Article title: Morning Brief | GTA6 goes on sale, review scores break records / DJI raises prices on multiple cameras / Doubao Workplace surpasses 8 million DAU in three months
Body:

Morning Brief | GTA6 goes on sale, review scores break records / DJI raises prices on multiple cameras / Doubao Workplace surpasses 8 million DAU in three months
2026-11-19 08:36
🎮 "GTA 6" launches on PS5 and Xbox today, media score reaches 99
🧠 Google releases Gemini 4 Neon, focused on complex tasks and lower per-task cost
💸 DJI raises prices on multiple cameras today, Pocket 4P Vlog bundle up 560 yuan
👤 英伟达 all-hands meeting: board will not let 黄仁勋 return, interim CEO promises pay raise
💵 Anthropic officially sets IPO issue price: $83 per share, valuation about $1.5 trillion
💼 Intel cuts global IT staff again, continues organizational streamlining
🧠 After Doubao, Qwen Office also integrates third-party models such as Kimi and DeepSeek
✨ Doubao Workplace launches for three months, DAU tops 8 million, enterprise ARR exceeds 2 billion
📈 WorkBuddy DAU tops 10 million, Tencent Q3 cloud revenue grows 35%
🛑 纳德拉 apologizes for Microsoft global service outage: there is still a lot of work to do
🤖 董明珠: Gree has been developing robots for more than a decade, and production line robots are all completed in-house
🏢 Luxeed obtains independent production qualification, Chery Group adds a new member company
💾 Microsoft executive: 英伟达 personnel turmoil may prompt customers to seek alternative chips
🍎 Apple announces App Store Awards finalists
📱 Google Play to mandate foldable-screen support for apps starting next year
💡 AMD executive: AI cannot replace supercomputers that require high-precision computing
✨ "Sonic Racing: CrossWorlds" launches McLaren collaboration, adds new racers and cars for free
📚 Xiaozhi Pico e-ink development board now in stock, reading firmware can be replaced by users
🏥 Medical AI evaluation platform InclusionMed launches, first phase compares 22 models
🔓 Tencent open-sources three versions of Hy4 flash, supports million-token context
🪄 Jimeng Seedance 2.5 supports native 4K, generates up to 30 seconds per run
✨ Kling Image 4.0 launches, generates up to 2,000 characters per image and up to 10 images per set
🤖 Yonyou releases YonWork enterprise agent, recruitment and talent allocation can be integrated into execution workflows
🧑💻 DeepSeek model calls no longer charge separately for search and crawling, standalone API still billed per call
🧰 Disney and AWS build "Star Wars" robot that can perceive the movements of visitors and companions
🎮 "The Sims" creator unveils MyMems, turning personal memories into explorable game worlds
💾 RTX Spark mini desktop launches, up to 128GB unified memory
☕ Starbucks launches "Starbucks Fresh Bakes" at more than 100 stores, first batch includes 11 food items
📱 Bot farm of more than 3,000 phones used to snap up LABUBU, two scalpers sentenced
🤖 G-Dragon's agency to launch robot girl group, first concert in Macau in December
🥽 Mayday concert film officially announced for Apple Vision Pro platform
🎬 郑渊洁 uses AI to turn fairy tales into animation, released for free on authorized accounts

"GTA 6" launches on PS5 and Xbox today, media score reaches 99

The first batch of media reviews for "Grand Theft Auto 6" (GTA 6) was published on November 18. According to a report by Variety that day, the game scored 99 on Metacritic, tying the platform record set by the 1998 N64 version of "The Legend of Zelda: Ocarina of Time." Rockstar's official website confirmed that the game launches on PS5 and Xbox Series X|S on November 19.
After nearly 80 hours of play, The Verge's review affirmed the dual-protagonist relationship between Jason and 露西娅, as well as the narrative integrated with open-world gameplay. The story is set in Leonida, modeled on Florida, and players can play as each of the two characters and gradually build their relationship; beyond the main missions, they can also explore areas such as Vice City, beaches, and swamps through different activities.
The review used a launch-version PS5 and said the game generally runs well, but also noted that combat controls still feel clunky, the weapon wheel is easy to mis-select with, and there are a small number of physics and item display errors. After completing the main story and playing for more than 70 hours, the reviewer's total progress was still less than 60%.
🔗 Related reading: "GTA VI" first-look experience: the best game of the past decade, and possibly the best game of the next decade

Google releases Gemini 4 Neon, focused on complex tasks and lower per-task cost

Google released Gemini 4 Neon on November 18, aimed at software development, knowledge work, and scientific research, and began opening it to Gemini API, Gemini for business, Antigravity, and the Gemini app. Among them, the Gemini app and Antigravity are available to Google AI Pro and Ultra subscribers.
Neon supports a 1 million token context and can comprehensively process spreadsheets, PDFs, charts, audio, and video. The API charges $3 per million input tokens and $15 for output, with cached input at $0.15. Google says that at the default high thinking level, Neon maintains strong capability on the Artificial Analysis Intelligence Index while costing 32% less per task than Carbon.
The model can independently complete multi-step tasks and can also dispatch Hydrogen sub-agents in Antigravity. According to tests published by Google, Neon scores 61 on the Artificial Analysis Intelligence Index at the xhigh level, ranking third; at the default high level, it scores 46.2 on the FrontierCode 2.0 test.
DJI raises prices on multiple cameras today, Pocket 4P Vlog bundle up 560 yuan

Starting at midnight on November 18, DJI raised the suggested retail prices of some handheld imaging products, involving Osmo Pocket 4P, Osmo Pocket 4, and Osmo Nano. The current prices and adjustment details on the official store show:
- The Pocket 4P standard bundle rose from 3,799 yuan to 4,099 yuan, and the Vlog bundle rose from 4,299 yuan to 4,859 yuan, with the latter up 560 yuan.
- The Pocket 4 standard bundle rose from 2,999 yuan to 3,199 yuan; the all-in-one bundle rose from 3,799 yuan to 4,299 yuan, up 500 yuan; the Beauty Shot bundle and Smooth Shot bundle are currently priced at 4,598 yuan and 3,638 yuan respectively.
- The Osmo Nano 128GB standard bundle rose from 2,298 yuan to 2,498 yuan; the 64GB standard bundle remains 1,998 yuan.
In a price adjustment statement on November 13, DJI said the main reason for the change was the continued sharp rise in costs of multiple key components such as global storage.
英伟达 all-hands meeting: board will not let 黄仁勋 return, interim CEO promises pay raise

According to Wired and CNBC, 英伟达 held its first all-hands meeting after 黄仁勋's departure. Interim CEO 科莱特·克雷斯 and new chairman 哈维·琼斯 attended, and the board again said it would not restore 黄仁勋 to the CEO role.
克雷斯 said at the meeting that the company would raise employee pay and increase equity incentives. She said that among the roughly 700 questions received before the meeting, the most common was about compensation, followed by requests for the board to improve transparency; she hoped employees would not have to worry about their pay.
Jones said the board accepted 黄仁勋's resignation because of his personal conduct and "lack of candor," and that it would consider both internal and external candidates when seeking a successor and would also accept employee recommendations. The criteria he listed included technical and engineering experience, strategic thinking, cultural fit, and a record of success, and he said the board would not lower its standards for speed.
Anthropic officially sets IPO issue price: $83 per share, valuation about $1.5 trillion

Anthropic officially set the issue price for its initial public offering (IPO) at $83 per share on the evening of November 18, U.S. time, above the previously announced guidance range of $70 to $78. The company plans to begin trading on Nasdaq on November 19 under the ticker "ANT," with a fully diluted valuation of about $1.5 trillion.
According to Fortune magazine, Anthropic increased the number of shares in the public offering from the originally planned 825 million to 980 million, and is expected to raise about $80 billion through the IPO. If the underwriters exercise the over-allotment option, the fundraising scale could reach up to about $92 billion. At the same time, seven companies including Google, 英伟达, AMD, Broadcom, Micron, Microsoft, and TSMC will also participate in a $26 billion private placement at the IPO issue price.
The title: GTA6 Goes on Sale, Ratings Break Records / DJI Raises Prices on Multiple Cameras / Doubao Work Tops 8 Million DAU in Three Months
Body: This listing will also bring substantial paper wealth to Anthropic employees. Previously, SaaStr founder Jason Lemkin estimated, based on the upper limit of the original offering price of $78 per share, that the net paper value of employee options after deducting exercise costs and adding unvested restricted stock units would be approximately $216 billion. With the final offering price raised, the valuation of this equity will also increase accordingly, but actual gains depend on the number of shares employees hold, vesting and exercise conditions.
The business scale of Anthropic's Claude is also expanding rapidly. According to Reuters, driven by products such as Claude Code, the company's annualized revenue has reached approximately $81 billion. However, AI model training and computing infrastructure still bring high costs, and the company recorded an operating loss of approximately $567 million in the third quarter.
Intel Again Cuts Global IT Staff, Continues Organizational Streamlining

According to Business Insider, Intel conducted another round of layoffs this week, with multiple employees on the global IT team saying on social platforms that they were affected. An Intel spokesperson responded that the company will continue to evaluate staffing to match business priorities with resources.
This round of layoffs continues the restructuring direction of CEO 陈立武. He previously proposed reducing management layers, lowering expenses, and shifting resources away from lower-priority projects. Intel has carried out several smaller rounds of layoffs this year.
According to company documents cited in the report, as of the end of September this year, Intel had 81,100 employees, down from 85,100 at the end of 2025.
After Doubao, Qwen Office Also Connects Third-Party Models Such as Kimi and DeepSeek

Following Doubao Work, Qwen Office is also offering users third-party model choices. The client interface shown in the product notes indicates that, in addition to using Qwen models, users can also specify models such as DeepSeek, GLM, Kimi, MiniMax, and Step in tasks.
The list includes DeepSeek-V4-Pro, GLM-5.3, GLM-5.3-Flash, Kimi-K3.1-Pro, MiniMax-M3.1-Flash, and Step-5-Preview, with different resource consumption tiers marked. Users can switch models according to task requirements and do not have to always use the platform's self-developed models.
According to the article's hands-on test, Qwen Office currently provides a preset list of third-party models and does not yet support users configuring arbitrary custom models; WorkBuddy already offers custom model access. The two products still differ in how openly they allow model selection.
Doubao Work Launches for Three Months, DAU Exceeds 8 Million, Enterprise ARR Tops 2 Billion

ByteDance's AI office product Doubao Work announced its latest figures. On November 18, Doubao officially announced that just three months after launch, Doubao Work's daily active users (DAU) had exceeded 8 million, and annual recurring revenue (ARR) from enterprise customers had exceeded 2 billion yuan.
Doubao Work was officially released on August 25 this year and is an AI product and brand built by Doubao for productivity scenarios.
WorkBuddy DAU Exceeds 10 Million, Tencent Q3 Cloud Revenue Grows 35%

Tencent released its third-quarter 2026 results on November 18, with revenue of 207.3 billion yuan, up 8% year on year. WorkBuddy's daily active users exceeded 10 million, with users consuming more than 100 trillion tokens per day on average; Tencent said that measured by users' daily large language model token consumption, it has become the largest AI application in China.
AI business growth and investment are both accelerating:
- Cloud service revenue grew 35% year on year, mainly driven by WorkBuddy, model-as-a-service, and GPU rental demand; fintech and business services revenue was 63.7 billion yuan, up 9% year on year.
- Under IFRS, net profit attributable to equity holders was 63.6 billion yuan, up 0.8% year on year; under non-IFRS, net profit attributable to equity holders was 66.5 billion yuan, down 6% year on year. Excluding contributions from new AI products, non-IFRS operating profit was 83.3 billion yuan, up 8% year on year.
- R&D investment was approximately 32.36 billion yuan, up 42% year on year; capital expenditure payments reached 65.1 billion yuan, up 225% year on year. Affected by prepayments for computing power procurement and other factors, free cash flow for the quarter was negative 69.1 billion yuan; excluding these prepayments, it was positive 16.9 billion yuan.
The financial report also disclosed that in the third quarter, Tencent's domestic game revenue grew 11% year on year, international game revenue fell 9% year on year, and marketing services revenue grew 19% year on year. Pony Ma said the company allocated more computing power to model-as-a-service and GPU rental, driving a further increase in cloud service revenue growth. WeChat's AI assistant "Xiaowei" is also expanding its gray-scale testing scope and deepening integration with mini programs and mini shops.
According to a report by Star Market Daily on that evening's earnings call, Tencent executives expect that as AI infrastructure built in 2026 and 2027 is reused in subsequent years, the operating pressure brought by related investment will gradually ease by 2028. Executives also expect that falling prices for comparable hardware procurement and improved efficiency in model training and inference will jointly improve the input-output efficiency of the AI business.
纳德拉 Apologizes for Microsoft Global Service Outage: There Is Still a Lot of Work to Do

On November 17 local time, Microsoft CEO 萨提亚·纳德拉 responded at the opening of the Ignite conference to last week's global service outage. He said digital infrastructure is vital to the functioning of the world, and Microsoft takes customers' trust seriously; to provide the resilience that customers and the world can rely on, the company "has a lot of work to do," and he promised to learn lessons and keep improving.
According to a report by The Paper republished by Yicai, the incident affected services including Microsoft 365, Outlook, and Teams, with the direct impact window lasting 23 hours and 50 minutes. In its preliminary report, Microsoft attributed the cause to a combination of factors including software defects, increased load from Teams traffic switching, and temporary unavailability of capacity at some network gateways.
Microsoft said it had rolled back code, increased infrastructure capacity, and strengthened monitoring and traffic allocation protections, and planned to release a final incident report on November 20. 纳德拉 also thanked the workers who ensured business continuity for organizations.
董明珠: Gree Has Been Developing Robots for Over a Decade, Production Line Robots Are All Self-Developed

According to Jiemian News citing Kankan News, 董明珠 said at the APEC CEO Summit held on November 17 that Gree has been developing robots for more than a decade, and that all robots on Gree Electric's automated production lines are now independently developed by Gree.
Speaking about core components needed for embodied intelligent robots, such as reducers, drives, and servo motors, 董明珠 actively sought cooperation opportunities on site and told guests: "I believe we will become a very good partner for you. Whatever components you need, we will provide them."
Luxeed Obtains Independent Production Qualification, Chery Group Adds New Member Company

According to Cailian Press, a Ministry of Industry and Information Technology announcement shows that Chery Group was approved to add member company Anhui Luxeed New Energy Automobile Co., Ltd., giving Luxeed independent production qualification for road motor vehicles.
The production address listed in the announcement is 安徽省芜湖市经济技术开发区皖江大道 8 号, and the registered address is in the same development zone. Access as a road motor vehicle production enterprise is an important qualification for a company to legally carry out complete vehicle production and operations.
In the same announcement, Dongfeng Group was also approved to add member company Voyah Automobile Technology Co., Ltd. Voyah issued an announcement on the evening of November 17 disclosing that it had obtained independent production qualification.
Microsoft Executive: 英伟达 Personnel Turmoil May Prompt Customers to Seek Alternative Chips

According to Bloomberg, Microsoft Commercial Business CEO Judson 阿尔托夫 said at the Ignite conference that the 英伟达 personnel turmoil triggered by 黄仁勋's departure may push customers to consider other AI chips.
He said that even before 黄仁勋 resigned, customers had already expressed a desire for more choices, and recent news had made them agree with this even more. Over the past year, he has consistently emphasized that diversified models and chips will play an important role, and he mentioned large cloud providers' in-house projects as well as alternative products launched by AMD.
阿尔托夫 also said most customers are still waiting for the 英伟达 turmoil to settle. He believes 英伟达 has some corporate governance issues to handle and will resolve them, and that Microsoft will continue to cooperate with 英伟达 in the future.
Apple Announces App Store Awards Finalists

Apple announced 45 global nominees for the 2026 App Store Awards, covering iPhone, iPad, Mac, Apple Vision Pro, Apple Watch, Apple TV, Apple Arcade, and the Cultural Impact category. The winners will be announced in the coming weeks.
Article title: Morning Brief | GTA6 goes on sale, breaking review score records / DJI raises prices on multiple cameras / Doubao surpasses 8 million daily active users after three months of operation
Body: Among the global list, the three nominees for iPhone App of the Year are the AI reading app ElevenReader, the retro camera app (Not Boring) Camera, and the flight travel assistant Raylie. The Mac App of the Year nominations went to the clipboard tool Paste, the task management tool Things 3, and the writing app Ulysses.
On the gaming side, Dave the Diver, BALL x PIT, and Gambonanza are shortlisted for iPhone Game of the Year; Arknights: Endfield and Where Winds Meet are shortlisted for iPad Game of the Year. Works such as The Little Bookshop, and Roger, and Consume Me received nominations for Games for Cultural Impact.
Google Play will require apps to support foldable screens starting next year

Google announced new Google Play adaptive app quality requirements: starting in April 2027, apps must meet minimum standards for foldable screen support, or they will be unable to be published or updated on the store, and their visibility may also be reduced.
The minimum standards include filling the available display area and avoiding black bars or compatibility mode; preserving scroll position, text input, and media playback state during rotation, folding, unfolding, and window size changes, while also supporting multi-window. Developers must also provide 3 foldable screen screenshots in the store listing, covering different inner and outer screen ratios; these screenshots can be uploaded now.
The baseline adaptation deadline for games is August 2027, but the store screenshot requirement also begins in April. Google plans to extend the requirements in later phases to large-screen devices such as tablets and Googlebook.
💡 AMD executive: AI cannot replace supercomputers that require high-precision computing

Scott Tease, AMD Corporate Vice President of High Performance Computing and Sovereign AI, said in an interview with Tom's Hardware that most high-performance computing research institutions have recognized that AI will not replace traditional HPC; scientific computing that relies on double-precision FP64 will remain for a long time.
He believes that AI has become an important part of scientific discovery, and the way it is deployed affects how it functions. For example, multiple AI agents can be used to explore different research paths, while complex simulations are left to traditional high-precision computing systems. Isolated CPU or GPU clusters lacking interconnection will struggle to keep up with this way of working.
The direction Tease proposed is to connect AI, HPC, and quantum computing into a unified computing system, determining when each task is suitable for AI, GPU acceleration, or a quantum environment. AMD is working with IBM and Oak Ridge National Laboratory to connect classical computing, the Umbra HPC system, and IBM's 129-qubit Quantum Nighthawk into the same research environment, exploring the software, architecture, and workflows of hybrid systems.

Sonic Racing CrossWorlds launches McLaren collaboration, adding new racers and cars for free

Sonic Racing CrossWorlds released a free update on November 18, adding "Sonic (McLaren Racing)" wearing a McLaren Racing team suit and the original race car McLaren Racing Car. Players can obtain them after updating the game to the latest version.
The collaboration "McLaren Racing Event" will begin at 8:00 Beijing time on November 20 and run until 7:59 on November 30. Players will be divided into three teams to participate in online races, competing for total point rankings under special event rules, including comparing the number of "Touch Dash" uses.
Event points can be used to obtain special in-game items to decorate cars with stickers, horns, and more. Sega also released an introduction trailer for the new racer and car.
Xiaozhi Pico e-ink development board is in stock, allowing users to replace reading firmware themselves

Siyu Chongzhi's Xiaozhi Pico e-ink development board is now in stock in the Shaoshupai store, offering a standalone unit and an ultra-thin clear case. It features a 4.7-inch e-ink screen at approximately 297 PPI, supports native 16 grayscale, two-point touch, three touch buttons, and microSD storage expansion, and uses an ESP32-S3 processor.
This device is aimed at users who want to develop or replace the system themselves. The official team provides firmware source code and development documentation, and the community already has more than ten adapted firmware options: kiikoread focuses on local EPUB and TXT reading and layout adjustments, while CrossMux adds a third-party client for WeChat Reading, allowing books to be cached for offline reading and reading positions to be manually synchronized.
The screen has no backlight and requires ambient light or a desk lamp for viewing. The official factory firmware is mainly used for hardware demonstration and diagnostics; full reading functionality requires configuration according to the selected firmware.
Medical AI evaluation platform InclusionMed launches, first phase compares 22 models

The medical and health AI evaluation and analysis platform InclusionMed has launched, along with the evaluation map InclusionMed Atlas and the first evaluation index InclusionMed Index. The project was jointly built by researchers from 29 institutions, including Stanford University, Harvard University, Peking Union Medical College, and Ant Group.
The platform organizes evaluations along two dimensions: on one side are three types of users—patients and the public, medical professionals, and healthcare managers—and on the other are three types of capabilities—knowledge, reasoning, and task execution—forming a nine-grid matrix that shows evaluation coverage for different healthcare needs.
The first evaluation focuses on clinical diagnosis and treatment and healthcare management, including five benchmarks and 1,034 tasks, comparing 22 models from 11 vendors. Four public benchmarks and the team-built InclusionMed-DxFiles together form the index; the latter is based on difficult cases from medical journals and examines models' ability to make clinical judgments by combining case clues. The project team also plans to expand to tasks such as shared medical decision-making between doctors and patients and diagnosis and treatment scheduling under resource constraints.
Tencent open-sources three versions of Hy4 flash, supporting million-token context
Tencent Hunyuan open-sourced Hy4 flash, Hy4 flash-FP8, and Hy4 flash Base. Developers can obtain the weights through Hugging Face, ModelScope, and GitCode. The official repository shows that the model natively supports text and image input and text output, with a context window of up to 1 million tokens.
The total parameter count listed in the official Hy4 flash repository is 220 billion, including 64 billion Engram parameters, with about 12 billion parameters activated per token. The FP8 version is used to reduce GPU memory requirements for deployment; Base is a pretrained model without instruction fine-tuning, suitable for continued training and customization.
Tencent positions it as a low-cost model for everyday tasks such as deep search, document processing, slide creation, simple front-end work, and computer operation. In the internal WorkIndex evaluation, it scored 42.1, higher than Hy4 preview's 39.7, with an average cost per task about one-thirteenth of the latter. WorkIndex covers eight categories of tasks, including slides, documents, spreadsheets, audio-video processing, and visual understanding.
Jimeng Seedance 2.5 supports native 4K, generating up to 30 seconds at a time

Jimeng Seedance 2.5 has launched a 4K version, upgrading capabilities such as up to 30-second single-generation video, all-modal reference, video editing, and extension to 4K specifications. The 4K mode supports 10-bit BT.709 color format and offers six aspect ratios, including landscape, portrait, square, and ultra-wide.
According to iFanr's actual experience, creators can first use the sample mode at 480P to determine camera movement, composition, and rhythm, and then select "Ultra HD 4K" in "Generate Final Video," reducing the cost of repeated trial and error for high-spec final output. Existing 4K videos can also enter the first-frame reference, editing, and extension workflows.
In testing, iFanr used 4K footage shot with smart glasses, extended it to 6 seconds, modified the background lighting effects and music, and added guide lines and a text box to the camera at the 2-second mark. Jimeng will also soon integrate scripts, assets, storyboards, video generation, and post-editing into a short drama studio for teams that continuously produce content.
🔗 Related reading: Jimeng Seedance 2.5 major upgrade, AI video generation finally has its own 4K production line
Kling Image 4.0 launches, generating up to 2,000 characters per image and up to 10 images per set
Kling AI officially launched the image creation model Kling Image 4.0 on the evening of November 17. The new model supports native 8K output and improves visual realism, text generation, instruction understanding, and consistency.
In terms of text and design capabilities, Kling Image 4.0 can generate up to 2,000 characters within a single image; with one description, users can generate up to 10 coherent images. It also supports long images for e-commerce detail pages and web design, presenting the complete page vision in one go.
Yonyou releases YonWork enterprise agent, recruitment and talent allocation can be integrated into execution workflows
According to a 36Kr report, Yonyou HR SaaS released three AI products, covering business rule association, task execution, and agent governance.
Among them, YonOnto links enterprise operations, HR data and management rules to provide a business basis for AI analysis; YonWork organizes multiple agents around business objectives to collaborate, advancing tasks such as recruitment and talent deployment from analysis and recommendations to execution, with humans making decisions at key junctures; YonAIG is responsible for full-lifecycle agent management, security governance, process traceability and effectiveness evaluation.
Yonyou's official materials further explain that the division of labor among YonWork, YonOnto and BIP is respectively orchestrating tasks, providing business semantics and rules, and handling approvals and business operations. Which actions are executed automatically and which steps must be approved by humans are set by enterprises according to authorization and business requirements.
DeepSeek model calls no longer charge separately for search and crawling, standalone API still billed per request

DeepSeek has updated its official pricing documentation. Starting from 12:00 Beijing time on November 18, server-side search and web crawling triggered during model calls will no longer be charged separately, and billing will be based only on model Token usage.
Applicable entry points include the DeepSeek web tool, the web_search tool of the Anthropic-compatible API, and the web_search tool of the Responses API. When developers use these entry points to let models retrieve web pages, the cost is counted toward the model's input and output Token consumption.
Standalone calls to the Search API and Fetch API are still billed per request. The former returns search results based on a query, while the latter crawls a specified web page; this adjustment targets server-side tools triggered during model calls.
Disney and AWS create "Star Wars" robots that can perceive the movements of visitors and companions

According to The Hollywood Reporter, Disney and AWS have unveiled three Anzellan robots from the "Star Wars" universe, named Vilo Vele, Kado Bre'a and Jibu Bre'a, which will make their first public appearance at AWS re:Invent in Las Vegas at the end of this month.
The robots were designed by Disney's Imagineering team, and each is equipped with cameras and microphones. AWS technology helps them simultaneously perceive the visitors in front of them, the activities in the workshop, and what the other two characters are doing, and then choose actions and alien-language lines accordingly.
AWS said the lines and actions were written by the Imagineering team, and AI guides the characters' reactions within this performance scope. The team also used the programming tool Kiro to speed up implementation. Disney hopes to use this to enable characters to interact personalizedly with different visitors; the currently announced showcase location is re:Invent.
The father of "The Sims" unveils MyMems, turning personal memories into an explorable game world

Will Wright, creator of "SimCity" and "The Sims," is working on a mobile game called MyMems that uses personal memories as material, planned for release in 2027. According to The Verge's interview with the development team, players can tell a past story to their phone and then place the generated memory fragments into the game map.
The game is developed by Gallium Studios and allows players to add AI characters representing relatives and friends. As memories accumulate, the world gradually forms a map related to personal experiences; players can set the world to private or share it with others. The team describes it as a cozy world-building game with deep social simulation.
Wright said that "The Sims" players often rebuild their own families in the game, while MyMems further centers the entire game around the player themselves. He hopes it feels like flipping through a family scrapbook: as the pages increase, characters, events and objects keep emerging, forming scenes of one's own life.
RTX Spark mini desktops launch, up to 128GB unified memory

英伟达 announced that RTX Spark compact desktops from ASUS, Dell and MSI are now available, with related models from Acer, Gigabyte, HP and Microsoft to follow. The products target local AI, model fine-tuning and creative work, and can also serve as standalone Windows computers for continuously running personal agents.
These devices are equipped with a 20-core Grace CPU, a Blackwell RTX GPU with 6144 CUDA cores, up to 128GB of unified memory, and 10-gigabit Ethernet. According to The Verge, the ASUS ProArt GR1X with 64GB memory and 1TB storage starts at $4,499.
On the software side, they support native CUDA and the use of Linux development tools via WSL. 英伟达 also expanded native Windows on Arm support for Adobe software and added DLSS 4.5 ray reconstruction in Blender 5.3; the latter requires the latest Game Ready or Studio driver.

Starbucks launches "Star Fresh Bake" at more than 100 stores, with 11 food items in the first batch

Starbucks China disclosed that its "Star Fresh Bake" series has launched in more than 100 stores across more than 30 cities nationwide, with 11 food items in the first batch, covering breakfast and afternoon tea.
This process moves baking into stores: dough is made in factories, delivered to stores as frozen ingredients via cold chain, and then store staff complete baking, cooling, shelving and same-day sales. Stores are equipped with professional ovens, frozen storage equipment and dedicated food cabinets, and staff receive corresponding operational training.
The first batch of products includes sausage puff pastry rolls, strawberry jam croissants, hazelnut cocoa spread croissants, hand-torn bagels and golden fan butterfly pastries, among others. For the two filled croissants, staff inject filling and sprinkle powder on site after fresh baking. The company said that since last September it has tested this model in nearly 10 stores, and food sales at some pilot stores nearly doubled.
Bot-controlling more than 3,000 phones to snap up LABUBU, two scalpers sentenced
According to Procuratorate Daily, the Jing'an District People's Procuratorate in Shanghai handled a case involving the use of plug-ins to bulk purchase trendy toys such as LABUBU and Hirono. The people involved could, at most, use cloud servers to control more than 3,000 phones simultaneously to place orders, and from June to August 2025 they successfully placed more than 28,000 orders in total.
Wu was responsible for developing the plug-in, renting cloud servers, purchasing algorithms, and using AI tools to write front-end code; Xia was responsible for making tutorials, recruiting "snappers" and forming purchasing groups. After the snappers logged in to the Pop Mart WeChat mini-program in advance, the program could automatically place orders at scheduled times, with a single phone sending a purchase request about every 2 seconds.
Through code analysis and real-device operation in a test environment, the procuratorial authorities confirmed that the program bypassed the platform's risk-control verification and could automatically submit and create valid orders.
On November 5, the court sentenced Wu to three years in prison with a four-year suspension and a fine of 50,000 yuan for the crime of illegally obtaining computer information system data; Xia was sentenced to three years in prison with a three-year suspension and a fine of 30,000 yuan.
G-Dragon's agency to launch robot girl group, with a concert in Macau in December first
Galaxy Corporation, the agency of G-Dragon, plans to launch a robot girl group globally in January 2027, and afterward launch a virtual AI girl group and a human girl group respectively. Galaxy founder and CEO Choi Yong-ho said in an interview with Maeil Business Newspaper that the company hopes to connect the three forms into an integrated operating team.
Before the girl group launch, Galaxy will hold a small robot K-pop concert in Macau, China, this December, and then plans to bring the performance to Dubai, Tokyo and other places. Choi Yong-ho said the robot hardware is sourced from China, the supporting software is developed by a Korean team, and K-pop content is layered on top; Unitree Robotics and Legend Star are its partners.
Galaxy already operates a robot-themed park in Seoul. According to figures disclosed by Choi Yong-ho, the park has been operating for about six months, with total visits exceeding 50,000, of which about 33,000 were paid admissions. In addition to K-pop concerts and taekwondo performances, the park offers interactive experiences such as robot operation, painting and food preparation.
Mayday concert film officially announced for Apple Vision Pro platform
Recently, the Mayday concert film was officially announced to land on the Apple Vision Pro platform, bringing fans who own this spatial computing headset a new immersive concert-viewing experience.
Unlike conventional playback on traditional flat screens or streaming platforms, the version coming to Apple Vision Pro will fully combine the headset's high-resolution Micro-OLED display and spatial audio technology. When users wear the device to watch, they can not only get a giant-screen surround view as if sitting in the front row of a concert venue, but also clearly capture the sense of direction of live instrument performances and the sound field atmosphere of tens of thousands singing together through spatial sound effects, bringing the audiovisual tension of offline tours directly into the living room.
As more and more top artists' concerts and sports events accelerate their integration into the spatial video ecosystem, the audiovisual entertainment experience of hardware devices is becoming increasingly three-dimensional.
郑渊洁 turns fairy tales into animations with AI, authorized account releases them for free
According to The Paper, 郑渊洁 resumed on November 17 a self-media account that had been inactive for two years, saying that although he does not understand animation production, he has used AI to turn his fairy tale works into animated short films.
He authorized the Douyin account "郑渊洁 Fairy Tale World" (Douyin ID 83425404836) to release these short films for readers to watch for free, and said that next he will turn "The Little Snake Who Doesn't Step on Manhole Covers" into an animation.
In the original text, 郑渊洁 expressed his judgment on the change in the threshold for creation: writers who do not understand animation can use AI to make animations, and people who do not understand literary creation may also use AI to write. He believes that enjoying cultural works for free and more people participating in creation will become a trend, and reminded writers to try paid AI before judging its impact on their profession.
Recommended Articles
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.