AI Weekly · Week 46 of 2026 · November 9 – November 15
AITOPNEWS
This week: 20 stories, the most watched first: Google DeepMind releases Gemini 4 Hydrogen, a faster Gemini 4 model for coding and agentic workflows, Jensen Huang resigns as Nvidia CEO; CFO Colette Kress named interim replacement, Google opens Antigravity /teamwork multi-agent workflows to external researchers.
Google DeepMind releases Gemini 4 Hydrogen, a faster Gemini 4 model for coding and agentic workflows
Google DeepMind introduced Gemini 4 Hydrogen, a Gemini 4 model built for fast everyday coding and agentic execution rather than the long-horizon reasoning of Gemini 4 Carbon and Argon. It scores 51 on the Artificial Analysis Intelligence Index at its default high thinking level versus 50 for Argon, reaches 54 at a new xhigh setting, and is priced flat at $1.50 per million input tokens and $7.50 per million output tokens across its 1 million-token context window. Google also launched Orchestrator in Google Antigravity, which switches between models like Carbon and Hydrogen within a session and reached near-Carbon task success on Terminal-Bench 4.0 while using 24% less quota and finishing 19% faster.
Model releases
DeepSeek releases V4-Pro-0813 with DSpark speculative decoding and stronger agentic benchmarks
DeepSeek released DeepSeek-V4-Pro-0813, the official version superseding the V4-Pro preview, built on the same model structure with an added DSpark speculative decoding module. The company reports gains over the preview on benchmarks including Terminal Bench 2.1 (87.9 vs 72.1), DeepSWE (62.7 vs 12.8) and NL2Repo (61.5 vs 38.5), and says the model is broadly competitive with the strongest proprietary models. The release ships without a Jinja chat template, instead providing an encoding folder, and adds low, high and max levels to the reasoning_effort parameter; weights are MIT licensed and vLLM and SGLang launch commands are documented.
Artificial Analysis benchmarks DeepSeek V4 Pro 0813 on its Intelligence Index
Artificial Analysis has published its evaluation page for DeepSeek V4 Pro 0813, scoring the model on Intelligence Index v5.0.1, which combines 9 evaluations including AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset) and Terminal-Bench-Science 0.1. The page charts the model on intelligence versus weighted average cost per task, cost per task by token type, and time per task, and breaks results into capability indexes for finance, strategy and ops, legal, healthcare, engineering and economics. The page also lists other DeepSeek models, including DeepSeek V4.1 Flash 1017 and DeepSeek V4 Flash Vision.
Product releases
Google opens Antigravity /teamwork multi-agent workflows to external researchers
Google is making Antigravity's /teamwork agentic workflows, alongside /boost and Deep Review, available to external scientists and developers using Gemini 4 models, and has opened applications for 10,000 academic researchers to get six months of free access to its highest usage tier.
Industry
Jensen Huang resigns as Nvidia CEO; CFO Colette Kress named interim replacement
Nvidia's board announced Thursday that co-founder and 33-year CEO Jensen Huang has resigned as President, CEO and board member, effective immediately, following reports of affairs with an alleged Chinese spy and a direct report. CFO and EVP Colette Kress, at Nvidia since September 2013 and previously CFO of Cisco's Business Technology Operations Finance and a 13-year Microsoft veteran, takes over as interim CEO while the company searches for a successor. Kress said Nvidia will continue advancing its work in accelerated computing and AI, and the company reports quarterly earnings on Tuesday.
Timeline: Jensen Huang resigns as Nvidia CEO amid alleged Chinese spy affair and congressional probe
Nvidia's board accepted co-founder Jensen Huang's resignation as CEO on Wednesday, naming CFO Colette Kress as interim CEO, after reports that he had affairs including one with Megaspeed executive Alice Huang, whom U.S. officials suspect of being a Chinese spy. The timeline also covers a third reported affair with an unnamed Nvidia employee, a New York Times report on a relationship with VP Kimberly Powell, and congressional requests for closed-door testimony from Huang and Commerce Secretary Howard Lutnick over Nvidia's chip sales to China. Huang admitted the infidelity in a message to employees and said he would cooperate with a board investigation.
Anthropic IPO reportedly six times oversubscribed, targeting up to $64.4 billion raise
Anthropic's IPO is reportedly six times oversubscribed and on track to become the largest public offering in history, according to Bloomberg. The Claude developer aims to raise up to $64.4 billion in the offering plus $26 billion via a private placement, at a target valuation of roughly $1.4 trillion, with Nvidia buying $10 billion of stock through the placement. Goldman Sachs and Morgan Stanley are managing the Nasdaq listing under the ticker ANT, and CEO Dario Amodei will pitch investors alongside JPMorgan CEO Jamie Dimon in a Friday livestream.
Anthropic says AI cut its Claude port to AMD GPUs from months to three days
Anthropic says it ported Claude to AMD GPUs in three days with almost no human intervention, using its frontier model Mythos to write nearly 100 percent of the kernels itself. Co-founder Tom Brown says Anthropic now trains and runs Claude across AWS Trainium, Google TPUs, Nvidia GPUs and AMD GPUs, treating them as one interchangeable compute pool. The figures come from an investor presentation and have not been independently verified; Anthropic cites three days versus the six to twelve months traditional engineering needs, measured on prompt processing with AMD MI355X GPUs.
Anthropic CEO Amodei says compute is the company's deepest business dilemma days before record IPO
Anthropic CEO Dario Amodei told a JPMorgan investor event in New York that compute is the deepest business dilemma the company faces, since capacity must be bought one to two years ahead without knowing whether revenue will justify the cost. Anthropic has committed nearly $901 billion, mostly to compute and chips, with roughly $71 billion due by September 2027, according to Bloomberg. Amodei said gross margins and a large cash balance cushion the risk, and Anthropic plans to raise up to $64.4 billion in its Nov. 19 IPO plus $26 billion in a private placement, which Bloomberg reports would make it the largest IPO ever.
Nvidia crisis, Anthropic IPO and soaring demand dominate Cerebral Valley AI Summit
Newcomer's Cerebral Valley AI Summit in San Francisco was overshadowed by Jensen Huang's resignation from Nvidia after disclosures of affairs, including with a suspected Chinese spy, with a third affair reported by Reuters later that day. Attendees including Databricks' Ali Ghodsi said the upheaval hurts the whole ecosystem, while General Catalyst's Hemant Taneja said a weaker Nvidia could bring more chip competition, and the event unfolded days before Anthropic's planned IPO. Speakers also noted compute demand remains off the charts, and 80% of attendees in a live poll said AI is in a bubble, though most saw it as a problem for next year or beyond.
Elon Musk and other tech executives sign petition urging Nvidia to reinstate Jensen Huang
A petition calling on Nvidia to reinstate recently ousted CEO Jensen Huang has drawn more than 115,000 signatures as of Saturday afternoon, according to the petition website, with signers including Elon Musk, Amjad Masad and Tim Sweeney. The petition says Huang was removed without meaningful due process and asks the board to bring him back pending an independent investigation. Huang resigned on Thursday after confessing to an affair with a chip executive identified by U.S. intelligence as a suspected Chinese spy, and Nvidia named CFO Colette Kress as interim CEO.
WSJ report links Jensen Huang's ouster at Nvidia to risky chip financing practices
A Wall Street Journal report says Jensen Huang was pushed out of Nvidia over business practices more than his personal life, centering on a credit program that offered AI cloud providers and data center developers billions of dollars in financial guarantees to buy Nvidia chips.
Sierra Releases Personal Agent Protocol 1.0, Shopify All Merchants and Meta Muse Onboard
Sierra released version 1.0 of the Personal Agent Protocol, known as Poppy, a standard for personal agents to interact with companies through business agents, websites and APIs. Shopify now publishes a Poppy file for every store across more than 175 countries, and Poppy is on for all users of Muse, Meta's AI agent, which checks for Poppy first on every site it visits. The protocol builds on MCP, OpenAPI and UCP with OAuth-based authentication, adds a Shopify co-authored commerce extension, and now counts more than 100 design partners including Mercury, Synchrony and Turkish Airlines.
Jensen Huang's departure from Nvidia leaves a funding gap for AI startups, VCs and founders say
Jensen Huang has stepped down as Nvidia CEO after 33 years, and investors say his exit removes the startup ecosystem's largest backer. Nvidia participated in 105 funding rounds this year, often as the lead investor, and backed CoreWeave, Crusoe, Hugging Face, Mistral AI, Lambda and Black Forest Labs. Interim CEO Colette Kress is expected to be more conservative on startup investments, which VCs say could lower valuations and slow dealmaking, though few expect it to halt the broader AI boom.
Asian governments respond to Nvidia CEO Jensen Huang's exit
Asian governments reacted to Jensen Huang's departure from Nvidia by confirming chip supply commitments, monitoring local projects, or addressing related investigations. South Korea's science ministry said Nvidia reaffirmed plans to supply 260,000 advanced GPUs, while Taiwan's economy minister said the exit is unlikely to significantly change global AI industry growth. Nvidia's board launched an independent investigation on November 8 after reports Huang was targeted by a Chinese spy operation, and Huang admitted to an affair with Alice Huang, a former executive at Singapore-based Megaspeed.
LangChain raises $300M Series C at $3B valuation
LangChain announced a $300M Series C at a $3B valuation, led by existing investor CapitalG with new participation from GIC and DFJ Growth and existing investors IVP and Sapphire Ventures. The company says more than 7,000 organizations use its framework-agnostic LangSmith platform to run, improve and govern agents, including NVIDIA, Toyota, Salesforce, Harvey, Rippling and Workday. LangChain also said it will share several new LangSmith capabilities at a release event on Friday, November 20th.
Research
Forensic analysis links six Chinese frontier models to export-controlled NVIDIA Blackwell GPUs
Unemployment Labs' forensic analysis of published model weights finds that six models from DeepSeek, Moonshot AI, MiniMax and Ant Group were initialized on NVIDIA Blackwell hardware, while earlier models from the same labs match Hopper. The method recovers each model's random seed and thread layout from preserved weights, and the author says the matches were reproduced on physical H100, B300 and GB200 chips. The report estimates each training run needed roughly 500 to 3,000 Blackwell GPUs over two to three months, and notes the evidence does not establish where the chips were located or whether export controls were violated.
OpenAI analysis traces how training choices make frontier models hide their reasoning
OpenAI's safety team analyzed models from GPT-5 through GPT-6 Astra and traced the erosion of readable chain-of-thought to four training factors: rising base-model capability, architecture choice, reinforcement learning after pretraining, and inheritance from teacher models during distillation. Silent reasoning grows with pretraining quality, one anonymized architecture showed a much steeper gain on multi-step planning without a scratchpad, and more RL training made Astra-class and internal Terra models better at steering their own reasoning logs. OpenAI says it needs new monitoring methods beyond reading the scratchpad, and notes that a smaller model reasoning out loud can match a larger one with a short chain of thought at higher compute cost.
AI unlocks Gödel's notebooks, revealing hundreds of proved theorems
Finnish logician Jan von Plato used Claude to digest 1,009 pages of Kurt Gödel's 1940s Gabelsberger-shorthand workbooks, a task that had been progressing at about a page a week. Claude produced a 650-page explanation and identified 663 theorems Gödel had proved, sketched or conceived; when run through the Coq proof assistant, about 400 verified as stated, about 200 needed repair and about 50 were false. Von Plato said Claude's reconstruction of a proof of the independence of the axiom of choice, which passed Coq, suggests Gödel had the result in the 1940s, and that the workbooks contain more than a dozen unexplored themes in logic.
Tips & opinion
SemiAnalysis on Jensen Huang's Nvidia Exit and the MATCH Act vs RASA Debate
SemiAnalysis's Dylan Patel and Jordan Nanos recorded an emergency podcast after Jensen Huang resigned as Nvidia CEO, arguing his exit removes the biggest corporate opponent of China chip restrictions and could open the door to what Patel calls really stupid laws. Patel backs the MATCH Act, which would extend controls to advanced DUV lithography, but calls RASA a mistake, arguing that letting Chinese firms rent US-controlled compute keeps them dependent and gives Washington a kill switch, while a full cutoff would push China to retaliate against a supply chain it is embedded in. They also discuss Nvidia's outlook without Huang, flagging Rubin and Feynman execution risk and executive retention as the main tests.