# AI news brief: Sunday, November 15, 2026

22 stories, most important first. Status is how solid each one is: official, confirmed, reported or rumor.

- [Nvidia opens probe into CEO Jensen Huang after US spy allegations](https://ai-news.ikangai.com/story/nvidia-opens-probe-into-ceo-jensen-huang-after-us-spy-allegations/index.md): reported. Nvidia opened an independent investigation into CEO Jensen Huang after U.S. intelligence agencies alleged that a woman he had an affair with is a Chinese spy.
- [OpenAI says GPT-6 Astra helped turn a lattice Standard Model idea into a detailed proof](https://ai-news.ikangai.com/story/openai-says-gpt-6-astra-helped-turn-a-lattice-standard-model-idea-into-a-detaile/index.md): reported. OpenAI describes a candidate nonperturbative lattice definition of the Standard Model and says its GPT-6 Astra model helped develop it into a detailed proof.
- [Anthropic's Fable 5.2 completes 41.3% of real freelance projects in Remote Labor Index](https://ai-news.ikangai.com/story/anthropic-s-fable-5-2-completes-41-3-of-real-freelance-projects-in-remote-labor/index.md): reported. Fable 5.2 completed 41.3% of real freelance projects to a professional standard in the Remote Labor Index, up from a top score of 2.5% a year earlier.
- [Google's Gemini 4 Carbon adds teams of agents that work together on tasks](https://ai-news.ikangai.com/story/google-s-gemini-4-carbon-adds-teams-of-agents-that-work-together-on-tasks/index.md): reported. Google's Gemini 4 Carbon introduces teams of agents that can work together on tasks, as major model releases keep accelerating.
- [US data brokers sell roughly $500M of RL training environments to Chinese labs](https://ai-news.ikangai.com/story/us-data-brokers-sell-roughly-500m-of-rl-training-environments-to-chinese-labs/index.md): reported. US data brokers are reportedly selling Chinese AI labs roughly half a billion dollars in reinforcement learning training environments.
- [Google ships four TPU variants in a year as it steps up Nvidia competition](https://ai-news.ikangai.com/story/google-ships-four-tpu-variants-in-a-year-as-it-steps-up-nvidia-competition/index.md): reported. Google is shipping four TPU variants in a year as it steps up competition with Nvidia in AI computing.
- [Stanford HAI: 2026 AI investment tops $1 trillion as youth employment declines deepen](https://ai-news.ikangai.com/story/stanford-hai-2026-ai-investment-tops-1-trillion-as-youth-employment-declines-dee/index.md): official. Stanford HAI says global corporate AI investment has passed $1 trillion in 2026 and that job declines in AI-exposed occupations have continued into this year.
- [Google DeepMind agents write Lean-verified quantum programs beating prior baselines](https://ai-news.ikangai.com/story/google-deepmind-agents-write-lean-verified-quantum-programs-beating-prior-baseli/index.md): official. Google DeepMind and Google Quantum AI build a quantum program library in the Lean theorem prover, where teams of Gemini 4 agents write and prove programs.
- [SII and GAIR apply agent-driven self-improvement to a full pre-training run](https://ai-news.ikangai.com/story/sii-and-gair-apply-agent-driven-self-improvement-to-a-full-pre-training-run/index.md): official. Researchers from SII and GAIR apply recursive self-improvement to an entire frontier pre-training run from scratch and report the details.
- [Anthropic spends $3.4B on Caterpillar gas turbines for 1.7 GW of power](https://ai-news.ikangai.com/story/anthropic-spends-3-4b-on-caterpillar-gas-turbines-for-1-7-gw-of-power/index.md): reported. Anthropic is spending $3.4 billion on Caterpillar natural gas turbines with a capacity of 1.7 gigawatts.
- [Test finds AI coding agents needed human cleanup on nearly three in four tasks](https://ai-news.ikangai.com/story/test-finds-ai-coding-agents-needed-human-cleanup-on-nearly-three-in-four-tasks/index.md): reported. In a test of real work, AI coding agents needed human cleanup on nearly three out of four tasks, which shows that producing code is not the same as delivering finished work.
- [DeepMind tests Gemini Robotics 2.5 on a Mercedes-Benz production line](https://ai-news.ikangai.com/story/deepmind-tests-gemini-robotics-2-5-on-a-mercedes-benz-production-line/index.md): reported. Gemini Robotics 2.5 is being tested on a Mercedes-Benz production line, where it reached a 96% success rate sorting metal parts in a pilot.
- [UW and Allen AI's s3 lets multi-agent teams learn their own organization](https://ai-news.ikangai.com/story/uw-and-allen-ai-s-s3-lets-multi-agent-teams-learn-their-own-organization/index.md): official. Researchers from the University of Washington and the Allen Institute for AI introduce s3, a reinforcement learning method that lets agents with no preset roles learn to coordinate.
- [SoMem lets agents organize their own memory as linked documents](https://ai-news.ikangai.com/story/somem-lets-agents-organize-their-own-memory-as-linked-documents/index.md): official. City University of Hong Kong and Tencent Jarvis Lab propose self-organized memory, in which agents decide how to summarize and link their own notes.
- [LongCapHarness boosts frozen models' long-context reasoning without training](https://ai-news.ikangai.com/story/longcapharness-boosts-frozen-models-long-context-reasoning-without-training/index.md): official. Microsoft Research Asia proposes LongCapHarness, which builds a harness around a frozen model rather than post-training it.
- [Berkeley and Illinois study finds agent costs come from context reads and tool calls](https://ai-news.ikangai.com/story/berkeley-and-illinois-study-finds-agent-costs-come-from-context-reads-and-tool-c/index.md): official. UC Berkeley and the University of Illinois Urbana-Champaign analyze agent cost and build Agency, a layer that cuts cost without changing the model.
- [KAIST's WikiTrace lets agents check stored lessons against their source runs](https://ai-news.ikangai.com/story/kaist-s-wikitrace-lets-agents-check-stored-lessons-against-their-source-runs/index.md): official. KAIST researchers describe memory-induced degradation and introduce WikiTrace, a memory system that links each lesson to the run it came from.
- [Google's ReproBridge raises agent ML reproduction success from 76.4% to 92.6%](https://ai-news.ikangai.com/story/google-s-reprobridge-raises-agent-ml-reproduction-success-from-76-4-to-92-6/index.md): official. Google researchers introduce ReproBridge, which packages one successful reproduction of an ML experiment so the next agent can reuse it.
- [Meta's AutoRefine recalibrates LLM code judges as coding agents change](https://ai-news.ikangai.com/story/meta-s-autorefine-recalibrates-llm-code-judges-as-coding-agents-change/index.md): official. Meta researchers present AutoRefine, a method for keeping LLM code judges calibrated as the coding agent they evaluate evolves.
- [Robbyant's LingBot-Touch gives different tactile sensors one shared representation](https://ai-news.ikangai.com/story/robbyant-s-lingbot-touch-gives-different-tactile-sensors-one-shared-representati/index.md): official. Robbyant introduces LingBot-Touch, a tactile foundation model that maps piezoresistive, magnetic and visuotactile sensors to one virtual fingertip.
- [datasette-apps 0.4a0 adds linking between apps and reports URL schemes to Datasette Agent](https://ai-news.ikangai.com/story/datasette-apps-0-4a0-adds-linking-between-apps-and-reports-url-schemes-to-datase/index.md): official. The 0.4a0 alpha of datasette-apps lets sandboxed apps link to each other through URL fragments and reports supported URL schemes to Datasette Agent.
- [datasette-agent 0.6a0 adds an agent prompt to the Datasette homepage](https://ai-news.ikangai.com/story/datasette-agent-0-6a0-adds-an-agent-prompt-to-the-datasette-homepage/index.md): official. The 0.6a0 alpha release of Datasette Agent adds a prompt box to the Datasette homepage and supports hash-mentions of tables and queries.

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.
