Monday, Nov 9, 2026
AI;DR
2026-11-09 · 15 stories- Anthropic signed a $3.4B deal, first reported by the Financial Times, to generate natural-gas power for its data centers. The arrangement is meant to secure a stable supply for its AI operations.
- anthropicGoldman Sachs is pulling Claude Code from its bankers and traders. The move changes access to Anthropic’s coding tool inside the bank, though no reason was given.
- regulationOfficials at Commerce Department AI safety center CAISSI were asked to informally report to a White House office, according to Garance Burke of the Associated Press. The request raised concerns about CAISSI’s independence.
- agentApple is targeting an autonomous Siri launch in 2027, possibly with iOS 28, to close its AI gap with Meta and OpenAI. John Ternus pressed Craig Federighi to add computer-use features that let Siri operate apps independently.+1
- anthropicMETR ruled out compute growth as the main cause of Anthropic’s roughly 2x AI progress speedup, estimating capability gains ran 1.9x faster since October 2025. Progress per dollar still fell 63% between early 2025 and 2026.
- fundingExcluding Anthropic’s IPO, AI companies sold or priced more than $300B of stock in 2026, with about $100B more pending. Alex Eichenstein of The Information said the pace raises questions about market capacity.
- deepseekDeepSeek’s open-source Harness AI app reached 1.54M daily active users by Nov. 8, up 545% in just over a month. About 60% used one of its roughly 35,000 plugins, helped by Creator Mode.
- Hedge funds extended their AI-stock selling streak to six weeks, with the largest 20-trading-day exit on record. In October, they cut net leverage 4.2 points to 73.4% and sold about $20B in technology shares, Goldman Sachs data shows.
- Llama-modchip ran 284B DeepSeek-V4-Flash on 59GiB of RAM at 44.4 tok/s prompt processing and 9.1–14.7 tok/s generation, using a 64GB CPU with 256K context and no GPU. It also ran 30B MoEs at 25–111 tok/s.
- Alibaba’s Co-Design Bench found coding agents still need human cleanup on nearly three of four real-world tasks. Even GPT-6 Astra solved only 26.4%, while the same models scored above 99% on older benchmarks.
- Anthropic-backed AI-safety super PAC Public First spent more than $63M in the 2026 midterms, and 10 of its 17 congressional candidates won. Bloomberg reported the gains came in races favoring stricter AI oversight.
- IndustrySilicon Valley says AI will cure disease, but drugmaker CEOs are not convinced. Their skepticism challenges the industry’s promises about AI’s medical impact and the speed of future breakthroughs.
- A small, wealthy AI-safety network organized in Signal chats and grew into a coalition that defeated the $100M super PAC Leading the Future in the 2026 midterms. Politico traced the shift in influence to the group’s organizing.
- regulationRights holders for Elvis and Eminem are suing Suno, adding two more lawsuits over AI music clones. The cases increase the legal pressure on the AI music company and its cloning practices.
- Inner Mongolia is China’s fastest-growing AI data-center region, with low construction costs and no public backlash. A wind-powered 800MW campus built 120MW within a year, alongside a 5MW SST-native 800V DC pilot and chiller-free cooling.