China AI & robotics developments, rewritten weekly-context-free for US readers. Published daily at 07:00 Beijing time.
Tue, Nov 10, 2026
Anthropic adds Sheets and whiteboard apps to ClaudeAnthropic is expanding Claude beyond chat with a spreadsheet tool called Claude Sheets and a new whiteboard app, moving the startup onto turf held by Google Workspace and Microsoft 365, according to Engadget.
The report describes the launch as a significant expansion of what Claude can do, positioning the AI assistant as a direct competitor to productivity suites from Google and Microsoft.
Engadget did not detail pricing, availability or specific features of either app in the summary provided. Anthropic, based in San Francisco, develops the Claude family of AI models and has been building out enterprise and workplace offerings.
The move follows a broader push by AI model developers into application software, where Google and Microsoft bundle assistants into their own document and spreadsheet products.
Anthropic discloses over $10 billion in new Microsoft, Google dealsAnthropic has disclosed more than $10 billion in new agreements with Microsoft and Google, while saying its previously announced deal with GPU cloud provider Lambda may be cancelled, according to The Information.
The filing-level disclosure, reported by The Information, puts fresh figures on Anthropic's cloud and compute ties to Microsoft and Google, both of which are investors in the AI developer behind the Claude models.
The possible cancellation of the Lambda agreement is notable because Lambda supplies GPU capacity that AI labs use to train and serve models. The report did not give a reason for the potential change.
Anthropic competes with OpenAI and Google in frontier model development, and its compute commitments are closely watched by investors tracking AI infrastructure spending.
BYD and Geely approached FIA about 12th F1 team, Ben Sulayem saysTwo of China's largest automakers, BYD and Geely, have approached Formula 1's governing body about entering a 12th team, FIA president Mohammed Ben Sulayem said, per Electrek. The disclosure signals Chinese EV and hybrid manufacturers are weighing a costly global motorsport push.
Ben Sulayem named the two companies while speaking at the New York Times DealBook Summit, according to Electrek. The FIA, based in Paris, oversees F1's technical and sporting rules and has been open to expanding the grid beyond its current 10 teams.
BYD, the Shenzhen-based EV and plug-in hybrid maker, and Geely, the Hangzhou-based owner of Volvo and Lotus, are China's largest automakers by volume. Neither company has confirmed a bid or commented publicly.
F1's current 10 teams include General Motors' Cadillac, which joined for the 2026 season. Any new entrant would need FIA approval and a commercial agreement with F1's rights holder.
Review tests Z.ai's ZCode coding agent on a real projectA developer's hands-on review of ZCode, the coding agent built on Z.ai's GLM-5.3 model, walks through using the tool on an actual project and assesses how it performs in practice.
The write-up, published on developer Flavio Copes' site and surfaced on Hacker News, is a first-person account of running ZCode against a real codebase rather than a benchmark.
The review focuses on the GLM-5.3 model that powers ZCode, Z.ai's coding-agent product. The post is tagged for AI coverage.
The input does not detail the project used, the results, or pricing, so specifics of the reviewer's conclusions are not summarized here.
Analysis argues US trails China in open-weight AI modelsA blog post circulating on Hacker News contends that China's leading open-weight AI models have outscored America's on a continuous basis since September 2024, framing the US as the laggard in a segment where released weights allow outside developers to inspect and build on models.
The post, published on village.dev and surfaced on Hacker News, offers no methodology in the supplied excerpt beyond the claim of a scoring gap dating to September 2024. It does not name the models or benchmarks involved.
Open-weight releases, in which developers publish model parameters for others to download and modify, have become a competitive front for Chinese labs. The piece's argument runs against the prevailing US narrative that domestic labs hold the lead.
The claims are the author's and could not be independently verified from the available text.
Mon, Nov 9, 2026
Claude Fable 5.2 tops global ranking, displacing Opus 5.5A model identified as Claude Fable 5.2 has taken the top spot in a global AI ranking, according to the Chinese tech outlet iFanr, which reported it displaced Opus 5.5 overnight. The report also cites a claim of 1,000 AI agents and a 30-hour Fermat's Last Theorem task, details that could not be independently verified.
The item, published by iFanr, gives no benchmark name, evaluation methodology, or sourcing for the ranking, and does not identify the organization that maintains it. The reference to "Opus 5.5" appears to denote a competing model version, but the input does not specify its developer.
The reported 30-hour completion of a Fermat's Last Theorem task and the figure of 1,000 AI agents are stated without technical detail. No independent confirmation was available in the source material.
Google rolls out new AI chips in bid to challenge NvidiaGoogle unveiled a new generation of its custom AI chips on Nov. 9, according to Bloomberg Tech, intensifying its effort to win customers from Nvidia, the dominant supplier of AI accelerators. The move matters for US buyers because it could widen the supply of alternatives to Nvidia's hardware.
Bloomberg Tech reported the rollout in a video segment published Nov. 9. The report did not specify chip names, performance figures, pricing or which customers would use them.
Google designs its own AI chips, known as TPUs (tensor processing units), for internal workloads and for rent through its cloud unit, positioning them as an alternative to Nvidia's GPUs.
No details on availability or manufacturing partners were given in the source material.
AI mini PC with 2017 Nvidia Tesla GPU sells for $699A mini PC pairing an Nvidia Tesla data-center GPU with 16GB of HBM2 memory is being sold for $699, according to TechRadar, which frames the 2017-era card as a bargain amid current memory shortages.
The listing centers on a Tesla-series accelerator originally built for data-center workloads, now repackaged in a small-form-factor desktop aimed at AI hobbyists and developers.
TechRadar attributes the pricing appeal to tight supplies of high-bandwidth memory, which have raised costs for newer AI hardware. The report does not name the system builder or specify the CPU and other components.
Older Tesla cards can require nonstandard power and cooling, and driver support varies by model, factors buyers typically weigh against newer consumer GPUs.
Xinhua, Ubtech unveil bionic humanoid robot news anchorsChina's state news agency Xinhua has introduced bionic humanoid robot news anchors built with Shenzhen-based Ubtech Robotics, extending a years-long push to put humanoid machines into media and service roles.
The robots were presented as on-air anchors, according to Unite.ai. Ubtech Robotics, founded in 2012 and listed in Hong Kong, is one of China's best-known humanoid and service robot makers and has supplied machines for exhibitions and commercial settings.
Xinhua has used AI-generated and robotic presenters before, part of a broader Chinese media effort to showcase domestic AI. The announcement adds to a crowded humanoid field in China that includes Unitree and Fourier Intelligence.
The item gave no deployment timeline, production figures or cost details.
Trump, Musk back Nvidia's Huang amid reported affair with alleged Chinese spyPresident Donald Trump, Elon Musk and prominent finance figures have publicly backed Nvidia CEO Jensen Huang after reports that he had an affair with a woman alleged to be a Chinese spy, according to The Guardian. The episode puts the head of the world's most valuable chip company at the center of US-China technology and security tensions.
Nvidia, based in Santa Clara, California, designs the AI accelerators that underpin most large-model training, making Huang a central figure in the US-China tech rivalry. The Guardian reported the support from Trump, Musk and finance executives without detailing the underlying allegations.
The report did not specify the timing of the alleged affair or identify the woman beyond describing her as an alleged Chinese spy. No charges or official findings were cited in the account.
Nvidia's chips are subject to US export controls affecting sales to China, a policy area where Huang has publicly engaged with Washington.
Nvidia shares fall on report of probe into Huang affair with alleged Chinese spyNvidia shares dropped after BBC Tech reported an investigation into an affair involving CEO Jensen Huang and a woman described as an alleged Chinese spy, a claim the input does not attribute to any named official or company statement.
The item, attributed to BBC Tech, gives no details on which body is investigating, what the allegations rest on, or Nvidia's response. The reported share move is likewise not quantified in the input.
Nvidia is the US chip designer whose GPUs underpin most large-scale AI training, making it central to US-China technology tensions. The input does not say whether the reported investigation is criminal, corporate, or congressional in nature.
No further facts were provided; details beyond the headline and summary could not be verified from the source text supplied.
Keln benchmarks inference rivals on DeepSeek, GLM modelsKeln published a comparison of AI inference providers — including Fireworks, Together AI, Baseten, Modal, DeepInfra and Z.ai — running DeepSeek V4.1 Flash 1017 and GLM-5.3 Flash, a vendor-authored test that gives US buyers another data point on serving costs and latency.
The benchmark was posted on Keln's own blog and surfaced on Hacker News. Keln is the author of the comparison, so results should be read as vendor-published rather than independent.
DeepSeek is the Hangzhou developer of the V-series models; Z.ai is the international brand of Zhipu AI, the Beijing company behind the GLM line. Fireworks, Together AI, Baseten, Modal and DeepInfra are US inference hosts that resell or serve open-weight models.
The summary does not disclose methodology, hardware or pricing figures, so the comparison cannot be independently assessed from the excerpt alone.
Report ties Nvidia CEO Huang to alleged Chinese spy, raising China questionsTechCrunch reported on Nov. 9 that Nvidia CEO Jensen Huang had a personal relationship with a person alleged to be a Chinese spy, a disclosure that puts new scrutiny on the chipmaker's push to keep selling AI chips to China.
Nvidia, based in Santa Clara, California, is the leading supplier of AI accelerators. Huang has publicly argued the U.S. should keep selling AI chips to China rather than cutting the market off, a stance that has drawn criticism in Washington.
The TechCrunch report, published Nov. 9, links the alleged relationship to questions about Nvidia's China ambitions. The report did not detail the nature or timing of the relationship beyond the allegation.
Nvidia's China business has been constrained by U.S. export controls. The company has not commented on the report in the provided material.
Sun, Nov 8, 2026
Cambricon chief denies 1M-chip shipment rumor, restates hardware ambitionsCambricon Technologies Chairman Chen Tianshi, 41, publicly denied reports that his company plans to ship more than one million AI chips next year, speaking at an Oct. 26 earnings briefing while reiterating the firm's aim to become a leading hardware giant of the AI era.
Cambricon, founded in 2016 and listed in Shanghai, designs AI accelerators and competes with Nvidia in China's domestic chip market. Chen, a former Chinese Academy of Sciences researcher, rarely gives interviews, according to the report from the Chinese-language outlet.
The briefing was covered by the magazine China Entrepreneur, which framed Chen's remarks around his stated ambition for Cambricon to become a major hardware player in the AI era. The report did not provide shipment figures or financial details.
The denial follows heightened investor attention on Chinese AI chip suppliers as US export controls limit access to Nvidia hardware.
AI short dramas surge on ByteDance's Hongguo, outpacing China's major streamersByteDance's short-drama app Hongguo reached 200 million daily active users by September 2026, more than the combined DAUs of iQiyi, Youku, Tencent Video and Mango TV, according to QuestMobile data cited by Huxiu. AI-generated dramas now account for most of its content, a signal for how cheap synthetic video is reshaping Chinese consumer media.
Hongguo, operated by ByteDance, the Beijing owner of TikTok and Doubao, hit 430 million monthly active users, the report said, surpassing Doubao's own MAU. The commentary, republished from a WeChat account, argues the AI-drama boom carries lessons for China's after-school tutoring industry.
The item gives no figures on AI-drama production costs or revenue, and the original author's specific claims about the education sector were not detailed in the summary.
MiniMax founder says closing final AI gap with US is hardest partMiniMax founder Yan Junjie said on a UK podcast that the last 5% of the gap between Chinese and US large models may be harder to close than the first 95%, framing the Shanghai startup's AGI path as one of catching up rather than leading.
Yan made the remarks on "Invest in Progress," a podcast from Scottish Mortgage Investment Trust, in conversation with its manager Lawrence Burns, according to Chinese tech outlet TMTPost.
MiniMax, founded in 2021, is among China's better-funded model developers and has released consumer apps and open-source models alongside its API business.
The interview focused on how MiniMax pursues AGI; the excerpt did not detail specific technical or funding plans.
ByteDance spins out five-year AI drug discovery unit into Anew LabsByteDance has carved its roughly five-year-old AI drug discovery operation out of the company and into a standalone firm called Anew Labs, according to Chinese tech outlet Huxiu, moving a long-running internal bet into a separate, externally funded entity.
The new company takes the team, algorithm platform and drug pipeline that had been built inside ByteDance, the Beijing parent of TikTok and Douyin. ByteDance retains a controlling stake, according to the report.
The article frames the spinout as part of a broader reshaping of China's AI pharma sector, with executives from large tech firms entering the field and roughly 60% of funding concentrating in leading players. The report did not disclose financial terms or the size of the investment.
Tencent AI chief says moats sit in products, not modelsTencent's chief AI scientist Yao Shunyu said at a Future Science Prize dialogue that competitive moats in AI lie more in the product layer than the model layer, according to IT Home citing Yicai.
Yao was responding to a question from Zhang Ya-qin, a Tsinghua chair professor and foreign member of the Chinese Academy of Engineering, about what actually protects a model company's business.
The report, dated Nov. 8, offered no further detail on Yao's argument or on Tencent's product strategy. Tencent operates the Hunyuan model family and WeChat; Yao joined the company as chief AI scientist.
The remark lands as Chinese model developers face price competition and rapid replication of base models, shifting attention to distribution and application layers.
China tech giants push into AI drug discoveryA TMTPost commentary examines why Huawei, ByteDance and OpenAI are moving into pharmaceuticals, framing AI drug discovery as a bet on replacing slow trial-and-error screening with computation. The piece offers analysis rather than new company disclosures.
The article argues that conventional drug hunting still relies on sifting large compound libraries through brute-force experiments, a process AI could compress. It cites Huawei, OpenAI and ByteDance as examples of technology firms entering the space.
No specific programs, investment figures or timelines are disclosed in the excerpt. ByteDance is the Beijing parent of TikTok and Douyin; Huawei is China's largest telecom equipment maker.
US readers should note the piece is opinion-driven and does not attribute concrete claims to the companies named.
Chinese AI models, led by Kimi, take 89% of open-source enterprise spendChinese AI systems, led by Moonshot AI's Kimi, account for 89 per cent of enterprise spending on open-source models, according to a South China Morning Post report surfaced on Hacker News. The figure signals that Chinese open-weight models are gaining commercial traction with corporate buyers despite US export controls.
Moonshot AI is the Beijing developer behind the Kimi chatbot. The SCMP report, cited on Hacker News, did not detail the methodology behind the 89 per cent figure or name the firms surveyed.
The number, if it holds, would mark a shift in enterprise adoption toward Chinese open-source models and away from US-led alternatives. The report did not specify a time frame or the size of the market measured.