A Guth Labs publication ↗

LG Electronics announced on November 16 that it signed a November 13 agreement with the Korea Institute for Robot Industry Advancement, or KIRIA, to develop humanoid robot safety and usability evaluations. The distinction matters: this is a newly disclosed agreement, not a certification awarded today. LG announcemen…

Sources: lg.co.kr, m.ajupress.com

Models

More models →

Agents

More agents →

Tools

More tools →

Research

More research →

Policy

More policy →

Business

More business →

Infrastructure

More infrastructure →

Robotics

More robotics →
All updates →
  1. Models

    LongCat-Video-Avatar 1.6 releases weights for longer audio-driven videos

    Meituan’s LongCat team announced LongCat-Video-Avatar 1.6 on November 15, updating its audio-driven human-video framework. The official repository describes more natural speech-related mouth motion, coordinated facial and body movement, and less a…

    Sources: github.com, huggingface.co
  2. Guth News

    Langfuse’s scheduled Cloud v4 cutoff puts legacy evaluations and ingestion on the migration checklist

    November 16 is the published cutoff date for Langfuse Cloud’s move to v4-only operation. The vendor’s August 17 announcement schedules removal of legacy APIs, features and ingestion for today. This is an effective-date update to an existing migrat…

    Sources: langfuse.com
  3. Guth News

    AMD confirms World Labs acquisition close as Fei-Fei Li joins its leadership

    AMD announced on November 16 that it had completed its acquisition of World Labs. The company’s announcement says Fei-Fei Li is joining AMD as executive vice president and chief scientist, reporting to Lisa Su. The announcement date and closing da…

    Sources: newsroom.amd.com, ir.amd.com
  4. Models

    Tencent launches Hy4 flash with lower API prices and a temporary free-access offer

    Tencent released Hy4 flash on November 16, positioning the model for everyday work rather than its most demanding software-engineering tasks. The company’s announcement describes a multimodal model with a one-million-token context window, 224 bill…

    Sources: tencent.com
  5. Models

    OpenAI names four organizations for embedded safety assessments

    OpenAI said on November 16 that it has signed agreements with Apollo Research, METR, Transluce and Goodfire for embedded assessments of model alignment and safeguards. The program gives outside assessors scoped access to internal systems and evide…

    Sources: openai.com
  6. Guth News

    Slurm update aims to bring quantum processors into familiar supercomputing workflows

    NVIDIA announced expanded quantum-computing support for Slurm on November 16, aiming to let supercomputing centers schedule quantum processors alongside their existing computing resources. Slurm 26.11 is expected later this month, according to the…

    Sources: blogs.nvidia.com
  7. Guth News

    NVIDIA outlines Rubin HPC and Vera CPU plans for scientific computing

    NVIDIA announced Rubin HPC at SC26 on November 16, describing a GPU designed for simulation and data processing in science, engineering and analytics. It plans more than 50,000 CUDA cores, sparse-data acceleration and seventh-generation NVLink con…

    Sources: blogs.nvidia.com, nvidianews.nvidia.com
  8. Models

    NVIDIA packages Nemotron models and chemistry data for scientific agents

    NVIDIA introduced Nemotron for Science at SC26 on November 16, combining its Nemotron 3.5 Super VL model with scientific evaluations, chemistry data and post-training recipes. The company says the model and LitmusBench v0.2 dataset are available o…

    Sources: nvidianews.nvidia.com
  9. Models

    AWS reports AI-agent campaigns aimed at exposed model credentials

    Amazon’s threat-intelligence team described cybercrime campaigns involving ARTEX AI agents in a November 13 report. Between October 3 and November 5, its MadPot decoy sensors received requests containing agents’ campaign memory, including accounts…

    Sources: aws.amazon.com
  10. Models

    Cloudflare joins SGLang’s new industry consortium, while maintainers retain technical control

    guthlabs.newsroom | November 16, 2026 LMSYS announced the SGLang Consortium on November 12 with NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk as founding members. The forum is intended to turn deployment experience into engineer…

    Sources: lmsys.org
  11. Agents

    Superfluid research targets the queue between local agents and their human user

    guthlabs.newsroom | November 16, 2026 A technical report submitted November 13 describes Superfluid, a local inference server designed to keep an interactive chat responsive while agents share its computing resources. Rather than replacing model r…

    Sources: arxiv.org, github.com
  12. Models

    Alibaba details Qwen4-Max and promises smaller open-weight models

    Alibaba has supplied a formal launch announcement for Qwen4-Max, adding architecture details and an open-weight roadmap to the production-API documentation Guth News covered earlier. In its November 16 announcement, Alibaba describes a mixture-of-…

    Sources: alibabacloud.com, qwen.ai +3
  13. Models

    Barr warns that shared AI models could turn individual trades into a systemic risk

    Federal Reserve Governor Michael Barr used November 16 prepared remarks to highlight a problem that individual AI evaluations can miss: many traders using similar models may react together. In his speech text, Barr described the possibility that m…

    Sources: federalreserve.gov, reuters.com
  14. Models

    GitHub Agentic Workflows changes model-routing defaults and connector setup

    GitHub Agentic Workflows v1.2.3 changes how eligible workflows choose models and connect to external work-tracking systems. The official release is dated November 13 and marked pre-release. This is newly discovered coverage, not a November 16 soft…

    Sources: github.com
  15. Models

    Huawei launches CloudRobo-WM1.2, but downloadable weights remain unverified

    Huawei Cloud presented its CloudRobo-WM1.2 robotics world model on November 16 and announced plans to open its weights and code, according to Cailian Press's post-event report. Huawei's official event agenda scheduled the model presentation and a …

    Sources: huaweicloud.com, m.cls.cn +1
  16. Models

    Craig details GPT-6 Astra’s role in a candidate lattice Standard Model

    Nathaniel Craig has described how he used GPT-6 Astra to develop a candidate lattice definition of the Standard Model. His account, published on OpenAI’s science blog on November 15, brings a research workflow into view, rather than announcing a n…

    Sources: openai.com, arxiv.org
  17. Models

    Apple opens a health-AI study focused on trust, comprehension and everyday use

    Apple announced the Apple Health AI Usage Study in the United States on November 15, pairing a virtual study and app with AI technology from OpenEvidence. The company says the research will examine user trust, response quality, comprehension and e…

    Sources: apple.com
  18. Models

    Qwen4-Max moves beyond preview with a documented production API

    Alibaba Cloud now documents a production qwen4-max endpoint and recommends it over the preview model. Its production model guide lists text, image and video input, text output, tool calling and a million-token context window. This is a substantive…

    Sources: help.aliyun.com, guthlabs.ai
  19. Agents

    Chrome DevTools MCP adds logpoints and network mocks for coding agents

    Chrome DevTools MCP’s November 11 release gives browser-connected coding agents two practical debugging additions: logpoints and page-specific network mocks. Version 1.13.0 also supports upstream mock responses, according to the project’s official…

    Sources: github.com
  20. Agents

    Soteria pairs verified plans with runtime checks to keep AI agents on task

    A November 13 research paper proposes a different approach to agent guardrails: give the system a checked plan it can return to when an action is blocked, rather than merely telling it to stop. Soteria, by Mike He, Ankush Desai, Sharad Malik and A…

    Sources: arxiv.org
  21. Models

    Huawei brings openPangu 2.1 Pro to MaaS, with 512K context and a developer focus

    Huawei Cloud has added openPangu-2.1-Pro to its model-as-a-service platform, giving developers a new hosted option aimed at agent workflows and code generation. The release record identifies version 20261115 and availability in the Southwest–Guiya…

    Sources: support.huaweicloud.com, ithome.com
  22. Agents

    E4S 26.11 brings agent tools and browser desktops to scientific-computing containers

    E4S is bringing more AI development tools into its scientific-computing software stack. A publicly accessible release announcement dated November 16 says version 26.11 is available with more than 125 HPC and AI packages. The project lists Hermes A…

    Sources: e4s.io
  23. Tools

    Philippines creates national AI council and adopts governance framework

    Philippine President Ferdinand Marcos Jr. signed Executive Order 131 on November 12, adopting an AI governance framework and creating a National Artificial Intelligence Council, according to the government’s Philippine News Agency. The signing was…

    Sources: pna.gov.ph, philstartech.com
  24. Agents

    SciMatrix 1.5 puts researchers inside multi-agent scientific discussions

    The Chinese Academy of Sciences’ Computer Network Information Center announced SciMatrix 1.5 on November 11, adding human intervention, selectable agent discussion rooms and a pilot collaborator-recommendation feature. This is newly covered here, …

    Sources: www1.cstcloud.cn
  25. Models

    AutoGPT adds remote Local PC execution targets in its latest beta

    AutoGPT’s platform beta 0.11.3, dated November 15, adds remote Local PC execution targets alongside native-mobile server foundations and support for Google’s Gemini 4 Hydrogen and Carbon models. The maintainers list the changes in their release no…

    Sources: github.com
  26. Agents

    Qwen Code adds temporary directory grants and memory-import controls

    Qwen Code released version 0.27.4 on November 15, extending managed-agent access while adding more explicit controls around risky actions. The project’s release page lists temporary local-directory grants, an interactive managed terminal and typed…

    Sources: github.com
  27. Models

    llama.cpp merges MiniMax-M3.1 support into its prerelease builds

    llama.cpp has merged support for MiniMax-M3.1 into its main development branch. Pull request #32301 was merged on November 15, and the project's release listing includes the change in prerelease build b12420. This is a runtime-compatibility update…

    Sources: github.com
  28. Agents

    Mercor raises $750 million to expand AI training and workplace evaluations

    The $20 billion valuation comes as Mercor takes its human-expertise business deeper into enterprise agents. Mercor announced a $750 million Series D on November 10 at a $20 billion valuation, as the AI training company expands its work helping bus…

    Sources: mercor.com, reuters.com
  29. Models

    Hugging Face publishes a recipe for shrinking multi-vector search

    Hugging Face published a practical guide on November 11 showing how retrieval models can learn to preserve useful information when their document representations are compressed. Tom Aarsen's Sentence Transformers example builds on earlier pooling …

    Sources: huggingface.co, sbert.net +1
  30. Tools

    Helsing launches Clarity military intelligence platform, with deployment details still undisclosed

    Helsing unveiled Clarity in Berlin on November 11, pitching a European-built AI intelligence platform that runs on military customers’ own infrastructure. The announcement moves the defence technology company further into software for interpreting…

    Sources: helsing.ai, reuters.com
  31. Models

    Hedra’s Model Factory pairs research agents with human-controlled deployment

    Hedra launched Model Factory on November 10, offering teams a managed way to adapt and run models for robotics, manufacturing and other physical-world applications. Its announcement describes GPU capacity, post-training and deployment as parts of …

    Sources: globenewswire.com, hedra.com +1
  32. Agents

    Google Workspace’s agent OAuth cutoff has passed: check the client, not just the token

    Google's previously announced OAuth transition for Workspace MCP integrations has passed its November 12 cutoff date. The company's October 12 release notes said rejection of requests from non-agent-designated OAuth clients would begin November 11…

    Sources: developers.google.com, support.google.com
  33. Agents

    Poppy 1.0 gives personal agents a common way to approach businesses

    Sierra announced Personal Agent Protocol 1.0 on November 13, expanding its initiative for interactions between personal agents and businesses. The company says Shopify stores now publish Poppy files automatically and Meta's Muse looks for the prot…

    Sources: sierra.ai, personalagentprotocol.org +1
  34. Guth News

    Dublin’s AI discussion becomes a work programme, not a funded buildout

    Publication review: November 15, 2026. Event: November 13, 2026. The earlier unpublished preview described an upcoming discussion; this article incorporates the subsequent results. The European Council now publishes results and António Costa’s clo…

    Sources: consilium.europa.eu
  35. Models

    Inkling 2 Small moves from hosted access to downloadable weights

    Thinking Machines Lab has released downloadable weights for Inkling 2 Small, changing the access story from its November5 hosted introduction. A clearly dated November12 update on the original announcement says the weights are available on Hugging…

    Sources: thinkingmachines.ai, huggingface.co +2
  36. Business

    Hesai reports rising robot-component shipments as its physical-AI business expands

    Hesai’s November 10 earnings release provided a new measure of its expansion beyond lidar: the company reported shipping 23,405 robotic actuation modules in the third quarter, up 96.8% from the previous quarter. Cumulative shipments reached 35,295…

    Sources: investor.hesaitech.com, hesaitech.com
  37. Agents

    Haystack 3.5 adds structured agent replies and new routing components

    Deepset’s Haystack 3.5.0 is available, adding controls for agent responses and tools for routing data through AI pipelines. GitHub dates the release to November 10, and PyPI lists the corresponding package files on the same day. GitHub release PyP…

    Sources: github.com, pypi.org
  38. Agents

    Graphon raises $35 million to expand memory infrastructure for AI agents

    Graphon Labs announced a $35 million Series A on November 10, led by RRE Ventures with participation from HP, Energy Impact Partners, Hitachi Ventures and Gaia VC. The round brings announced funding to $43.3 million. Company announcement The compa…

    Sources: businesswire.com, thenextweb.com
  39. Models

    ElevenLabs launches Scribe v3 and Live with new speech-recognition controls

    ElevenLabs launched Scribe v3 for recordings and Scribe v3 Live for streaming audio on November 10. Both are available through its API, with Live also offered in ElevenAgents. The streaming model adds speaker labeling and conversation-aware turn d…

    Sources: elevenlabs.io
  40. Infrastructure

    EIA forecasts rising U.S. electricity use in its November outlook

    The U.S. Energy Information Administration's November outlook forecasts electricity consumption of 4,306 billion kilowatt-hours in 2026 and 4,371 billion in 2027, compared with 4,216 billion in 2025. The report was released November 10, using a fo…

    Sources: eia.gov
  41. Agents

    Microsoft 365 disruption exposes an AI-agent dependency bottleneck

    Updated November 15, 2026: Anthropic marked its Microsoft 365 connector incident resolved at 21:49 UTC on November 13. Its final notice dates widespread mail, calendar and search errors in the connector and Claude for Outlook from 16:03 UTC Novemb…

    Sources: status.claude.com, wsj.com
  42. Models

    Edge0 releases large local-agent models, with platform and licensing limits worth reading

    Edge0 has published 145B and 120B mixture-of-experts models derived from Qwen3.8-Flash-Next. The project's release log dates the announcement November 13; its indexed blog carries November 12. Exact release time and the reason for that date differ…

    Sources: github.com, huggingface.co
  43. Models

    Perplexity adds source-span extraction to Decider, with deployment evidence still worth checking

    Perplexity’s November 12 announcement introduces Decider v1.3 and Flash v1.3, based on Qwen backbones of 27B and 4B respectively. The new locate feature returns matching text and character positions, rather than writing a free-form answer. The com…

    Sources: perplexity.ai, huggingface.co
  44. Agents

    Agent memory can carry harmless preferences into dangerous contexts

    Persistent memory is supposed to make an assistant more useful. A November 5 preprint asks what happens when it remembers a preference but loses the limits that made it reasonable. In The Memory Trap, researchers compare six frontier models with a…

    Sources: arxiv.org, github.com
  45. Agents

    Feedzai takes agent detection to the bank login, but detection is not consent

    Feedzai announced a Digital Trust expansion on November 12 intended to help banks recognize AI agents entering customer accounts. The company describes more persistent device recognition, shared-network fraud signals and cryptographic device bindi…

    Sources: prnewswire.com, biometricupdate.com
  46. Guth News

    Apple explains the privacy boundary around ambient AI sensing

    Apple’s November12 security report explains how Secure Exclaves separate sensitive camera and microphone processing from the main operating system. Apple distinguishes ambient sensing, which releases limited metadata, from active capture, which ma…

    Sources: security.apple.com, macrumors.com
  47. Guth News

    CoreWeave's reported $4 billion loan plan puts customer contracts at the center of AI financing

    Bloomberg reported on November 11, 2026 that CoreWeave was discussing an investment-grade loan of at least $4 billion. Semafor's November 13 report, syndicated by Yahoo Finance, adds that the proposed debt would be backed by Meta contracts, with M…

    Sources: bloomberg.com, finance.yahoo.com +1
  48. Models

    WorldParticle reuses one AI architecture across materials, while training separate models

    MIT highlighted WorldParticle on November 9, describing research from MIT, the University of British Columbia and collaborators into a shared architecture for simulating materials. The project covers phenomena ranging from moving cloth and sand to…

    Sources: csail.mit.edu, arxiv.org
  49. Models

    Tether introduces BotPsy, a small open-weight model for local tool use

    Tether AI Research introduced BotPsy on November 11, a 0.8-billion-parameter model based on Qwen3.5-0.8B and specialized for tool calling. Its team reports stronger results than its base model across agent benchmarks, including a BFCL v4 score of …

    Sources: huggingface.co
  50. Tools

    Sumary raises $25 million to expand AI accounting software across Europe

    Sumary announced a $25 million seed round led by Singular on November 10, with existing backers byFounders, Partech and Tenity participating. The accounting-software company plans to expand its commercial, implementation and research teams, with B…

    Sources: sumary.com, techfundingnews.com
  51. Business

    Sparxity closes Series A, taking cumulative funding to $90 million

    Sparxity announced a Series A financing on November 10 that takes cumulative funding secured by its team to $90 million. The release does not disclose the size of this individual round, a distinction also noted by The Next Web. Company release, in…

    Sources: kommunikasjon.ntb.no, thenextweb.com
  52. Research

    Sage-Taiyi research puts revisable scientific assumptions at the center of AI discovery

    A new preprint introduces Sage-Taiyi, a research-agent architecture that revises its working assumptions as evidence accumulates and uses those revisions to choose what to investigate next. The paper was submitted on November 9 at 17:38 UTC. arXiv…

    Sources: arxiv.org, github.com
  53. Robotics

    Robbyant releases LingBot-Touch to help robot hands share tactile understanding

    Robbyant, Ant Group’s embodied-AI company, announced LingBot-Touch on November 10, offering a tactile perception layer intended to work across different robot hands and control models. The release targets an integration problem: touch sensors base…

    Sources: via.tt.se, arxiv.org +2
  54. Business

    Ring returns to mobile security with a drone and business patrol robot

    Ring announced two AI-guided security devices on November 10: an indoor Drone Cam and a business-focused Rover Cam. Its AI Pilot combines cameras, sensors, vision models and cloud processing to map spaces and navigate. These are announced products…

    Sources: aboutamazon.com, geekwire.com
  55. Guth News

    Chinese AI firms reportedly seek overseas GPU capacity as cloud-access controls loom

    Chinese AI companies are trying to secure Nvidia-powered capacity outside China amid concern that Washington could tighten remote-access controls, Reuters reported on November 12, citing three unnamed people. The report describes discussions and r…

    Sources: reuters.com, clerk.house.gov +2
  56. Tools

    Rackspace plans roughly 2 MW of CPU capacity for Blacksmith’s build-and-test workloads

    Rackspace and Blacksmith announced a multiyear agreement on November 10 for dedicated AMD EPYC compute and storage representing approximately 2 MW of capacity in Northern Virginia. Deployment is expected to begin in early 2027. The contract target…

    Sources: ir.rackspace.com, blacksmith.sh
  57. Agents

    CapLite puts small-agent write permissions in code, with an explicit utility cost

    A November 12 preprint proposes a defense for small open-weight agents that does not ask their language model to author the security rules. CapLite derives policy from tool schemas, fixes write operations before reading tool results and checks con…

    Sources: arxiv.org
  58. Agents

    PRISM tests which tools an agent actually needs before granting access

    A November 8 preprint introduces PRISM, a method for reducing an agent's tool privileges before execution. It tests tool removals in resettable sandboxes, learns which capabilities a task needs and groups tasks into reusable roles. This is newly r…

    Sources: arxiv.org
  59. Agents

    LangChain’s reported $300 million round backs the agent-engineering layer

    LangChain raised $300 million at a $3 billion valuation in a round led by Alphabet’s CapitalG, according to The Wall Street Journal. The report names IVP, Sapphire Ventures, GIC and DFJ Growth as other participants. LangChain subsequently confirme…

    Sources: wsj.com, siliconangle.com +1
  60. Agents

    Mirendil raises $1 billion for agent-led AI research, with product proof still the test

    Mirendil announced a $1 billion Series A on November 12, led by Kleiner Perkins and Clearlake Capital. Its announcement lists Google among participating investors and describes Lor, its agent-based research platform, as a private beta. Company ann…

    Sources: mirendil.com, prnewswire.com +1
  61. Models

    Qwen4-Max enters a time-limited preview through Alibaba’s Token Plan

    Alibaba has opened Qwen4-Max-Preview, moving its next model generation beyond the release-watch stage. Official QwenCloud documentation puts the start at 10 a.m. Beijing time on November 11, or 02:00 UTC, and says the preview will end at the end o…

    Sources: docs.qwencloud.com
  62. Agents

    Pydantic AI 2.67 adds durable dynamic agents and records guardrail decisions for replay

    Pydantic AI released version 2.67.0 on November 10, extending its agent framework’s support for recoverable workflows while changing several compatibility requirements. Among the changes are durable execution for dynamically created subagents, per…

    Sources: github.com
  63. Business

    Oura agrees to buy Counsel Health in $625 million cash deal

    Oura has agreed to acquire Counsel Health, extending its wearable-health business into physician-supervised care. The November 10 announcement describes a definitive agreement rather than a completed acquisition. Oura Latham & Watkins, which repre…

    Sources: ouraring.com, lw.com
  64. Research

    OpenAI–Upwork study finds larger AI project budgets, not a comparable hourly-rate premium

    OpenAI and Upwork published research on November 10 suggesting that AI is changing how freelance projects are assembled. The evidence points to different combinations of skills and larger assignments, rather than a straightforward story of either …

    Sources: openai.com, cdn.openai.com