AI News: November 15, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. More Than 115,000 People Sign a Petition Asking Nvidia to Reinstate Jensen Huang, Including Elon Musk and Garry Tan
Nvidia. A petition calling on Nvidia’s board to bring back Jensen Huang had attracted more than 115,000 signatures as of Saturday afternoon, TechCrunch reported. The petition site lists 5,965 current Nvidia employees, 1,108 former employees and 63,063 shareholders among the signers, with Elon Musk, Replit CEO Amjad Masad, Epic Games CEO Tim Sweeney and Y Combinator CEO Garry Tan named as notable supporters. It says Huang was removed “without meaningful due process” and asks the board to reinstate him while an independent investigation establishes the facts. Huang resigned on Thursday and CFO Colette Kress is interim CEO. Source
2. Asian Governments Respond to Huang’s Exit by Confirming Chip Supply and Probing Megaspeed
Nvidia. TechCrunch surveyed reactions across Asia. South Korea’s Science Ministry said Nvidia reaffirmed its plan to supply 260,000 advanced GPUs and continue joint AI projects, and the government’s AI strategy council held an emergency meeting on Friday. China’s foreign ministry rejected allegations that a businesswoman linked to Huang was a spy and said it was not aware of reports tying restricted Nvidia chips to a Shanghai data center associated with her. Reuters, cited by TechCrunch, reports that Huang’s exit creates uncertainty over more than $3 billion in Chinese AI computing projects that depend on Nvidia’s restricted processors. Singapore’s Ministry of Home Affairs said its investigation into Megaspeed and six other individuals is ongoing. Source
3. TechCrunch Asks What Nvidia’s Security Team Knew, After a Chief Security Officer Change in August
Nvidia. Nvidia’s longtime chief security officer, David Reber, left in August citing health reasons, and was replaced by Saša Zdjelar, who previously worked at ExxonMobil, Salesforce and ReversingLabs. TechCrunch notes that nothing public connects either change to the Megaspeed matter or to Huang’s relationships, and that Nvidia and Reber did not answer questions. The piece quotes a retired intelligence officer who says foreign services gather leverage mostly through ordinary conversation rather than hacking, and that security teams are often underfunded and defined too narrowly to catch it. It also points out that Nvidia’s top lawyer was engaged with the government’s Megaspeed inquiry more than a year ago, according to The New York Times. Source
4. Amodei Calls Compute “the Deepest Business Dilemma” at a JPMorgan Event Days Before Anthropic’s IPO
Anthropic. Speaking with JPMorgan CEO Jamie Dimon on Friday, Dario Amodei said compute must be bought one to two years ahead, before anyone knows whether revenue will justify it, and that buying too much risks financial overextension. Bloomberg, cited by The Decoder, puts Anthropic’s commitments at nearly $901 billion, mostly compute and chips, with about $71 billion due by September 2027. Amodei named gross margins and cash on the balance sheet as the buffers, and said Anthropic buys relatively little compute upfront and can turn to the spot market. Anthropic plans to raise up to $64.4 billion in the IPO plus $26 billion in a private placement, with a planned Nov. 19 listing, and reported an $81 billion annualized revenue run rate at the end of October. Source
5. Gemini 4 Carbon Takes the Top Spot on MathArena, at Roughly 15 Times the Cost of GPT-6.1 Sol
Google. Gemini 4 Carbon is the first non-OpenAI model in a while to lead the MathArena math benchmark, displacing GPT-6.1 Sol, with Claude Fable 5.2 in second place. With the reasoning settings used in testing, Carbon and Fable 5.2 cost about 15 and 40 times as much as GPT-6.1 Sol. Carbon’s most common failure is exceeding the time limit, and in one case it launched a script that ran for more than eleven hours without checking its progress. The Decoder notes that OpenAI’s strongest math results are not yet publicly available. Source
6. OpenAI Folds Its Robotics Researchers Into Research and Applied Teams, Betting General Models Can Drive Robots
OpenAI. OpenAI reorganized its robotics group, moving researchers to the main research team under Chief Research Officer Mark Chen and the applied robotics and hardware teams into Greg Brockman’s Applied organization. OpenAI has since clarified that robotics work continues, after early reports described the team as disbanded. OpenAI apparently believes general models such as GPT-6 Astra can control robots directly. In an Nvidia study cited in the piece, an Astra-powered agent reached a 74.8 percent success rate on simulated tasks against 51 percent for the best specialized models, in simulation only. Source
7. Runway Open-Sources Praxis-1, a Video-Pretrained Model That Controls Robots, Under Apache 2.0
Runway. Runway released the weights and code for Praxis-1, which takes camera images, text instructions and robot position data and predicts both how the scene will change and the robot movements to produce it. It was introduced in late September with partner-only access. On Nvidia’s RoboLab-120 simulation leaderboard Praxis-1 scored 46.2 percent, ahead of FLUX 3 Action at 42.9 percent among World Action Models, and on a real Franka arm it succeeded 88 times in 100 attempts against 80 for FLUX 3 Action. Runway ran all tests itself and they have not been independently verified. Adapting the model still needs robot data, for example about 277 hours of DROID recordings for the Franka arm. Source
8. Former Anthropic Researcher Jacob Coxon Says an IPO Roadshow Slide Shows Algorithmic Efficiency Is Accelerating
Anthropic. Jacob Coxon, who previously trained models at OpenAI and then Anthropic, says a compute-versus-time chart in Anthropic’s IPO roadshow materials was a key reason he left after about four months. He argues the compressed axis makes a 10x and a 100x increase look similar, and that even so the curve implies efficiency gains are speeding up, so Anthropic could retrain Mythos with a fraction of the original compute. He also says algorithmic improvements are “on the verge of being automated” and calls for a pause in frontier model development next year. These are the claims of a former employee as reported by The Decoder, and Anthropic’s response was not included. Source