Daily Briefing

A blog post from Fig argues that open-ended data, rather than model capability, is the constraint on AI systems performing real knowledge work. The claim, surfaced on Hacker News, adds to debate over what blocks enterprise AI adoption.

Tencent has agreed to lease roughly 100,000 AI chips from Oracle under a deal valued at about $7 billion, according to a report, a move that would give the Chinese internet giant large-scale compute capacity from a US supplier.

US robotics startup Figure has published a video in which its F.02 humanoid walks into a furnace, a send-off the company said was suggested by Arnold Schwarzenegger himself. The stunt marks the end of the robot's service as Figure shifts resources to its newer F.03 fleet.

A Chinese state-backed company has disclosed a deal involving Nvidia's Blackwell chips, according to Bloomberg Tech, a transaction that could draw scrutiny in Washington given ongoing US export controls on advanced AI silicon to China.

Nvidia is drawing scrutiny over cases involving the alleged smuggling of its AI chips into China, according to Bloomberg Tech, a focus for US policymakers weighing export controls on advanced semiconductors.

Meta released Muse, its first AI agent product, on Sept. 8; it hit No. 1 on the US App Store free chart with over 2.5 million downloads and 642,000 US mobile daily actives, the Chinese tech outlet Huxiu reported, citing the article's author Shen Huaizheng.

Tencent has agreed to spend roughly $7 billion on compute services from Oracle, with data centers in Southeast Asia and about 100,000 advanced chips involved, according to foreign media reports cited by Chinese outlet Huxiu. The deal lands as US export controls restrict where high-performance chips can go, including shipments to third countries tied to China-based firms.

The Federal Trade Commission has opened an investigation into OpenAI, Anthropic and other artificial intelligence companies over potential dangers posed by their products, an agency spokesperson said, according to CNBC. The inquiry adds a new regulatory front for US AI developers already facing scrutiny over safety and competition.

The US Federal Trade Commission has launched an industry-wide investigation into Anthropic, OpenAI and other AI developers, according to the Guardian, examining potential harms from their models. The inquiry adds regulatory scrutiny for the sector's largest US players.

Google is facing internal skepticism from employees about its newest Gemini model, according to Bloomberg Tech, a signal of pressure inside the company as it competes with rivals in frontier AI.

The US Federal Trade Commission is escalating a broad investigation into Anthropic, OpenAI and other frontier AI labs, planning subpoena-like demands for information on potential consumer harms, according to government officials cited by the New York Post. The move signals tighter US scrutiny of the companies whose models underpin much of the commercial AI boom.

Baidu is framing "context" — a user's files, chat history and accumulated work — as the key competitive variable in AI-powered office software, according to an interview with a company executive published by Leiphone on Aug. 14.

Figure AI's Helix 2.5 model sent robots into 30 unfamiliar US homes to make beds and fold towels, while China's Qiyuan launched two personal robots priced at roughly $2,800 to $4,200, highlighting a divide over whether home robots should work or keep company.

Huawei executive Jin Yuzhi, who heads its Yinwang auto software unit, said automakers should scale back self-developed driver-assistance systems because sustained R&D costs outstrip what their sales can support, according to an interview published Sept. 28 by Huxiu.

Grasshopper Manufacture, the Japanese studio behind the No More Heroes series, said it will split from Chinese publisher NetEase Games and return to independent operation on Oct. 1, 2026, according to Insider Gaming.

DeepSeek, the Hangzhou-based AI developer known for its low-cost reasoning models, has released open-source software tools designed to run on Huawei's Ascend AI chips, according to The Information. The move signals a push to build a domestic alternative to Nvidia's CUDA software ecosystem, which remains the industry standard for training and running AI models.

Anthropic has raised concerns that Z.ai's open-weight GLM-5.3 model nearly matches its Claude Mythos on cyber capabilities, a warning that puts a Chinese open-source release at the center of the AI security debate in Washington.

Apple has set aside an internal plan that could have cut roughly 5,000 AppleCare support jobs by shifting customer service work to AI, according to Bloomberg's Mark Gurman. The reversal, reported Sept. 29, signals that even the largest US tech firms are finding full AI substitution of human support harder to justify than expected.

China's generative AI user base has surpassed 700 million people, more than half the country's population, according to data reported by the South China Morning Post. The figure underscores how quickly AI tools have reached mainstream scale in the world's second-largest economy, a market US AI developers and investors watch closely.

Former SEC chair Gary Gensler said Chinese AI models pose a significant challenge to US artificial intelligence companies, according to Bloomberg Tech, underscoring how open-weight releases from China are reshaping competitive dynamics in the sector.