error, turning verification costs that should be borne by the committers themselves into public platform expenses. Octoverse data shows that in September over 40% of pull requests were initiated by AI agents. The article suggests that testing teams respond in four aspects: moving gates earlier, tiered triggering, Agent governance, and metric upgrades.

Why read thisUsing GitHub commit volume and availability data, the article breaks down the verification cost transfer problem caused by AI Agents treating CI as a debugger.

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

AAlibaba Cloud Developer Community AI18:33

AI score25

1D Convolution and Residual Modules: How Rolling Bearing Fault Diagnosis Accuracy Improved from 83% to 95%

In rolling bearing fault diagnosis, a single-layer 1D-CNN achieves a test accuracy of only about 83%, and inner race fault samples are often misclassified as rolling element faults; after switching to a CNN+ResNet with 5 residual blocks, the accuracy rises to 95%+. The residual connection, with its shortcut of output = F(x) + x, provides a lossless channel for gradients, solving the vanishing gradient and degradation problems of deep networks and enabling the network to distinguish inner race faults from rolling element faults after deepening. The two sets of experiments use the same data, splits, and evaluation method; the only difference is whether residuals are added.

#Tutorials / practice#English text is being prepared.

YYicai18:10

AI score44

China Southern Power Grid hosts the power industry's first embodied intelligence application competition, with robots performing tasks such as substation inspections

On November 3, the semifinals (practical competition) of the 2026 Power Embodied Intelligence Application Competition kicked off at China Southern Power Grid's Guangdong Dongguan Power Supply Bureau. This is the first embodied intelligence application competition in China's power industry, featuring three major assessment events: substation equipment inspection, emergency power outage operations, and voltage testing before electric meter replacement. Ultimately, 6 teams were selected to advance to the finals. The competition was launched in September this year, with a total of 140 units registering and forming 93 participating teams; 15 teams advanced to the offline practical competition. In April this year, State Grid issued the "2026 Embodied Intelligence Development Plan," planning to procure 8,500 embodied intelligence devices with a total investment of approximately 6.8 billion yuan.

#Industry news#Embodied AI

CChina.com Technology17:55

AI score22

ESI 20th Anniversary: Mao Daqun Discusses Data Governance in the AI Era and the Ruizhi OPT Agent Architecture

At the 20th anniversary customer conference, ESI General Manager Mao Daqun proposed the "AI and data double helix" concept and released the Ruizhi OPT agent work architecture, in which data governance is completed by one person plus a group of silicon-based experts plus N sets of reusable Skills. Under this architecture, governance work that originally required a team six months to a year can be completed by one person, master data project implementation cycles are reduced by 50%, manual work drops by 70%, and AI programming overall efficiency improves by about 3 times. ESI has served more than 13,000 customers, and IDC has ranked it first in China's data governance solutions market for five consecutive years (2021–2025).

Same story · 3 reports

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

CChina.com Technology17:55

AI score40

Yixin Huachen Releases Ruizhi OPT Multimodal Data Governance Platform, Adopting an AI-Native Architecture

Yixin Huachen released the Ruizhi OPT multimodal data governance platform on October 30, 2026. It adopts an AI-native architecture and comes built in with more than 30 data governance expert agents, over 50 governance skills, and more than 100 knowledge bases, and encapsulates governance capabilities such as metadata, data standards, and data quality as a CLI. The platform adopts a four-layer architecture and provides three working modes: fully automated execution, human-machine collaboration, and manual revision. Multimodal governance covers six stages from collection and aggregation to scenario-based delivery.

Same story · 3 reports

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

CChina.com Technology17:55

AI score35

Zhongtan Taifeng's steel energy closed-loop project lands, validating cross-industry methodology for industrial AI

Zhongtan Taifeng disclosed an energy medium balance and comprehensive cost optimization project implemented at a large steel enterprise. After commissioning, the gas flaring rate dropped from 4.9% to below 2%, with annualized comprehensive cost savings exceeding 100 million yuan. The project is organized as a closed loop of "process prediction—dispatch recommendations—effect review," with large models responsible for understanding dispatch language and process knowledge, domain models undertaking deterministic calculations, and actions involving production safety and key dispatch boundaries retaining manual confirmation and rollback mechanisms. Zhongtan Taifeng emphasized that the annualized savings figure is affected by capacity utilization, medium prices, and statistical boundaries; what is replicable is the mechanism for locating, explaining, and verifying cost anomalies.

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

AAnt inclusionAI: New GitHub Repository17:49First-party

AI score46

Ant inclusionAI Open-Sources dInfer: An Efficient Inference Framework for Diffusion Language Models

Ant inclusionAI has open-sourced dInfer, an efficient inference framework for diffusion language models (dLLMs) that splits inference into four modules—the model, the diffusion iteration manager, the decoder, and the KV-cache manager—and supports batched inference. It supports the LLaDA, LLaDA-MoE, and LLaDA2 series; on a single 8×H800 node, LLaDA-MoE exceeds 1,100 TPS on HumanEval at batch size 1 and averages 800+ TPS across six benchmarks, 10× faster than Fast-dLLM while maintaining accuracy.

#English text is being prepared.#Reasoning#Open-source ecosystem

JJiwei.com17:21

AI score18

The China Electronics Standardization Association held a health measurement standards seminar in Shenzhen, discussing two group standards: "Bioelectrical Impedance Chip" and "Four-Electrode Body Fat Scale"

The Home Health Auxiliary Management Device Branch of the China Electronics Standardization Association held a "Chip Health · Smart Future" health measurement standards seminar in Shenzhen, with nearly 40 representatives from industry, academia, research and application attending. The seminar discussed the draft group standards for "Bioelectrical Impedance Chip" and "Four-Electrode Body Fat Scale," and solicited standardization needs for health electronic devices from the industry. The drafting groups for the two standards, represented by Chipsea Technology and Kangyou Health respectively, presented their reports. The former specifies technical indicators such as four-electrode measurement, the dynamic range and measurement error of impedance measurement, and common-mode rejection ratio; the latter standardizes metrological performance, impedance measurement performance and body fat percentage measurement performance, and classifies products into single-frequency, dual-frequency and multi-frequency according to measurement frequency.

#Industry news#English text is being prepared.

JJiwei.com17:21

AI score45

CXL Specification Advances to 4.0, QLC Orders Already Booked Through 2028: The Two Storage Roadmaps Diverge in Pace

The CXL Consortium has released the 4.0 specification, doubling link speeds from 64 GT/s to 128 GT/s, but Intel's Diamond Rapids platform supporting CXL 3.x has been delayed to a 2027 launch, and Samsung, SK hynix, and Micron have successively halted or scaled back their in-house CXL controllers. On the QLC side, Micron's 245TB 6600 ION has officially shipped, with per-drive capacity roughly quadrupling from 61.44TB over three years, and Seagate's nearline drive shipment allocations are already booked through 2028.

#Industry news#English text is being prepared.

JJiwei.com17:21

AI score13

Wanrun Technology Q3 Report: Non-GAAP Net Profit Up 214.72%, Storage Business Rides AI Boom

Wanrun Technology's revenue for the first three quarters of 2026 was 3.383 billion yuan, down 8.90% year-on-year, while non-GAAP net profit was 27.1435 million yuan, up 214.72% year-on-year, marking a clear recovery in the profitability quality of its core business. Revenue from the semiconductor memory business in the first half of the year was 490 million yuan, up 18.06% year-on-year, with gross margin rising to 9.81%; the company's inventory at the end of the period increased to 365 million yuan for stockpiling, and it completed a 130 million yuan capital increase in Wanrun Semiconductor with iSoftStone. AI servers' demand for DRAM is 8 to 10 times that of ordinary servers, and SEMI China expects the global HBM market to grow 58% to 54.6 billion US dollars in 2026.

#English text is being prepared.

CCailian Press17:10

AI score42

Goldman Sachs CEO Solomon: Compute Demand Far Exceeds Supply, AI Financing Will Remain Strong Over the Next 12 to 24 Months

Goldman Sachs Chairman and CEO David Solomon said that compute demand far exceeds supply, and that strong financing activity over the next 12 to 24 months will meet part of that demand, with investor demand for subscriptions to large AI financing deals being "very strong." He believes that AI infrastructure construction is still at an early stage, but rising financing costs will change the returns on some projects; Goldman Sachs has secured the lead underwriting role for Anthropic's IPO and is co-leading Meta's $95 billion equity financing plan.

#Industry news#English text is being prepared.

YYicai17:08Opinion

AI score58

Is AI Financing About to Hit a Minsky Moment?

Li Wei, a professor at Cheung Kong Graduate School of Business, analyzes AI infrastructure financing risks, noting that Anthropic finalized approximately $42 billion in syndicated loans and about $18 billion in subordinated debt through the "Project Polaris," with Broadcom providing a guarantee and Blackstone contributing half of the funding. The article also uses Meta's Hyperion data center as an example, explaining that it moved $27 billion in debt off its balance sheet through an SPV, with a debt-to-asset ratio of about 90%, higher than Meta's own ratio of about 43%. The author believes there is insufficient evidence to assert that a Minsky moment is imminent, but signs such as rising issuance of junk-rated data center bonds and falling prices of some bonds are worth watching.

Why read thisThe article uses Anthropic's debt financing and Meta's off-balance-sheet data center structure as examples to outline the leverage and maturity mismatch risks in AI infrastructure financing.

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

YYicai17:08

AI score30

From Missing Out on AI to AI Overweight: The Pricing Lesson of STAR 50's 47% Drawdown in Four Months

The STAR 50 fell from 2208 points at the end of June to 1161 points at the end of October, a drawdown of about 47% in four months, while Cambricon's revenue for the first three quarters was 9.526 billion yuan, up 106.76% year-on-year, yet its stock price still fell 36.8% in the third quarter. The author Xue Hong believes the industry logic has not changed; what has changed is that the market has shifted from the income statement to the cash flow statement, from growth rate to duration, and capital expenditure has turned from orders on the numerator side into liabilities on the denominator side. He suggests first looking at who ultimately pays and how far this "right" has already been priced in, and then looking at whether the account's time horizon can wait for it to be realized.

#Industry news