A Guth Labs publication ↗
Guth News

Models

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
  1. 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 The earlier unpublished preview described an upcoming discussion; this article incorporates the subsequent results.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
  2. 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
  3. 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
  4. 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
  5. 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
  6. 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
  7. 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
  8. 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
  9. 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
  10. 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
  11. 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
  12. 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
  13. 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
  14. 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
  15. 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
  16. 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
  17. 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
  18. 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
  19. 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
  20. 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
  21. Models

    Corelight finds hidden dependency changes beneath pinned MCP servers

    Corelight has documented a deployment risk in AI assistants connected to administrative tools: a server can keep the same release identifier while its supporting code changes. The company’s November 10, 2026 study concerns installation choices, no…

    Sources: corelight.com, docs.astral.sh
  22. Models

    OpenAI study finds training gains can make reasoning traces harder to monitor

    OpenAI’s November 10 research report examines how model training affects the usefulness of written reasoning traces for safety oversight. The company says observations from GPT-6 Astra prompted the study. OpenAI research The researchers report tha…

    Sources: alignment.openai.com
  23. Agents

    OpenAI opens Business Hub and customer-facing agents inside ChatGPT

    OpenAI introduced Businesses in ChatGPT in beta on November 10, giving companies a way to claim profiles, update their information and inspect aggregate engagement data. Eligible businesses can use Business Hub wherever ChatGPT is available; Busin…

    Sources: openai.com
  24. Models

    Ollama adds Ripple support and lower Apple-silicon memory use; next release remains a preview

    Ollama’s stable 0.40.16 release adds Ripple Preview, Ricursive Intelligence’s 65-billion-parameter coding model, alongside fixes aimed at local inference and agent integrations. The GitHub release is dated November 9. Stable release The release no…

    Sources: github.com
  25. Models

    New York expands independent AI testing and opens frontier-developer registration

    New York announced a statewide AI-testing network on November 10 and opened registration for large frontier developers ahead of January compliance requirements. The initiative joins technical scrutiny with a concrete reporting process. The New Yor…

    Sources: suny.edu
  26. Agents

    n8n releases Nod, a local personal agent with open-source code

    n8n has introduced Nod, a standalone personal AI agent for Mac and Windows, as a research preview. Announced November 10, the app comes from the new n8n Labs team and is separate from the company’s workflow-automation product. Users connect their …

    Sources: blog.n8n.io, github.com
  27. Models

    MiniMax releases H3-Image weights, with substantial geographic and commercial limits

    MiniMax has released H3-Image, an image-generation and editing model with downloadable weights. The company’s Chinese announcement and model license are dated November 10; its English announcement is dated November 11. MiniMax says the system can …

    Sources: minimax.cn, minimax.io +1
  28. Models

    LongCat-2.5’s public weights add a self-hosting route beyond its hosted API

    Availability rechecked November 15, 2026. The exact first-public opening date remains unconfirmed. Meituan’s LongCat-2.5 is available as downloadable weights on Hugging Face under the MIT license, adding a self-hosting route to its hosted API. The…

    Sources: huggingface.co
  29. Guth News

    llama.cpp adds FP8 cache support and fixes Muse Spark structured replies in November 11 prereleases

    The llama.cpp project has published prerelease builds that add scaled FP8 key-value cache support and repair a structured-output problem affecting Muse Spark. GitHub lists build b12274 at 02:52 on November 11 and b12282 at 03:31. Both are explicit…

    Sources: github.com
  30. Tools

    GRAIL opens a €1 million call for robotics data and evaluation tools

    Europe’s GRAIL project opened its Genesis call on November 9 to support the data and evaluation infrastructure behind physical AI. The call offers a €1 million budget and up to €100,000 for each selected project, with applications closing February…

    Sources: grail-robotics.ai, exteriors.gencat.cat
  31. Models

    Cerebras grows cloud inference while contracted capacity remains a future delivery promise

    Cerebras reported third-quarter GAAP revenue of $210.8 million and non-GAAP core revenue of $230.9 million on November 12. The distinction matters: the two measures are not interchangeable. The company also reported 700 gross megawatts of data-cen…

    Sources: investors.cerebras.ai, reuters.com +1
  32. Agents

    Google's agent governance reaches outside its cloud, with a security tradeoff to understand

    Google has opened a limited preview of Agent Gateway for cross-cloud, extending centralized agent governance to CNCF-conformant Kubernetes clusters outside Google Cloud. Its November 11 release notes name Amazon EKS, Azure AKS, OpenShift and bare-…

    Sources: docs.cloud.google.com
  33. Models

    Baidu’s Token Exchange makes model verification part of the marketplace

    Baidu formally introduced Token Exchange at its November 12 World conference, positioning the service as a marketplace where customers can compare model providers rather than simply buy another model. Its company announcement describes fingerprint…

    Sources: nasdaq.com, yicaiglobal.com
  34. Research

    InfraRSI research targets faster AI training through coordinated infrastructure changes

    A research team has introduced InfraRSI, an agent-based system that searches for ways to accelerate distributed language-model training by changing launch settings and code together. The preprint was submitted to arXiv on November 8. Paper. Rather…

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

    IFM releases PAN 2 world model with training code and open weights

    The Institute of Foundation Models at Mohamed bin Zayed University of Artificial Intelligence announced PAN 2 on November 10, releasing a world-model system intended to simulate the consequences of actions before an agent takes them. IFM says its …

    Sources: ifm.ai, github.com
  36. Guth News

    Cloudflare adds per-Workflow concurrency caps, with limits developers need to account for

    Cloudflare has added a way to limit how many instances of an individual Workflow run simultaneously. The November 10 change gives developers a more targeted control over workloads that share an account, including applications calling databases or …

    Sources: developers.cloudflare.com
  37. Tools

    Cloudflare distributes $301,300 to Vite ecosystem projects and maintainers

    Cloudflare said on November 9 that it is distributing $301,300 to 44 projects and maintainers in the first allocation from its Vite Ecosystem Fund. The money comes from an earlier $1 million commitment, with further quarterly distributions planned…

    Sources: blog.cloudflare.com
  38. Agents

    Cloudflare adds analytics and tracing for Markdown delivered to AI agents

    Cloudflare announced new analytics, logs and tracing for Markdown for Agents on November 9, giving website operators more ways to investigate how their pages are converted for AI systems. Its announcement describes the service as generally availab…

    Sources: developers.cloudflare.com
  39. Agents

    Natera’s lab-support agents show why fast answers still need measured accuracy

    Natera and AWS published a November 12 case study of SWARM, an equipment-support agent system built on Bedrock AgentCore. It reports reducing first response from 17.4 hours to approximately eight minutes, with automated diagnostic reports covering…

    Sources: aws.amazon.com
  40. Models

    MMDM exposes the gap between choosing one answer and choosing every valid action

    A new benchmark asks a practical question about single-pass decision models: can they identify all valid choices when the options depend on one another? MMDM, submitted to arXiv on November 10, evaluates 28 models across 5,448 questions in 16 cate…

    Sources: arxiv.org
  41. Models

    Gemini 4 Hydrogen moves beyond rumors into Google’s official API lineup

    Google now lists Gemini 4 Hydrogen as generally available in its November 12 Gemini API release notes. The model identifier is gemini-4-hydrogen, with access on the free tier and standard paid pricing of $1.50 per million input tokens and $7.50 pe…

    Sources: aistudio.google.com
  42. Research

    OpenAI pledges $2 million for child-safety research, with evidence still to come

    guthlabs.newsroom · November 12, 2026 OpenAI announced on November 12 that it has committed $2 million to research on AI and children, following its signing of a child-safety charter earlier this week at the Paris Peace Forum. The pledge supports …

    Sources: openai.com
  43. Models

    Cloudflare brings multimodal EmbeddingGemma 2 to Workers AI

    November 12, 2026 | guthlabs.newsroom Cloudflare has added Google's EmbeddingGemma 2 to Workers AI, giving developers a hosted embedding option for mixed collections of text, code, images, video and audio. The announcement is a new Cloudflare avai…

    Sources: developers.cloudflare.com
  44. Models

    MotionGraphicsBench tests whether AI animations actually follow the brief

    November 7, 2026 | guthlabs.newsroom Researchers affiliated with Figma and Stanford have introduced MotionGraphicsBench, a benchmark of 100 tasks that ask language models to animate supplied static SVG artwork. Providing the design upfront separat…

    Sources: arxiv.org, huggingface.co
  45. Agents

    Cloudflare expands agent-hosting examples, with recovery limits developers need to check

    Cloudflare has published new routes for running coding agents on its infrastructure, covering Codex, Exo and Nanocode. The November 9 announcements focus on keeping conversations and work recoverable when a connection or process is interrupted, ra…

    Sources: developers.cloudflare.com
  46. Models

    Claude 5.2 migrations need a conversation-state check, not just a new model name

    November 9, 2026 | guthlabs.newsroom Anthropic's November 9 platform release notes describe integration changes that deserve attention alongside Claude Fable 5.2 and Mythos 5.2's launch. An agent fleet cannot assume that replacing a model identifi…

    Sources: platform.claude.com
  47. Robotics

    Google rolls out Gemini Robotics 2.5 to partners, with factory reliability still under test

    Google DeepMind introduced Gemini Robotics 2.5 on November 10 and began rolling it out to robotics partners. The vision-language-action model turns visual input and instructions into physical movements. Google reports a 96% success rate for automo…

    Sources: deepmind.google, arstechnica.com +1
  48. Guth News

    Google begins rolling out Gemini Live on Apple CarPlay

    Google began rolling out its dedicated Gemini app on Apple CarPlay on November 11, extending its voice assistant to another setting where users can talk through information from their phones. The company says the release is global and covers suppo…

    Sources: blog.google, 9to5google.com +1
  49. Agents

    Google previews Antigravity Swarm for coordinated coding-agent teams

    Google began rolling out a preview of Antigravity’s /swarm command on November 10, adding a shared coordination system for multiple agents working on a task. The announcement accompanied Gemini 4 Carbon’s arrival in Antigravity, but the two featur…

    Sources: antigravity.google
  50. Models

    OpenAI sets December 11 retirement for Advanced and Standard Voice, with video exception

    OpenAI plans to remove Advanced and Standard Voice as selectable ChatGPT voice modes on December 11, 2026, moving remaining users to GPT-Live. The company announced the change in its November 10 release notes. Live video and screen sharing are an …

    Sources: help.openai.com
  51. Business

    UK confirms backing for Fractile as FT reports $7 billion valuation

    The UK government has confirmed that Sovereign AI and the British Business Bank invested in Fractile’s Series C, turning earlier reporting about state participation into an on-record announcement. The November 11 release describes a funding round …

    Sources: gov.uk, ft.com +3
  52. Models

    Bilt selects Cerebras for its neighborhood AI platform and concierge

    Bilt has selected Cerebras to provide AI infrastructure for Neighborhood OS and Bilt Concierge, bringing a specialist inference provider into its consumer-commerce platform. Announced November 10, the agreement targets services such as rewards red…

    Sources: investors.cerebras.ai, newsroom.biltrewards.com
  53. Agents

    Flashpoint announces threat-intelligence agents for early 2027

    Flashpoint announced a set of AI agents on November 10 aimed at recurring threat-intelligence work, from investigations and briefings to alert configuration and triage. The company’s FAQ says the new offering will be available in early 2027. Flash…

    Sources: flashpoint.io, siliconangle.com
  54. Models

    Bessent outlines AI testing and incident-alert plans with China

    US Treasury Secretary Scott Bessent has set out the next priorities for AI cooperation with China ahead of talks in Shenzhen, while warning that the countries will continue to disagree on important issues. Reuters reported his November 10 remarks.…

    Sources: streetinsider.com, content.govdelivery.com
  55. Guth News

    SMIC warns AI chip supply could remain tight through 2027

    Reuters reports that SMIC said on November13 that AI-related semiconductor supply would remain tight through2027, with demand still strong. This is management’s outlook, not an independently established future shortage. The Reuters report was veri…

    Sources: www1.hkexnews.hk, reuters.com
  56. Agents

    Mount Thor raises $45 million to turn Apple hardware into AI-agent infrastructure

    Mount Thor announced US$45 million in combined Seed and SeriesA equity financing on November12 to expand managed Apple-hardware infrastructure. CRV and Conviction co-led the SeriesA. The company also disclosed additional non-dilutive hardware fina…

    Sources: blog.mountthor.com, 9to5mac.com
  57. Business

    TSMC approves up to $1 billion for frontier labs, not a completed billion-dollar purchase

    TSMC's board approved investments of up to $500 million each in Anthropic and OpenAI on November10. The two authorizations are different: Anthropic common shares at its per-share IPO price, and OpenAI preferred stock. The official board resolution…

    Sources: pr.tsmc.com, reuters.com
  58. Models

    TNG tests plug-in coding expertise without retraining a model's existing weights

    November 10, 2026 | guthlabs.newsroom TNG Technology Consulting's November 10 research account explores a way to specialize an open mixture-of-experts model while leaving its existing weights frozen. Its experiment adds independently trained Rust …

    Sources: huggingface.co
  59. Models

    Pydantic AI 2.68 extends durable delegation, but the new child-workflow path is opt-in

    Pydantic AI's 2.68.0 release adds conversation persistence, streamed cost information for Google and Bedrock models, and a Gemini Decisions API backend. Its GitHub release title is dated November 11, while the publication display reads November 12…

    Sources: github.com
  60. Agents

    ElevenLabs moves from better transcripts to agents that remember and follow through

    ElevenLabs is extending its speech business into the work around a conversation. At its November 11 New York summit, the company opened free early access to ElevenAI, a workplace app for meeting notes and follow-up tasks, while previewing persiste…

    Sources: elevenlabs.io, thenextweb.com
  61. Agents

    Nebuly raises $21 million to put AI agents on a business-value scorecard

    Nebuly announced a $21 million Series A led by Planven on November 12 to expand its enterprise AI-measurement business. The company plans to grow its commercial team in Europe and the United States and continue developing its platform. Company ann…

    Sources: nebuly.com, thenextweb.com
  62. Models

    Baidu lists ERNIE 5.5 Flash Preview for API access with a million-token context

    Baidu’s Qianfan documentation now lists ERNIE-5.5-Flash-Preview as an API model, with a one-million-token context window. The catalog, now dated November 11, confirms the service specification. Baidu model catalog The catalog includes the preview …

    Sources: cloud.baidu.com, openrouter.ai +1
  63. Agents

    Nex-N2.6 targets less repetitive agent reasoning with open FP8 weights

    Nex AGI's Nex-N2.6 is available as a public FP8-weight repository and an OpenRouter model listing. The model card describes a roughly 321-billion-parameter multimodal agent model based on GLM-5.3-Flash, with an Apache-2.0 license. Its intended wor…

    Sources: huggingface.co, openrouter.ai
  64. Guth News

    Orakl raises $34 million for AI cancer-drug prediction, not a replacement for clinical trials

    Orakl Oncology has announced a $34million Series A co-led by Norrsken VC and Singular. The November12 financing backs an AI-and-biology approach to deciding which cancer drugs merit further development. The amount and round are corroborated by the…

    Sources: norrsken.vc, infraviacapital.com +1
  65. Models

    EUROPA shares a trillion-scale training recipe, not a finished frontier model

    Event/source date: 2026-11-12 technical reports; experiment execution dates unknown The EUROPA consortium has published an engineering account of training a near-trillion-parameter mixture-of-experts configuration on JUPITER, Europe’s shared publi…

    Sources: pytorch.org, domyn.com
  66. Models

    Torus raises $500 million and launches modular AI compute for existing power sites

    Torus announced $500 million in growth capital on November 11 alongside Foundry, a modular system intended to place AI inference capacity at commercial and industrial sites with available electricity. Magnetar led the financing. The company says t…

    Sources: torus.co, thenextweb.com
  67. Guth News

    Tendrils confirms $30 million to develop a CPU built around parallel computing

    Tendrils confirmed a $30 million round led by Benchmark on November 10 to develop a new general-purpose processor. The company names the UK Sovereign AI Fund and Visionaries Club among its backers and says Jack Altman has joined its board. Company…

    Sources: tendrils.co, newcomer.co +1
  68. Models

    DeepL launches Aure to carry changing tone into translated speech

    DeepL announced Aure on November 11, a voice-translation model that the company says follows changes in a speaker’s delivery throughout a conversation. The launch targets a familiar limitation of automated interpretation: translated words can be a…

    Sources: deepl.com, the-decoder.com
  69. Models

    Vercel’s gateway data separates cheap decisions from expensive generation

    Vercel’s November 9 AI Gateway Production Index reports that decision models accounted for 26% of its October requests but just 0.53% of estimated spending. The gap suggests that request counts alone can badly misrepresent the economics of a mixed…

    Sources: vercel.com
  70. Research

    Databricks plans $3 billion expansion across Europe and the Middle East

    Databricks announced on November 11 that it plans to invest approximately $3 billion across Europe and the Middle East over three years. The company describes the commitment as its largest regional investment outside North America, with spending i…

    Sources: databricks.com
  71. Models

    Salesforce opens SOV-Transcribe-1, but active parameters are not the whole deployment bill

    Event/source dates: 2026-11-12 release announcement; exact time unknown Salesforce released SOV-Transcribe-1 on November12 as an Apache 2.0 open-weight model for non-streaming speech transcription. The release lets developers steer transcript form…

    Sources: huggingface.co, artificialanalysis.ai
  72. Agents

    Rox raises $100 million to expand enterprise revenue agents

    Rox announced a $100 million Series B financing at a $1.3 billion valuation on November 11. General Catalyst and March Capital co-led the round, according to the company’s announcement. The proceeds will support AI research, custom models, evaluat…

    Sources: prnewswire.com, rox.com
  73. Robotics

    Path Robotics raises $300 million as it looks beyond automated welding

    Path Robotics announced $300 million in Series E financing at a $2.45 billion post-money valuation on November 10. Georgian led the round. The Ohio company's next objective is to expand the physical-AI technology it developed for welding into more…

    Sources: path-robotics.com, siliconangle.com +1
  74. Policy

    AI’s power race faces a gas lock-in risk, but announced plants are not built capacity

    BloombergNEF’s November 11 report identifies a climate risk in the rush to energize data centers: temporary on-site gas generation can become a durable part of the system. It tracks 181GW of announced gas generation proposed alongside facilities, …

    Sources: about.bnef.com
  75. Agents

    PuppyGraph raises $10 million to give enterprise agents a structured route through company data

    PuppyGraph announced a $10 million Series A on November 10, bringing its total funding to $15 million as it expands a graph-query engine for enterprise AI agents. Particle Future and defy.vc participated. The company's announcement positions the s…

    Sources: puppygraph.com, siliconangle.com
  76. Models

    Anthropic’s November 12 policy puts hardware safety outside the model

    Anthropic’s revised Usage Policy carries a November 12 effective date, bringing a practical checkpoint for developers connecting Claude to equipment. This is the effective date of a policy announced October 8, not a new model release or a new stat…

    Sources: anthropic.com
  77. Agents

    Anthropic releases benchmark for agents that must share resources

    Anthropic announced the release of MAMUBench on November 11, giving developers a way to test what happens when different users’ AI assistants compete for the same budget, calendar or group booking. The accompanying research finds that assigning ev…

    Sources: alignment.anthropic.com, arxiv.org +1
  78. Models

    Octave’s Cerebre deal targets the missing context behind industrial AI

    Octave announced a deal for Cerebre on November 11, adding software that connects engineering drawings, plant data and operational knowledge into a plant model. The acquisition announcement puts industrial information management, rather than a new…

    Sources: investors.octave.com, dealroom.co
  79. Models

    DeepMind releases AlphaGenome 2 to predict genetic effects in specific cell types

    Google DeepMind announced AlphaGenome 2 on November 11, extending its DNA prediction work from broad tissue patterns to particular cell types and biological states. The model is available through the AlphaGenome API for noncommercial research, wit…

    Sources: deepmind.google, alphagenomedocs.com
  80. Models

    Cloudflare’s AI attack report puts credential boundaries ahead of model hype

    Cloudflare says it observed more than 50 AI penetration-testing framework families scanning customers over three days. Its November 13 report also describes attackers using compromised, long-lived CI/CD credentials to move between cloud environmen…

    Sources: blog.cloudflare.com
  81. Models

    Tandem broadens local DeepSeek serving, but a 24GB GPU is only part of the machine

    Tandem-vLLM 0.2.0 expands a community DeepSeek-V4.1-Flash serving engine beyond Blackwell GPUs. The public project now lists Ampere, Ada and Hopper support and a second-GPU helper mode. Its design holds compressed experts in host RAM and caches fr…

    Sources: github.com, reddit.com
  82. Models

    Cohere adds cybersecurity and code review to its sovereign-AI push

    Cohere said on November 13 that it completed its acquisition of AISLE and welcomed the cubic code-review team. The additions place software security beside model development in its enterprise strategy. Cohere announcement AISLE supplies vulnerabil…

    Sources: cohere.com, betakit.com
  83. Tools

    DeepSeek opens its app toolkit to developers, with hosted tools rather than a new weight release

    DeepSeek has expanded its developer offering from model responses toward the surrounding machinery of a chat application. Its November 13 changelog opens app-style search, document input, link reading, safety checks and standalone speech and OCR s…

    Sources: api-docs.deepseek.com
  84. Agents

    SteerDistill teaches low-bit agents to correct their own failures, but checkpoints are still pending

    Low-bit AI agents need more than smaller weights. They also need to preserve the details that make a tool action correct. SteerDistill, a preprint submitted November 10, targets that problem by training quantized agents on corrections made in thei…

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

    Infrathon shows why a short agent benchmark can pick the wrong long-running worker

    A November 9 preprint argues that the time allowed for an agent to work can change which model looks strongest. Infrathon studies infrastructure optimization across 25 tasks, six models and 150 runs. The authors report that 135 runs improved their…

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

    Two local-model studies show why disagreement is not the same as useful uncertainty

    Two recent preprints test an appealing idea for local AI: use inexpensive precision changes to decide when an answer deserves more computation. The results are more qualified than a simple “second opinion” pitch. In One Model, Two Precisions, publ…

    Sources: arxiv.org
  87. Models

    ci-proto brings local forecasting a long-history alternative to attention

    Not every useful local model needs to chat. Continuous Intelligence's ci-proto targets numerical forecasts, with a design that substitutes recurrent processing for temporal self-attention. The November 12 research preprint describes a Mamba-3-base…

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

    Autoware releases AutoE2E driving model, with noncommercial checkpoints and road-testing limits

    TIER IV and the Autoware Foundation have expanded their public autonomous-driving tools with AutoE2E, a 79.9-million-parameter model released alongside a broader set of driving-model options on November 11. The release gives developers code, train…

    Sources: tier4.co.jp, autoware.org +2
  89. Infrastructure

    AMD’s EPYC 9006 SP7 servers begin shipping as independent tests show a substantial gain

    AMD announced general availability of its sixth-generation EPYC SP7 server processors on November 10, moving its Zen 6-based platform from earlier previews into shipping systems. The company says server partners are delivering machines, while Micr…

    Sources: newsroom.amd.com, phoronix.com
  90. Models

    SpeakLeash lists DFlash draft models for Bielik 11B v3.1

    SpeakLeash's catalog lists DFlash and DFlash2 companion models for Bielik 11B v3.1. Its November 10 announcement scheduled the draft-model release for November 11. The exact first-public-availability time remains unverified; relative repository up…

    Sources: huggingface.co, bielik.ai
  91. Models

    Ragas 0.4.4 treats the evaluator itself as an attack surface

    Ragas released version 0.4.4 on November 14, adding a security fix to the software many teams use to evaluate retrieval-augmented AI. PyPI's release record confirms the package and date. The maintainers' release notes say the patch addresses CVE-2…

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

    Plurai raises $13 million and opens agent simulation to developers

    Plurai announced $13 million in seed funding and general availability of its AI-agent simulation platform on November 11. Team8 led the round. The company says developers can now use a free tier after access was previously limited to selected cust…

    Sources: prnewswire.com, plurai.ai
  93. Models

    Adobe and Google plan closer Acrobat and Gemini workflows, with launch details still pending

    Adobe and Google announced an expanded partnership on November 10 that would put more Adobe tools within reach of users of Gmail, Drive and Gemini, while bringing Google models into additional Adobe workflows. The companies presented the plans at …

    Sources: news.adobe.com, blog.google +1
  94. Models

    Salesforce details an 80-hour embedding-model research loop, not a new autonomous free-for-all

    November 11, 2026 | guthlabs.newsroom Salesforce’s November 11 newsroom account describes agents improving an embedding model in roughly 80 hours. The company says its SFR-AutoR&D system produced SFR-Embedding v4 from EmbeddingGemma 2, reporting a…

    Sources: salesforce.com, raihanjoty.github.io +1
  95. Models

    Browser Use cuts reported model cost by preserving the prompt cache

    Browser Use 1.0.4 changes screenshot cleanup so the browser agent does not continually invalidate its provider’s prompt cache. The maintainers report model cost per task of $0.208 versus $0.434 in a 200-task benchmark, with essentially unchanged s…

    Sources: github.com