Newshot

Morning Brief | 黄仁勋 says willing to return to 英伟达/李飞飞 joins AMD, reporting directly to 苏姿丰/Hongguo Short Drama tackles portrait infringement

AI summary

黄仁勋 speaks out for the first time since resigning: if the board approves, willing to return to 英伟达 as CEO. 英伟达's quarterly revenue exceeds $100 billion for the first time, with data center revenue up 110%. AMD completes acquisition of World Labs, with 李飞飞 serving as Executive Vice President and Chief Scientist. OpenAI releases GPT-6.2 Astra and Sol, with Astra API prices cut by 20%. Apple iPhone Duo has over 200,000 compatible apps, more than doubling in a week. macOS 27.3 tightens app file access permissions. Hongguo Short Drama tackles portrait infringement, having handled 43,992 suspected problem works. Spotify opens profile protection to all artists, allowing pre-review of new songs under their names. Apple reportedly plans to sign a 4- to 5-year DRAM long-term agreement with Samsung, with final negotiations underway. Automakers' 60-day payment period enters annual research and evaluation, and intentional payment delays will lead to talks. Civilization developer Firaxis lays off about one-third of staff, studio head to leave. Kunlun Chip M100 expected to meet mass production conditions by year-end, Baidu AI cloud revenue up 94%. Lenovo's PC and tablet market sales share surpass

Translation(model translation, not the source's own text)

Article title: Morning Brief | 黄仁勋 Says He Is Willing to Return to 英伟达/李飞飞 Joins AMD, Reporting Directly to 苏姿丰/Hongguo Short Dramas Addresses Portrait Infringement

Body:

Morning Brief | 黄仁勋 Says He Is Willing to Return to 英伟达/李飞飞 Joins AMD, Reporting Directly to 苏姿丰/Hongguo Short Dramas Addresses Portrait Infringement

2026-11-18 08:32

💾 黄仁勋 Speaks Out for the First Time After Resigning: If the Board Approves, He Is Willing to Return to 英伟达 as CEO

📈 英伟达 Quarterly Revenue Tops $100 Billion for the First Time, Data Center Revenue Up 110%

💰 AMD Completes Acquisition of World Labs, 李飞飞 Appointed Executive Vice President and Chief Scientist

🏷️ OpenAI Releases GPT-6.2 Astra and Sol, Astra API Price Cut by 20%

📱 Over 200,000 Apps Adapted for Apple iPhone Duo, More Than Doubling in a Week

💻 macOS 27.3 Tightens App File Access Permissions

🎬 Hongguo Short Dramas Addresses Portrait Infringement, 43,992 Suspected Problematic Works Handled

🎵 Spotify Opens Profile Protection to All Artists, Allowing Pre-Review of New Songs Under Their Names

💬 Apple Reportedly Plans 4- to 5-Year DRAM Long-Term Agreement with Samsung, Final Negotiations Underway

🚗 Automakers' 60-Day Payment Terms Enter Annual Research and Evaluation, Intentional Payment Delays Will Trigger Talks

📉 Civilization Developer Firaxis Lays Off About One-Third of Staff, Studio Head to Depart

📈 Kunlunxin M100 Expected to Meet Mass Production Conditions by Year-End, Baidu AI Cloud Revenue Up 94%

💻 Lenovo PC and Tablet Market Sales Share Surpasses Apple, IDC Estimates $19.5 Billion

🤖 何小鹏 Expected to See November Sales Exceed 60,000 Units for the First Time; Humanoid Robots to Launch for Commercial Customers in Q2 Next Year

✨ NBA 2K27 to Collaborate with GTA 6, Vice City-Themed Court Launching November 27

🏢 王兴兴: Robots Need Retraining When Scenarios Change, Generalization Capability Is the Primary Challenge

🧠 Shing-Tung Yau: Following AI Conclusions Completely Will Cause Us to Lose the Ability to Ask Questions Independently

💼 Voyah Obtains Independent Production Qualification for Road Motor Vehicles

🏛️ MIIT Approves 6G Innovation Pilots in Eight Provinces and Municipalities Including Beijing and Shanghai

💾 AMD Chips Enter 202 TOP500 Supercomputers, Lux Debuts at No. 14

🏢 英伟达 Establishes Dedicated Board Committee to Oversee National Security and Chip Export Compliance

🎬 Hongguo Short Dramas Recruiting TV Product Manager, Plans to Build Large-Screen Product from Scratch

💵 哈萨比斯's Isomorphic Labs Raises $5 Billion to Advance AI Drug Development

📈 JD.com Q3 Revenue Up 3.5%, New Business Operating Loss Narrows

🧠 Lei Jun: AI Adoption Scenarios Still Limited, Few Companies Actually Making Money

💡 李彦宏: If AI Applications Cannot Boost Revenue and Profit, Bubble Risk Will Intensify

🎧 ¥359, Sony MDR-3R Launches in China, Recreating the Original Walkman Headphones

🎮 任天堂 eShop Adds Family Group Gifting, Allowing Gifts of Games, DLC and Upgrade Packs

🪄 ¥519,900, Dongfeng Mengshi M817 Ultimate Launched

✨ Huawei Starlight OSTA O1 Pro Opens Reservations, First Sale on Evening of November 20

🔓 SenseTime Open-Sources Loop-9B Official Version, Reusing Model Layers to Expand Reasoning Depth

🪟 Microsoft MAI Code 2 Launches on Copilot, Simultaneously Expanding Image, Voice and Security Models

🧰 VSCO's First Native Mac Photo Editing App Launches, Can Batch Edit 500 Photos

💬 Xiaohongshu Tests "Dada" App Internally, Users Can Create AI Companions and Add Them to Group Chats

🪄 INMO Future Opens Reservations, Featuring Binocular Full-Color Display and Spatial Interaction

📷 DJI Price Adjustments Take Effect Today, Osmo Pocket 4P Standard Kit Now ¥4,099

✨ WorkBuddy Supports Cross-Device Migration, Can Import Conversations, Artifacts and Scheduled Tasks

🎮 TGA 2026 Nominations Announced, GTA 6 Leads, Phantom Blade Zero Competes for Game of the Year

🛍️ Master Kong Uses Doubao Work Platform to Track Noodle Cake Loss Across 77 Production Lines

🔌 Two Departments Require Popularization of "One-Click TV Watching," Crack Down on Illegal Pharmaceutical Ads Targeting Seniors

👀 Avengers 5 Shanghai Fan Event Set for December 10, Downey and Other Creatives to Visit China

🎬 Netflix's Three-Body Problem Season 2 Debuts First Trailer, Launching January 14 Next Year

🎨 Call Me Bae Season 2 Set for December 17, Trailer Released Simultaneously

Top Stories
Top Stories

黄仁勋 Speaks Out for the First Time After Resigning: If the Board Approves, He Is Willing to Return to 英伟达 as CEO

黄仁勋 published a long post on social media, publicly responding for the first time to calls from outside for him to return to 英伟达. He said that if the board believes his return is in the best interests of shareholders, employees, the company and the United States, he is willing to serve as 英伟达 CEO again.

In the statement, 黄仁勋 said he did not meet 英伟达's behavioral requirements for every employee and made serious mistakes in his personal life that also created problems for the board.

He said he would take full responsibility and cooperate with the board's ongoing independent investigation. He also mentioned that after leaving 英伟达, he received widespread support from employees, investors, business partners and customers, and that many people contacted him directly as 英伟达 faces complex technological, market, competitive and geopolitical situations. In the statement, he did not ask the board to reinstate him, but left the decision to the board.

However, 英伟达 Board Chairman 哈维·琼斯 said on the earnings call that the company does not plan to have 黄仁勋 return.

哈维·琼斯 said, we hope to appoint a permanent CEO as soon as possible, but we will not rush the decision just to save time; we must first find the most suitable candidate.

🔗 Related reading: Just now, 黄仁勋 speaks out for the first time after resigning: willing to return to 英伟达 as CEO

Big Companies
Big Companies

英伟达 Quarterly Revenue Tops $100 Billion for the First Time, Data Center Revenue Up 110%

Image
Image

英伟达 reported results for the third quarter of fiscal 2027 ended October 25, with revenue of $115.8 billion, up 103% year over year and 20% quarter over quarter, exceeding $100 billion in a single quarter for the first time. GAAP net income was approximately $65.3 billion, up 105% year over year.

  • Data center revenue was $107.6 billion, up 110% year over year; edge computing revenue, including the consumer gaming PC business, was $8.2 billion, up 42% year over year.
  • GAAP and non-GAAP gross margins were 74.8% and 74.9%, respectively; GAAP diluted earnings per share were $2.70, and non-GAAP were $2.68.
  • Free cash flow was $43.777 billion; $28.1 billion was returned to shareholders through buybacks and cash dividends.

Interim CEO, Executive Vice President and CFO 科莱特·克雷斯 said Vera Rubin is seeing strong early growth and the product roadmap is on track. The company said Vera Rubin NVL72 has begun shipping and is running at partners including CoreWeave, Google Cloud and Microsoft Azure.

英伟达 expects fourth-quarter revenue of $130 billion, plus or minus 3%. The quarter has one more week than usual, and the guidance does not include data center computing revenue from China.

AMD Completes Acquisition of World Labs, 李飞飞 Appointed Executive Vice President and Chief Scientist

Image
Image

AMD announced on November 16 that it has completed its acquisition of AI research company World Labs. World Labs founder 李飞飞 has joined AMD as Executive Vice President and Chief Scientist, reporting directly to AMD Chairman and CEO 苏姿丰.

World Labs develops AI models capable of creating and simulating interactive 3D environments, and also researches robot learning and simulation technologies. After the acquisition, the team will continue fundamental model research and work with AMD engineers to apply research results on new model capabilities to the development of future AI infrastructure.

苏姿丰 said that providing infrastructure for the next phase of superintelligence requires optimizing both computing platforms and models. AMD hopes to use the World Labs team's model research to participate in the design of next-generation hardware, software and systems, serving customers and developers in the open AI ecosystem.

OpenAI Releases GPT-6.2 Astra and Sol, Astra API Price Cut by 20%

OpenAI released GPT-6.2 Astra and Sol on November 17, aimed at complex professional projects and daily work, respectively. Both models are available to Plus, Pro, Enterprise and Edu users of ChatGPT Work and Codex; enterprise and education workspaces require administrators to enable the corresponding access permissions.

Astra's updates focus on producing documents, spreadsheets and slides according to industry conventions, as well as code, webpages and 3D design. OpenAI said that on Agents' Last Exam v1.1, which tests multi-step professional workflows, Astra improved by 5.2 percentage points over the previous generation; on the company's internal GDPval v2 test, it improved by 6.2 percentage points while using 22% fewer output tokens.

Sol focuses on efficiency. OpenAI said that on the FrontierCode 2.0 test, which checks the quality of modifications to real codebases, when both models use the highest reasoning settings, Sol's score was within 0.5 percentage points of Astra, at about one-fifth the cost per task.

On the API side, Astra is priced at $8 per million input tokens and $40 per million output tokens, 20% lower than the previous generation; Sol keeps the previous generation's price of $2 per million input tokens and $10 per million output tokens. Pro 500 users can use Ultrafast with both models, and Pro 200 users are getting the mode for the first time, starting with Sol.

Article title: Morning Brief | 黄仁勋 Says He Is Willing to Return to 英伟达/李飞飞 Joins AMD, Reporting Directly to 苏姿丰/Hongguo Short Dramas Cracks Down on Portrait Infringement

Body: Both models' cybersecurity capabilities reach the Critical threshold of OpenAI's Preparedness Framework. The company has added restrictions for high-risk cybersecurity requests and plans to introduce the new models into Daybreak, which is aimed at vetted security researchers.

Apple's iPhone Duo Has Over 200,000 Compatible Apps, More Than Doubling in Just Over a Week

Image
Image

Recently, Apple's Senior Vice President of Worldwide Marketing Greg Joswiak announced in a post on X that more than 200,000 apps have now been optimized for the iPhone Duo, a figure that doubled in just over a week.

Apple previously disclosed that just over two weeks after the iPhone Duo went on sale, more than 100,000 apps on the App Store had been built specifically for the foldable phone.

Greg Joswiak said he was excited by how quickly developers had embraced the iPhone Duo and looked forward to seeing the new apps they release next.

Currently, a number of commonly used apps in China have been adapted for the iPhone Duo, including Kimi, Baidu Maps, DiDi, Youku Video, as well as Goodnotes, MarginNote 5, Tide and other apps, covering scenarios such as AI assistants, travel, video and entertainment, and study and productivity.

macOS 27.3 Tightens App File Access Permissions

Image
Image

Apple has further tightened app permission management in macOS 27.3 Beta 2. For apps requesting "Full Disk Access" or "System and App Protection Coverage" permissions, the system will show users a red warning: after authorization, the app may read, modify, delete and export personal data, view hidden photos, and even browse websites as the user. Users must confirm the risks item by item before they can continue granting authorization.

The new version of the system also allows users to open only specified folders to an app, or limit the apps it can control, and can be set to ask again before every operation. When a third-party tool controls other apps in the background, the system will also display a corresponding prompt. Apple says users can adjust or remove the relevant permissions at any time in "Privacy & Security" in "System Settings."

Apple previously warned developers that some apps are using the "Full Disk Access" permission in ways that may endanger user data, and specifically mentioned the risks brought by the growing autonomous capabilities of AI agents.

🔗 Related reading: Mac tightens permissions, Apple starts guarding against AI

Hongguo Short Dramas Cracks Down on Portrait Infringement, Has Handled 43,992 Suspected Problematic Works

Image
Image

Hongguo Short Dramas announced special portrait infringement governance measures on November 17, covering platform self-review, pre-production checks and infringement complaints. The platform said that as of October this year, it had cumulatively intercepted, removed and fixed 43,992 works suspected of portrait infringement, 99% of which were proactively intercepted by the platform in advance.

The pre-production check tool launched in August this year has been opened free of charge to creators and supports detecting and adjusting problematic materials during the material production stage. Hongguo said the tool has cumulatively checked nearly 800,000 material cases and intercepted more than 90,000 non-compliant materials in advance.

Since April this year, the platform has cumulatively received more than 1,500 reports and complaints of portrait infringement, and after review, about 60% involved relatively clear infringement problems. Users can select "Report" in the menu in the upper right corner of a work's playback page to provide relevant clues.

Hongguo said some creators use face-merging technology or modify original portrait materials, causing some advance interceptions to fail, and the platform will continue to improve its technical interception capabilities and improve complaint handling.

Spotify Opens Profile Protection to All Artists, Allowing Them to Review New Songs Under Their Names First

Image
Image

Spotify is opening its "Artist Profile Protection" tool to all artists. The tool previously underwent about eight months of testing to address issues such as impersonation uploads, release metadata errors and works by artists with the same name being misattributed.

According to TechCrunch, artists can enable protection in Spotify for Artists and receive alerts about new works submitted under their names. If a work does not belong to them, they can unlink it so it no longer appears on their profile.

Under the default setting, works are still released at the originally scheduled time, and artists need to intervene proactively. Artists who want to vet releases in advance can enable advanced settings, requiring each release to be approved or rejected by them before it appears on their profile. Spotify said thousands of artists participated in improving this process during testing.

Apple Reportedly Plans 4- to 5-Year DRAM Supply Agreement with Samsung, in Final Negotiations

Image
Image

According to a November 17 report by South Korea's Maeil Business Newspaper, Apple and Samsung Electronics are in final negotiations on a long-term DRAM supply agreement. The agreement under discussion is planned to last 4 to 5 years starting in the first half of next year, set a minimum supply volume, and have prices ultimately determined based on future market conditions.

The report, citing semiconductor industry insiders, said Apple typically used short-term contracts in the past and sought price cuts when renewing them. This time it is seeking long-term supply to secure the DRAM needed for core products such as iPhone and MacBook amid tightening memory chip supply.

Maeil Business Newspaper also said Samsung Electronics, SK Hynix and Micron have successively signed long-term supply agreements with major customers, and Apple recently reached a long-term agreement with SK Hynix as well, accepting terms more favorable to the memory supplier.

Automakers' 60-Day Payment Terms Enter Annual Survey and Assessment; Intentional Payment Delays Will Lead to Talks

Image
Image

The "60-day payment terms" promised by automakers will enter an annual survey and assessment. The China Association of Automobile Manufacturers issued a notice on November 16 stating that, entrusted by the Ministry of Industry and Information Technology, it will assess the supplier payment situation of 17 key automobile companies.

The assessment adopts a method of "regular reporting by automakers + spot checks of sample suppliers," randomly selecting no fewer than 15 suppliers from each automaker's suppliers each year, and confirming payment situations through document verification and on-site surveys. Automakers must submit the previous year's report by January 31 each year and the report for the first half of the current year by July 31.

The verification covers whether payment terms for small and medium-sized enterprises exceed 60 days, cash payment conditions, and whether payment terms for non-small and medium-sized enterprises meet the 60-day requirement and whether the proportion of cash and bank acceptance bill payments is no less than 80%. After the assessment results are submitted to the Ministry of Industry and Information Technology, they will be made public according to procedures.

According to Cailian Press, a relevant official from the Ministry of Industry and Information Technology said that for automakers with large accounts payable, intentional lengthening of payment terms or more complaints, talks will be conducted jointly with relevant departments and rectification will be urged.

Civilization Developer Firaxis Lays Off About One-Third of Staff, Studio Head to Leave

Image
Image

According to Bloomberg, Firaxis, the developer of the Civilization and XCOM series, cut about 60 positions on November 17, accounting for about one-third of its total employees. 2K Games, under parent company Take-Two, also confirmed that studio head Heather Hazen will leave.

In a statement provided to Ars Technica, Firaxis said it is restructuring and refocusing on its core strength in strategy games, reducing the size of its team, and focusing on design, technology and player experience.

Kunlunxin M100 Expected to Be Ready for Mass Production by Year-End; Baidu AI Cloud Revenue Grows 94%

Baidu disclosed in its third-quarter earnings announcement that the Kunlunxin M100, optimized for large-scale AI inference, is expected to be ready for mass production before the end of this year. The chip was first publicly unveiled at the World Artificial Intelligence Conference in July this year.

The financial report published on November 16 shows that Baidu's third-quarter revenue was about 30.8 billion yuan, down 1% year on year; under GAAP, net profit attributable to Baidu was about 1.4 billion yuan, and non-GAAP net profit was about 2.4 billion yuan.

In terms of business structure, intelligent cloud infrastructure revenue was 8.2 billion yuan, up 94% year on year, of which GPU cloud revenue rose 191% year on year; core AI new business revenue was 13.2 billion yuan, accounting for 53% of Baidu's general business revenue. Online marketing services revenue was about 11.9 billion yuan, down 22% year on year.

Baidu's third-quarter operating cash flow was positive, reaching about 2.8 billion yuan. The company also said in the financial report that token usage revenue from external customers on the Qianfan MaaS platform rose 1159% year on year.

Lenovo's PC and Tablet Market Sales Share Surpasses Apple, IDC Estimates $19.5 Billion

Image
Image

IDC's Worldwide Quarterly Personal Computing Device Tracker for the third quarter of 2026, provided by Lenovo, shows that in the market including desktops, notebooks, workstations and tablets, Lenovo's sales share was 21.4% and its shipment share was 20.3%, ranking first in both.

Based on IDC-estimated end-user spending, Lenovo's sales in the quarter were $19.5 billion, up 7.4% year on year; Apple's were $19 billion, with a sales share of 20.9%, ranking second. Lenovo's sales share rose 1.3 percentage points compared with the same period last year.

In terms of shipments, Lenovo shipped 19.1 million units and Apple 17.2 million units. Global personal computing device shipments in the same period were about 94.1 million units, down 19.3% year on year.

何小鹏 Expected to See Monthly Sales Exceed 60,000 Units for the First Time in November; Humanoid Robots to Be Released to Commercial Customers in Q2 Next Year

Values written like {{NAME_ab12cd34}} are placeholders for real values. Use them exactly as written, in replies and in tool calls alike: they are filled in with the real values before anything runs. Don't ask to verify them, and don't replace them with examples.

Article title: Morning Brief | 黄仁勋 Says He Is Willing to Return to 英伟达/李飞飞 Joins AMD, Reports Directly to 苏姿丰/Hongguo Short Dramas Addresses Portrait Infringement

Body:

Image
Image

According to Yicai, 何小鹏 said on XPeng Group's third-quarter earnings call that November sales are expected to exceed 60,000 vehicles for the first time. The company's fourth-quarter delivery guidance is 169,000 to 174,000 vehicles; third-quarter actual deliveries were 118,390 vehicles, a quarterly record.

On the overseas business side, XPeng delivered 27,882 vehicles in the third quarter, with overseas deliveries exceeding 13,000 vehicles in October. The company's fourth-quarter international business sales target is more than 40,000 vehicles, and it expects full-year overseas sales to exceed 100,000 vehicles for the first time; according to plan, XPeng NGP will gradually be delivered to global customers in 2027.

何小鹏 also updated the humanoid robot plan: the first batch of mass-produced robots is planned to enter some XPeng stores for trials in the first quarter of next year, and IRON will be officially released to external commercial customers in the second quarter of 2027. He said that next year he will devote more energy to improving operations and strengthening the automotive business's self-sustaining capacity, while advancing robot mass production and delivery.

"NBA 2K27" Will Collaborate with "GTA 6", Vice City-Themed Court to Launch November 27

Image
Image

2K announced the third-season collaboration for "NBA 2K27", bringing the scene style of "GTA 6"'s Vice City into the basketball game. The season is called "Vice City Takeover", launching November 27 and running for seven weeks until January 15 next year.

The City will add a Vice City-themed basketball park, featuring beachside courts, neon signs, pastel hotels, and a statue of the alligator mascot Macca. Players can also obtain Vice City- and Los Santos-themed jerseys, as well as clothing from "GTA" fictional brands such as Burger Shot and Cluckin' Bell.

MyTEAM will also add Vice City-themed cards, limited-time challenges, and season rewards. 2K is launching a digital "NBA 2K27 Vice City Edition" at the same time, including the third-season Pro Pass and collaboration-themed items.

From November 27 to January 15 next year, GTA+ members on PS5 and Xbox Series X|S can play the standard edition of "NBA 2K27" for free; starting December 17, they can also claim exclusive collaboration items.

王兴兴: Robots Need Retraining When the Scene Changes, Generalization Capability Is the Primary Challenge

Image
Image

According to "Science and Technology Innovation Board Daily", 王兴兴 said at the APEC CEO Summit on November 17 that the results publicly demonstrated by AI and robots are only the tip of the iceberg, and the underlying technology is still iterating rapidly; however, to make robots adapt to different home and work environments, the first problem to solve is the insufficient generalization capability of models.

He explained that large language models have already played an increasingly valuable role in personal life and office work, raising people's expectations for technology applications, and people therefore hope that robots can also use AI to truly work and solve practical problems at work.

He pointed out that robots often need retraining when changing application scenarios, making it difficult for a single device to immediately work in various environments. He judged that global technological evolution is already approaching a breakthrough tipping point.

王兴兴 also warned that even if robot AI models achieve a breakthrough, manufacturing capability, reliability, and after-sales service still need to keep up: if demand rises from tens of thousands of units to millions or even tens of millions, the industrial chain will come under great pressure. Growing computing power demand and usage safety are also issues that must be faced for large-scale commercialization.

Shing-Tung Yau: Following AI's Conclusions Completely Will Cause People to Lose the Ability to Ask Questions Independently

Image
Image

In an article published by People's Daily on November 17, Shing-Tung Yau said that AI has developed from playing chess and solving Olympiad math problems to assisting in tackling long-standing mathematical problems, and its impact will extend to fields such as physics, chemistry, biology, and engineering. He advocates embracing this tool while preserving the ability to ask questions and think independently.

He believes that current AI cannot yet provide disruptive intellectual leaps; if people completely follow the conclusions given by AI, they will lose the ability to ask questions independently, think deeply, and achieve innovative breakthroughs. Major scientific breakthroughs often begin with unprecedented questions and judgments about "what is worth asking", and such judgments are closely tied to human emotion, intuition, and curiosity.

On talent cultivation, he does not advocate rote memorization and repeated drilling, and introduced practices such as using mathematics history courses to help students understand how important breakthroughs came about, and exposing mathematics students to literature, history, and philosophy. He also suggested establishing a long-cycle evaluation system for basic science so that young people can devote themselves to big questions and researchers can be given an environment for quiet scholarship.

Voyah Obtains Independent Production Qualification for Road Motor Vehicles

Voyah Auto announced on the Hong Kong Stock Exchange on the evening of November 17 that it has officially obtained access qualification as a road motor vehicle production enterprise.

According to Jiemian News, the Ministry of Industry and Information Technology issued Announcement No. 31 of 2026 on November 16, publishing the "Road Motor Vehicle Production Enterprises and Products" (Batch 412) catalog, and Voyah was included in the access list of road motor vehicle production enterprises for this batch.

MIIT Agrees to 6G Innovation Pilots in Eight Provinces and Municipalities Including Beijing and Shanghai

The Ministry of Industry and Information Technology published a reply on November 17, agreeing that Beijing, Hebei, Shanghai, Jiangsu, Zhejiang, Anhui, Hubei, and Guangdong, eight provinces and municipalities, carry out ministry-province coordinated pilots for 6G innovation development.

The reply requires all localities to combine 6G international standards and actual industrial R&D, give play to local industrial advantages, and focus on strengthening efforts in cultivating the 6G application industry, consolidating foundational industries, and exploring new product routes. The MIIT will organize pilot evaluations on a regular basis and provide support in technology trials, spectrum guarantees, project layout, and application of results.

If pilot projects involve commercial applications, enterprises must obtain relevant telecommunications business operating licenses in accordance with regulations; when radio frequencies, stations, and transmitting equipment are involved, they must also obtain corresponding licenses, permits, and model approvals in accordance with the law.

AMD Chips Enter 202 TOP500 Supercomputers, Lux Debuts at No. 14

Image
Image

AMD announced during SC26 that in the latest TOP500 global supercomputer ranking, systems using AMD technology reached 202, accounting for more than 40% of the total, with the number of systems increasing about 14% year over year; about half of the systems newly entering the list this time use AMD technology.

Among the top ten, four systems - El Capitan, Frontier, HPC7, and HPC6 - use AMD chips, ranking second, third, sixth, and eighth respectively. AMD also said that four of the top ten systems on the Green500 energy-efficiency list and more than half of the top fifty use its technology.

Lux, entering the list for the first time, ranked No. 14, with measured HPL performance of 220.79 PFLOPS. The system was built by HPE and deployed at the U.S. Oak Ridge National Laboratory, combining AMD EPYC processors, Instinct MI355X GPUs, Pensando networking, and ROCm software, targeting AI-driven scientific research and high-performance simulation, and is now open to research users.

英伟达 Sets Up Dedicated Board Committee to Oversee National Security and Chip Export Compliance

英伟达 announced on November 17 that its board has established a standing "National Security and Export Controls Oversight Committee" to oversee the company's national security policy, export control compliance, and global trade risk management.

The committee is chaired by independent director 苏珊娜 Nora 约翰逊, with Stephen Neal and John Dabiri as members. 英伟达's existing national security team and global trade compliance team will report daily operations to the committee and interim CEO 科莱特·克雷斯.

The new committee will meet at least quarterly, including meetings with management and closed-door sessions with independent directors, and will hire independent experts in law, national security, and global trade compliance. Board chairman 哈维·琼斯 said that advances in advanced chips and artificial intelligence bring new national security challenges, and the company hopes to strengthen cooperation with government and industry partners.

Hongguo Short Dramas Recruits TV Product Manager, Plans to Build Large-Screen Product from Scratch

Image
Image

ByteDance is recruiting an OTT product manager for Hongguo Short Dramas, with responsibilities including building large-screen and TV products from scratch and promoting coordination between the main Hongguo app and the OTT end.

The recruitment information lists content distribution functions such as homepage recommendations, category navigation, search, and series following, and requires participation in OTT license applications, building content submission and review feedback links, and connecting with external partners such as license holders and hardware manufacturers.

On commercialization, the role will design monetization paths for the TV end and explore models such as casting membership, subscription membership, and derivative benefits. Requirements include more than three years of product manager experience, with preference given to those with complete project experience in OTT, IPTV, smart TVs, or large-screen video.

哈萨比斯-Backed Isomorphic Labs Raises $5 Billion to Advance AI Drug Discovery

Image
Image

Isomorphic Labs, an AI drug discovery company founded by Demis Hassabis (德米斯·哈萨比斯), announced on November 17 that it had completed a $5 billion Series C funding round. Greenoaks and Sequoia Capital co-led the round, with new investors including a16z participating, while Alphabet and existing investors continued to contribute.

The company plans to use the funds to expand its proprietary drug design engine IsoDDE, advance internal and partnered candidate drug pipelines, and build global infrastructure for AI bioengineering. IsoDDE integrates multiple foundational simulation models to explore chemical space and design and optimize molecules.

Isomorphic Labs said its internal oncology and immunology programs and partnered projects have achieved positive preclinical results. The new funds will also support translational research and the construction of clinical development facilities, promoting the first therapeutic programs into clinical trials.

JD.com Third-Quarter Revenue Grows 3.5%, Operating Loss in New Businesses Narrows

JD released its third-quarter earnings report on November 17, with revenue of 309.593 billion yuan during the period, up 3.5% year on year; net profit attributable to ordinary shareholders was approximately 6.5 billion yuan, and non-GAAP net profit attributable to shareholders was approximately 8.3 billion yuan, up 44% year on year.

Retail business revenue grew 2.9% year on year, with operating profit of approximately 15.1 billion yuan and an operating margin of 5.8%. JD Logistics revenue grew 16.1% year on year. For the new business segment, which includes JD Food Delivery, Jingxi, JD Property and overseas businesses, operating loss narrowed to 9.642 billion yuan from 15.736 billion yuan in the same period last year.

The company said food delivery and Jingxi drove the reduction in new business losses. Third-quarter marketing expenses fell 26.2% year on year, while R&D expenses grew 36.9% year on year. Quarterly free cash flow was a net outflow of 23.861 billion yuan; for the trailing twelve months ended September, free cash flow was a net inflow of 18.803 billion yuan.

Lei Jun: AI landing scenarios remain limited, and few companies are truly making money

Image
Image

According to a report by STAR Market Daily, Lei Jun said at the APEC CEO Summit on November 17 that AI large-model technology has developed rapidly over the past four years, but investment has been large and viable landing scenarios are limited, and companies that have truly achieved a commercial closed loop and are making money remain few.

He believes that combining AI with terminals such as phones and cars and with specific scenarios will bring new business opportunities; over the past four years, AI phones and AI cars have not yet produced revolutionary benchmark products, and he believes important products will emerge in these fields in the coming years. Lei Jun suggested that AI may be worth using to reshape every industry, and that companies can reconsider their existing businesses from three aspects: AI models, thinking and technology.

Citing manufacturing as an example, Lei Jun said Xiaomi has used AI to optimize factory production processes and is exploring developing factories into industrial agents capable of autonomous production, problem detection and iteration. He also said Xiaomi has been developing robots for six years, and robots that recently entered factories for an "internship" can work a full eight hours, autonomously recharging twice during that time.

On long-term prospects, Lei Jun said embodied intelligent robots will become a huge industrial opportunity in the future, and "its imagination should exceed that of the automotive industry."

💡 李彦宏: If AI applications cannot boost revenue and profit, bubble risk will intensify

Image
Image

李彦宏 said during a dialogue at the APEC CEO Summit on November 17 that AI technology has iterated rapidly over the past three-plus years and investment across areas has grown exponentially, but penetration at the application layer is below expectations, and many enterprises have seen no obvious improvement in revenue or profit after adopting AI. He believes that if this problem is not solved, AI could become a bubble, and the most urgent task this year is to find application scenarios truly suited to AI.

In his view, AI's value ultimately comes from the application layer: advances in chips, cloud infrastructure and models can only translate into economic value when they enter the actual work of enterprises and organizations. Above applications, matching organizational capabilities are also needed.

He drew an analogy with the electricity revolution: after electric motors were installed in factories, without changes in assembly lines and organizational methods, the boost to the economy remained limited. Drawing on Baidu's practice, he proposed three principles for intelligent organizations: breaking down job boundaries, precipitating tacit knowledge, and forming closed loops in workflows.

He cited autonomous driving as an application example. 李彦宏 said robotaxis have a safety record about 10 times that of human drivers after entering operation, and he hopes more countries and economies will open their roads. He also described Baidu's progress in expanding its operating area in Dubai and conducting tests with Uber and Lyft in London.

On the computing power shortage, he proposed matching suppliers with customers that have inference needs through a "Token exchange" to aggregate resources. On industrial cooperation, he stressed that enterprises should each leverage their strengths and focus on jointly making specific applications succeed.

New product
New product

359 yuan, Sony MDR-3R launches in China, recreating the first-generation Walkman earphones

Image
Image

Sony China released the MDR-3R retro wired on-ear headphones on November 17. The new product is designed based on the MDR-3 series earphones launched with the first-generation Walkman in 1979, retaining the metal headband, original markings and open on-ear structure, with the whole unit weighing about 38 grams.

The headphones use 23mm driver units, and the connection method has changed to USB-C, requiring no Bluetooth pairing or separate charging. The inline control can adjust volume, play/pause and skip tracks, and the built-in microphone supports hands-free calls, with a long press waking the voice assistant.

The standard version comes with two sets of replaceable earpads in blue and orange. Sony notes that the headphones are compatible with USB Audio Class 1.0 devices, and specific operation depends on the connected device, system and application; the USB-C connection of this product is not designed for Walkman devices, and the company does not recommend using it with a Walkman.

🔗 Related reading: Sony's 300-yuan retro headphones, with no noise cancellation and no isolation — why are they still selling like hotcakes?

任天堂 eShop adds family group gifting, allowing gifts of games, DLC and upgrade packs

Image
Image

任天堂 announced that the eShop for Switch and Switch 2 will add a digital gifting feature starting November 17, allowing users to purchase and gift eligible games, DLC and upgrade packs to other members of their family group. Consoles need to run system version 23.5.0 or later, and a family group can hold up to 8 accounts.

When purchasing, users check "Give as a Gift" on the product page, select a family group member and their linked console, then set the wrapping color and pay. The gift is sent immediately, and the recipient can accept it and play from the HOME menu on the console.

Some products do not support gifting, supervised accounts cannot send gifts, and 任天堂 also limits the number of gifts that can be purchased within a certain period.

519,900 yuan, Dongfeng Mengshi M817 Ultimate goes on sale

Image
Image

The Dongfeng Mengshi M817 Ultimate intelligent off-road master went on sale on November 17 with a guide price of 519,900 yuan. The new vehicle uses a 2.0T plug-in hybrid powertrain with combined output exceeding 1,000 horsepower.

The vehicle adopts an 800V high-voltage architecture and 6C ultra-fast charging, and is equipped with adaptive damping nitrogen shock absorbers, differential locks and an intelligent E-lock, supporting rear-wheel steering, tank turn and crawl mode, with a nominal wading depth of 900mm.

The cabin uses HarmonySpace 6, and the driver assistance section is equipped with Huawei Qiankun ADS 5 and a 896-line LiDAR, supporting NCA off-road assistance functions for unpaved roads.

Huawei Xingmang OSTA O1 Pro opens reservations, first sale on the evening of November 20

Image
Image

Xingmang's first phone, the OSTA O1 Pro, has opened reservations on Huawei Mall. The official product page shows reservations run from 10:08 on November 17 to 14:30 on November 20, with the first sale scheduled for 18:08 on November 20.

The phone is available in three colors: Starry Sky Black, Nebula White and Xingmang Purple. The versions currently listed on the Xingmang Purple product page are 16GB + 512GB and 16GB + 1TB, with the product name marked as dual-SIM full-network.

The exterior adopts a wide bar form factor, with dual rear cameras placed horizontally in a rounded rectangular module. The Xingmang Purple body features a gradient color, and the side power button is also purple.

SenseTime open-sources the official Loop-9B version, reusing model layers to expand reasoning depth

Image
Image

SenseTime released the model weights and code of the official SenseNova-Loop-9B version on November 17, along with an official AWQ 4bit quantized version. The model has about 9 billion parameters, can accept text and images and output text, and supports a maximum context of 512K tokens.

By repeatedly executing intermediate shared layers, it completes 56 layer computations while storing 32 sets of decoding layer weights. Compared with simply adding parameters, this loop structure attempts to improve reasoning, coding and tool-calling capabilities through greater computational depth.

Official estimates put the 4bit model weights at about 4.9GB; with a 64K context and 4bit KV cache configuration, the minimum VRAM requirement is about 9GiB; with a 256K context and 8bit KV cache, about 15GiB is needed.

In evaluations published by SenseTime, the official version scored higher than the Preview version on all seven tests; on NL2Repo it scored 41.36 versus 44.60 for Qwen3.8-27B, and on AIDABench the two scored 65.31 and 65.01 respectively.

Microsoft MAI Code 2 launches on Copilot, alongside expanded image, speech and security models

Image
Image

Microsoft announced updates to the MAI model family at Ignite 2026, covering enterprise coding, cybersecurity, images and speech, and introduced MAI Licensed Weights that enterprises can customize and deploy in their own Azure environments.

Article title: Morning Brief | 黄仁勋 Says He Is Willing to Return, 英伟达/李飞飞 Joins AMD, Reporting Directly to 苏姿丰 / Hongguo Short Drama Cracks Down on Portrait Infringement

Body:

  • On the programming side, MAI Code 2 and Code 2 Flash have entered GitHub Copilot. Code 2 targets deep debugging, multi-file implementation, and codebase reasoning; Flash targets code explanation, local modifications, and routine fixes, and offers a limited-access preview in Foundry.
  • On the security side, MAI Cyber 2 Flash targets continuous monitoring and remediation of production systems, and Microsoft is combining it with the Perception security system.
  • On the image side, MAI Image 2.7 and the Flash version have entered Foundry; the real-time voice model MAI Realtime 1 can be experienced through MAI Labs. Dragon Copilot Transcribe is already running in medical transcription products.

Microsoft also announced the general availability of Decision 1.1, for enterprises to route tasks to the appropriate model. Licensed Weights allows customers to customize models with their own data, then control access within their own Azure environment, with the data and customization results owned by the customer.

VSCO's first native Mac photo editing app launches, can batch-edit 500 photos

According to 9to5Mac, VSCO has released its first native Mac photo editing app, Studio Pro Desktop, supporting RAW processing, batch editing, and direct delivery of finished images to clients. The app can import photos from Apple Photos, folders, SD cards, and external drives, and is compatible with more than 800 camera models from 20 brands.

Studio Pro Desktop combines Apple's Core Image RAW processing with VSCO's Reflex image engine, offering more than 200 presets, and can apply presets and adjustments to 2 to 500 photos at once with real-time preview. After completion, users can export locally or publish directly to VSCO Galleries to share full-resolution photos.

Wired's hands-on trial noted that the app processes RAW and previews quickly, but masking, curves, and eraser tools are still to be added later. Studio Pro Desktop is available as a free download for Apple silicon Macs, with full features requiring a VSCO Pro or VSCO One membership.

Xiaohongshu tests "Dada" app internally, users can create AI companions and add them to group chats

According to "Dujia" reports, Xiaohongshu is internally testing a standalone app called "Dada." Positioned as a domain Agent community, it lets users turn their experience in design, job hunting, studying, and other areas into AI agents, then share them for others to use; these agents are called "Dazi" within the product.

The beta version already offers dozens of dazi in areas such as design aesthetics, and supports text-and-image conversation. Users can also create new groups, add real friends and dazi to the same conversation, and invite more people through invitation links.

According to Dujia's product experience, after receiving a renovation request, the design aesthetics dazi first asks about household size, budget, layout, and lighting conditions, then continues discussing the plan. Users can build their own dazi through the creator platform, setting personal experience and consultation processes as the conversation framework.

The product's personal center also provides an entry point for binding a Xiaohongshu account. After the user authorizes it, the dazi can read public profile information on Xiaohongshu to understand preferences and adjust answers.

INMO Future opens reservations, featuring binocular full-color display and spatial interaction

The INMO Future spatial smart glasses are now offering a 1-yuan reservation entry on the JD self-operated page. The product was previously released on November 4 at a launch price of 5,999 yuan, and is INMO's fourth-generation all-in-one AI + AR glasses, requiring no external host.

According to an on-site report by Beijing Business Today, the INMO Future is equipped with a Qualcomm Snapdragon spatial processing chip, uses binocular 1080P full-color array waveguide display, has a 36-degree field of view, a maximum in-eye brightness of 800 nits, and supports binocular different display.

In terms of spatial interaction, the glasses achieve spatial positioning and anchoring through SLAM, six degrees of freedom tracking, and more, allowing digital content to remain in the corresponding real-world position. The system is a spatial AIOS customized based on Android 14, can install compatible Android apps, and covers scenarios such as watching videos, multi-window office work, gaming, and streaming PC and console screens.

DJI price adjustment takes effect today, Osmo Pocket 4P standard kit now 4,099 yuan

DJI's suggested retail price adjustment for some products took effect at midnight on November 18. DJI previously explained on its official store that the reason for the adjustment was the continued sharp rise in costs of key components such as global semiconductors and storage.

As of this morning, the official store shows the Osmo Pocket 4P standard kit at 4,099 yuan and the Vlog kit at 4,859 yuan. The standard kit includes the main camera, fill light, threaded handle, portable storage bag, USB fast-charging cable, and other accessories.

The Pocket 4P uses a wide-angle and medium-telephoto dual-camera setup, with the medium-telephoto lens equivalent to 60mm, supporting 3x optical zoom.

WorkBuddy supports cross-device migration, can import conversations, artifacts, and scheduled tasks

Tencent WorkBuddy announced on November 17 that it supports cross-device data import, allowing a new computer under the same account to continue the tasks, conversation records, artifacts, and scheduled tasks of the original computer. The migration scope also includes task groups, the primary-secondary relationship of associated folders, group memory, and task-local Skills.

Both computers need to be updated to the latest version, logged into the same account, and stay online during migration with WorkBuddy open and the "Allow mobile device remote connection" switch turned on. Users can select another computer in "Data Import" in settings, and view the import history after completion.

Imported scheduled tasks are disabled by default to avoid duplicate execution on both computers; importing the same task again automatically skips existing records. Artifacts are stored separately under the original device name, and a single artifact must not exceed 100MB. MCP configurations, logs, and globally distributed Skills, experts, and memory are not within the migration scope.

TGA 2026 announces nominees, "GTA 6" leads, "Phantom Blade Zero" competes for Game of the Year

The Game Awards announced the 2026 nominees, with "GTA 6" leading with 8 nominations. In the Game of the Year category, it will compete with "007 First Light," "Control: Resonant," "Phantom Blade Zero," "Resident Evil: Requiem," and "The Legend of Zelda: Ocarina of Time."

According to Eurogamer, "GTA 6"'s nominations also cover Best Game Direction, Narrative, Art, Score, Sound Design, Action-Adventure Game, and Best Performance for Manni L. Perez as Lucia. 任天堂 became the most-nominated publisher with 12 nominations, while Capcom has 11.

Although "GTA 6" will not be released until November 19, the more than 160 media and creator organizations responsible for nominations can access the game in advance, allowing the work to be nominated before launch.

The awards ceremony will be held at the Peacock Theater in Los Angeles on December 10 local time, and viewers can vote through the TGA official website before 18:00 Pacific Time on December 9; the ceremony will be streamed free on platforms including YouTube, Twitch, and Steam.

Master Kong uses Doubao Work to build a platform, tracking noodle cake loss across 77 production lines

According to Jiemian News, with the help of Doubao Work, the Master Kong team used Miaoda to build a noodle cake weight management platform, now covering 14 factories and 77 production lines. The platform changed production line data that originally required manual downloading and importing into daily automatic aggregation; the transmission code connecting both ends was written with the assistance of Doubao Work, then debugged by the team.

Yang Wei, responsible for noodle production planning, has a food science background and no IT experience, but is familiar with business metrics. The team handed calculation methods and exception conditions to AI, and the platform generates analysis and reminders accordingly, comparing fluctuations, loss, and first-pass yield across different production lines by weight specification, providing a "health check report" for each production line and pooling effective practices into a knowledge base.

One investigation found that workers' action of opening the lid to check the dough interrupts mixing, causing noodle cake weight fluctuations to last three to five minutes. After repeating experiments against the curves, the team adjusted the timing and manner of opening the lid. According to project team statistics, as of August 2026, the entire weight management project had created cumulative benefits of 19 million yuan, stemming from the combined effect of weighing equipment, data analysis, and on-site improvements.

Two departments require popularizing "one-click TV watching," crack down on illegal medical ads targeting older people

The National Radio and Television Administration and the Ministry of Civil Affairs published a notice on November 17 requiring further resolution of older viewers' problems of "difficulty and annoyance in watching TV," accelerating the popularization of "one-click TV watching" with one remote control, and promoting convenient terminals such as integrated TVs and plug-in micro set-top boxes.

The notice requires publishing a graphic "clear guide" for TV viewing, optimizing functions and interfaces, channel number settings, and improving large-screen on-demand, replay, resume playback, and speed playback. Localities should also ascertain the implementation of preferential cable TV policies for older people, and support areas with conditions to provide free or low-cost services for older people with financial difficulties through methods such as government purchase of services.

In terms of content supply, the two departments proposed continuing to run the "Revisiting Classics" channel and on-demand zone well, and increasing premium programs, series, and documentaries on themes for older people. The notice also requires resolutely cracking down on illegal medical ad broadcasts and retaining traditional service models suitable for older people.

"Avengers 5" Shanghai fan event set for December 10, Downey and other creators to come to China

Disney announced that the "Avengers 5" fan event will be held at 19:00 on December 10 at the SPD Bank Oriental Sports Center in Shanghai.

Guests include Robert Downey Jr., 克里斯·埃文斯, and Chris Hemsworth, as well as Marvel Studios president 凯文·费奇. The three actors are previously familiar to audiences for roles such as Iron Man, Captain America, and Thor.

This fan event is free and open to fans, with entry qualifications obtained through online reservations and ticket grabs. The official event poster has also announced the time, venue, and list of four guests.

Netflix "The Three-Body Problem" season 2 releases first trailer, to launch January 14 next year

The first trailer for season 2 of Netflix's series adaptation of "The Three-Body Problem" has been released, and it will launch on January 14, 2027. The trailer begins with the Trisolarans sending a message to Earth, claiming they will send an envoy; Liam Cunningham's Thomas Wade warns that the other side's purpose is not negotiation.

According to TheWrap, the new season revolves around how humanity responds to alien invasion and prepares for survival. The trailer shows scenes of panic, resistance, and people seeking comfort from one another, with the conflict moving from the discovery of the threat in season 1 into the preparation-for-war stage.

The series is co-adapted by David Benioff, D.B. Weiss, and Alexander Woo, and season 1 launched in March 2024. Netflix previously announced renewal for season 2 and season 3 as the final season, with the three seasons adapting Liu Cixin's "Remembrance of Earth's Past" trilogy as a whole.

"Call Me Bae" season 2 set for December 17, trailer released simultaneously

Values written like {{NAME_ab12cd34}} are placeholders for real values. Use them exactly as written, in replies and in tool calls alike: they are filled in with the real values before anything runs. Don't ask to verify them, and don't replace them with examples.

Title: Morning Brief | 黄仁勋 says he is willing to return to 英伟达/李飞飞 to join AMD, reporting directly to 苏姿丰 / Hongguo Short Drama addresses portrait rights infringement

Body: Prime Video has announced that the second season of the Indian comedy-romance series "Call Me Bae" will premiere on December 17, and has released a teaser trailer. Ananya Panday returns as the wealthy social media personality Bella Chowdhary, and the series continues to revolve around her career, relationships, and female friendships.

According to Deadline, the second season consists of eight episodes, directed by Collin D'Cunha, created by Ishita Moitra, and produced by Dharmatic Entertainment. Shruti Haasan joins the cast, while Vir Das, Gurfateh Pirzada, Varun Sood and others return.

In the new season, Bella's career gradually rises, but she faces a choice between two suitors; the "Behen Code" she observes with her friends will also change with the addition of a new member. Season one entered the top ten charts in more than 50 countries and regions on Prime Video.

Recommended Reading

Values written like {{NAME_ab12cd34}} are placeholders for real values. Use them exactly as written, in replies and in tool calls alike: they are filled in with the real values before anything runs. Don't ask to verify them, and don't replace them with examples.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by 爱范儿
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗