A Guth Labs publication ↗
Guth News

Tools

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
  1. 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
  2. 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
  3. Agents

    Cloudflare expands Radar bot profiles with identity and content-use fields

    Cloudflare’s public Radar Bots Directory now draws on BotBase and presents more detail about how tracked bots and agents identify themselves and use web content. The November 9 update adds an identity section to bot profiles, with access level, co…

    Sources: developers.cloudflare.com
  4. 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
  5. 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
  6. Agents

    Cloudflare opens Kitesurf: an agent browser engine with deployment caveats

    Cloudflare has released Kitesurf, its experimental browser engine for AI agents, as open source under Apache 2.0. The November 12 announcement makes the code available for inspection and modification rather than establishing production readiness. …

    Sources: developers.cloudflare.com, github.com
  7. 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
  8. Agents

    World details Cloudflare’s agent-trust push, but local agents remain outside its initial scope

    World’s November 12 explanation of Trusted Agent Delegation gives agent builders a practical boundary to watch: the initiative initially focuses on operator-hosted agents, not on-device or locally hosted ones. Its goal is to help websites assess t…

    Sources: world.org, blog.cloudflare.com +1
  9. Robotics

    T-Rex separates robot planning from fast touch corrections

    guthlabs.newsroom · November 12, 2026 Panasonic's November 5 research announcement highlights a robot-control approach that gives touch its own faster feedback loop. T-Rex combines vision-and-language motion planning with tactile corrections, aimi…

    Sources: news.panasonic.com, arxiv.org +1
  10. 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
  11. 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
  12. 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
  13. 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
  14. 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
  15. 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
  16. Agents

    Bay raises $37 million Series A to expand controls on AI agent actions

    Bay emerged from stealth on November 10 with $50 million in total financing, comprising a $37 million Series A led by Battery Ventures and a previously undisclosed $13 million seed round. YL Ventures and Striker Venture Partners led the seed; the …

    Sources: businesswire.com, thenextweb.com
  17. 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
  18. Tools

    Boston Dynamics uses wheeled Atlas to test learned control at motor limits

    Boston Dynamics has described an Atlas experiment that replaces the humanoid’s feet with scooter wheels to study how learned whole-body control behaves near the limits of its motors. The company identifies the machine as a research platform, with …

    Sources: bostondynamics.com
  19. Policy

    England proposes safety certification for school AI, with supplier rules still open

    England's Department for Education opened a consultation on November 10 on a safety-certification scheme for education technology and AI products. It is a proposal for education settings in England, not an already-operating certification program o…

    Sources: gov.uk, consult.education.gov.uk
  20. Agents

    Ancep raises $8.5 million for AI-assisted commercial insurance placement

    Ancep announced $8.5 million in funding on November 10 to expand AI tools for commercial insurance brokers. Creandum led the round, with Kleiner Perkins and Vicus Ventures participating. The London company’s legal adviser, Wilson Sonsini, confirme…

    Sources: wsgr.com, ancep.com +1
  21. 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
  22. Guth News

    Claude Meetings makes the permission boundary part of the product

    Event/source dates: Earlier independent reports 2026-11-10; official rollout article dated 2026-11-12; exact first availability unresolved Claude Meetings puts an AI participant inside Google Meet, where it can take notes and help produce work dur…

    Sources: support.claude.com, claude.com +1
  23. Infrastructure

    CoreWeave adds managed PostgreSQL beside AI workloads

    CoreWeave announced general availability of CoreWeave Database on November 11, adding a managed PostgreSQL service inside customers’ CoreWeave Kubernetes Service clusters. The launch gives teams using its GPU infrastructure another option for keep…

    Sources: investors.coreweave.com
  24. Robotics

    Serve opens a delivery-robot API, with live access still subject to approval

    Serve Robotics announced an AV Delivery Developer Portal on November 11, giving outside developers, businesses and AI assistants a software interface to its sidewalk delivery fleet. A sandbox lets builders rehearse deliveries without dispatching a…

    Sources: investors.serverobotics.com, developer.serverobotics.com
  25. 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
  26. 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
  27. 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
  28. 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
  29. 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
  30. 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
  31. 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
  32. 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
  33. 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
  34. 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
  35. 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
  36. 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
  37. 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
  38. Infrastructure

    Armada starts European production of modular AI data centers

    Armada and Johnson Controls announced on November 11 that their Galleon Forge Two operation in Dublin has entered continuous production. The companies said the first Irish-built Leviathan modular data center has been completed for Fossefall in Nor…

    Sources: armada.ai
  39. 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
  40. 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
  41. 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
  42. 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
  43. Robotics

    GXO’s Apollo expansion is a performance-gated plan, not 1,000 deployed robots

    GXO and Apptronik have outlined a path from warehouse trials to a larger humanoid fleet. Their November 12 letter of intent expands testing of Apollo and develops a Robotics-as-a-Service commercial framework. It is not evidence that a thousand rob…

    Sources: gxo.com, supplychaindigital.com
  44. Models

    A short traffic spike can leave agent inference slow long after the spike ends

    A new preprint argues that agent-serving systems can remain slow after a temporary traffic surge has passed. In controlled tests, Wen Huang and Jack Liu report that prefix-cache misses created a feedback loop: slower requests kept more sessions al…

    Sources: arxiv.org
  45. Tools

    NXP packages the integration work behind physical AI, starting with drones

    Event: November 9, 2026, 03:00 EST / 08:00 UTC announcement. Catch-up coverage, not breaking November 13 news. NXP's new robotics platform combines hardware reference designs with an open-source software development kit intended to reduce the inte…

    Sources: investors.nxp.com, allaboutcircuits.com
  46. Tools

    Ramp’s AI spending update shows why usage and revenue need separate scorecards

    Ramp’s November11 economic update says observed business AI spending rose 1.65% in October. Its newly introduced department breakdown puts engineering at 72% of classified spending and customer support/success at 13.3%, up from 0.8% in June. These…

    Sources: econlab.substack.com, ramp.com +1
  47. Tools

    Limited EU AI Act product rules take effect in Northern Ireland

    A UK–EU decision bringing product-related parts of the EU AI Act into Northern Ireland took effect on November 11. The Withdrawal Agreement Joint Committee adopted Decision No 3/2026 in Brussels on November 10, with its final text setting entry in…

    Sources: assets.publishing.service.gov.uk
  48. Agents

    AWS separates web search from the agent gateway, not from the need to verify results

    AWS has made AgentCore web search callable directly from application code, without first creating an agent gateway. The general-availability notice is dated November10; the more detailed walkthrough appeared November11. This is newly discovered co…

    Sources: aws.amazon.com
  49. Agents

    United restricts unauthorized AI agents across its booking channels

    United Airlines said on November 11 that it is restricting unauthorized third-party personal AI assistants on its digital platforms and through travel agencies and aggregators. The move extends the contest over agent-driven commerce into airline d…

    Sources: united.mediaroom.com, travelweekly.com +1
  50. Agents

    AWS reduces AgentCore startup and idle costs, without making agent work automatically safe

    AWS announced an updated Amazon Bedrock AgentCore runtime on November 12, aimed at reducing the overhead of running many isolated agent sessions. The company says startup can be up to 15 times faster, with sessions supporting up to eight virtual C…

    Sources: aws.amazon.com
  51. Infrastructure

    NVIDIA puts OpenShell into CNCF and previews more Kubernetes infrastructure

    NVIDIA and the Cloud Native Computing Foundation announced on November 10 that OpenShell, a runtime for controlling autonomous AI agents, is joining CNCF as a Sandbox project. The move places the project within a community governance structure as …

    Sources: cncf.io, blogs.nvidia.com
  52. Agents

    AWS introduces UAEF to compare AI agents across frameworks

    AWS has introduced the Universal Agent Evaluation Framework, an open-source system for testing agents built with different software stacks. Its November 11 announcement describes 36 metrics across seven dimensions and adapters that put execution t…

    Sources: aws.amazon.com, awslabs.github.io
  53. Models

    IBM gives Granite an explicit instruction-data format, without promising injection immunity

    IBM released experimental Granite 4.1 LLMON previews on November12, using structured prompt blocks to distinguish instructions from the data they operate on. The release includes models named 3B and 8B and a toolkit. This is an evaluation release,…

    Sources: huggingface.co
  54. Agents

    Deepnote tests whether proactive agents know when not to help

    Deepnote's Agent Initiative Benchmark tests 12 models across 50 workplace situations. Its useful distinction is that an accurate answer can still be an unwanted interruption. The creator reports that models stayed silent in roughly six out of ten …

    Sources: deepnote.com, github.com
  55. Tools

    Google brings existing subscriber credits into AI Studio

    Google has moved redemption of existing AI Pro and Ultra developer credits into AI Studio. The November 13 change combines developer-profile setup, Cloud billing and balance tracking rather than introducing a new universal free-API allowance. Goog…

    Sources: blog.google
  56. Models

    Google opens a waitlist for Tabletop, an AI-built space for group planning

    Google Labs introduced Tabletop on November 11, an experiment that lets groups generate interactive tools and use them together in real time. Access is rolling out through a waitlist for adults in the United States. Google's announcement Users can…

    Sources: blog.google
  57. Agents

    Google launches Gemini Decision 1 for the small choices inside AI agents

    Google has introduced Gemini Decision 1 and a dedicated Decisions API for bounded tasks such as choosing a tool, routing a request or grading retrieved context. Its November 11 API changelog confirms the release through the Gemini API. According t…

    Sources: blog.google, aistudio.google.com
  58. Models

    Claris previews Harbor, a private AI server that can pool Macs over Thunderbolt

    Apple subsidiary Claris is privately testing Harbor, an app for running open-weight AI models on Apple silicon Macs. Its public preview page describes a single-Mac endpoint and the option to connect multiple machines over Thunderbolt 5, combining …

    Sources: content.claris.com, macrumors.com
  59. Agents

    Claude Code tightens approval and credential boundaries in 2.1.325

    Anthropic's November 13 Claude Code release changes sandboxed Bash to the normal permission flow by default. It also addresses credential selection in background sessions, provider-policy enforcement during token counting, and several approval-rul…

    Sources: github.com
  60. Tools

    Vercel fixes approval replay so validation cannot rewrite the action that was approved

    Vercel’s November 13 AI SDK 7.0.173 release fixes a subtle approval boundary: validating or refining a tool call must not change the arguments a person has already approved. The patch prevents in-place schema validation and input refinement from m…

    Sources: github.com
  61. Models

    MOSS-Live separates a flowing conversation from the work running behind it

    OpenMOSS's November 7 overview describes an assistant that can keep talking while a separate background component reasons or uses tools. MOSS-Live combines simultaneous listening and speaking with live visual input and a shared coordination layer …

    Sources: huggingface.co, github.com +1
  62. Tools

    Gemini adds financial context through Google Wallet, not authority to move money

    Google is adding Google Wallet as a connected app in Gemini, giving US users a way to ask financial questions grounded in their spending. The rollout covers web, Android and iOS. Google describes budgeting, recurring-charge discovery and visualiza…

    Sources: blog.google, plaid.com +1
  63. Agents

    Devin Apps turns the coding agent’s machine into a private app host

    Cognition introduced Devin Apps on November 13, giving its coding agent a way to keep internal web tools running on the same cloud computer that built them. A team describes a dashboard, form or review queue, then shares a stable URL that requires…

    Sources: devin.ai, docs.devin.ai +1
  64. Models

    ForeSpark targets wasted draft tokens, with a noncommercial licensing catch

    Stable ID: forespark-joint-decoding-budget-2026-11-10 Event/source date: 2026-11-10T01:37:38Z preprint v1; repository release timing not established First covered and last updated: 2026-11-13T03:00:45.531952+00:00 (UTC). ForeSpark proposes a pract…

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

    PEAR brings open-weight robot reasoning out of simulation

    A new robotics preprint reports that simulation-only reinforcement learning can teach an open 27-billion-parameter reasoning model to control a physical robot. The PEAR team reports 18 successful trials out of 30 on a real dual-arm YAM, versus non…

    Sources: mit-han-lab.github.io, github.com
  66. Models

    Redwood tests whether simulated safety checks can stand in for real computers

    A research report offers a limited but useful result for safer AI evaluations: monitors reading a model’s internal activations caught attacks at similar rates in simulated and real computer environments. Researchers tested six probes against Qwen3…

    Sources: github.com
  67. Agents

    Kindo raises $30 million for the control layer around enterprise agents

    Kindo announced $30 million in new financing on November 12, co-led by Crosspoint Capital and Forgepoint Capital. That brings its cumulative funding to $57.6 million, not a $57.6 million new round. The distinction matters because some headline tre…

    Sources: kindo.ai, forgepointcap.com +2
  68. Robotics

    Xilect raises RMB 500 million for robot chips, with first silicon still on a 2028 roadmap

    Xilect, the robot-computing venture backed by MetaX and UBTech, announced RMB 500 million in angel financing on November 12. Its important qualifier is the schedule: a partner development platform is planned for 2027, with the first SoC product an…

    Sources: xilect.com, technode.com
  69. Agents

    PhysLoop checks the physics before trusting an AI-generated video

    A convincing video is not necessarily a physically coherent one. PhysLoop, a November 12 research preprint, proposes checking simulated events before refining their appearance, then checking the finished video again. That second check matters: mak…

    Sources: arxiv.org
  70. Policy

    Runway opens Praxis-1 robot-policy weights and an experimental consumer-GPU path

    Runway made Praxis-1 publicly available on November 11, turning its earlier world-action-model announcement into a release robotics teams can inspect and adapt. The company says the weights, inference code and fine-tuning tools are covered by Apac…

    Sources: runway.com, github.com
  71. Models

    FOAM trims SOAP’s training overhead, but its public code is still pending

    NAVER AI Lab researchers propose FOAM, a lighter version of the SOAP optimizer, in a paper submitted November12. They report competitive validation loss with substantially less optimizer-state storage than SOAP. The central pretraining comparison …

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

    Timon opens its coding-agent workstation, with a boundary between local control and model access

    Timon's maker recently announced that the coding-agent workstation is now public under an AGPL/MIT licensing split. The release offers another way to coordinate existing coding assistants instead of asking developers to adopt a new model. The make…

    Sources: reddit.com, github.com
  73. Agents

    Qwen Code 1.0 preview combines Qoder's agent harness with the open-source coding tool

    Announcement: November 11, 2026. Preview release; not general availability. Alibaba's Qwen Code project has published a 1.0 preview that combines agent capabilities from Qoder and Qwen Code. The release packages code understanding, implementation,…

    Sources: github.com, pandaily.com +1
  74. Models

    Sophon brings local full-duplex voice to Macs, with important limits

    The most interesting part of Sophon-Realtime’s latest updates is not another voice benchmark. It is a new way to run an interruptible speech conversation locally, plus an unusually specific account of what that path cannot do yet. Adventists AI’s …

    Sources: github.com
  75. Models

    Self-hosted Claude agents move the service, not necessarily the model

    Anthropic’s November 12 release lets customers operate the Managed Agents service on their own infrastructure, rather than moving only the tool sandboxes. Agents, sessions and their histories can be stored there. Access requires the account team, …

    Sources: platform.claude.com, thenextweb.com
  76. Agents

    Microsoft releases .NET 11 with a broader toolkit for agent applications

    Microsoft released .NET 11 on November 10, combining its annual application-platform update with tools for building and working alongside AI agents. The release includes C# 15 and is a Standard Term Support version, with two years of support. .NET…

    Sources: devblogs.microsoft.com, learn.microsoft.com
  77. Agents

    AWS Emulator turns cloud behavior into a local test target for coding agents

    AWS has released a local emulator that lets developers and coding agents exercise AWS-style APIs before deploying to the cloud. The useful distinction is not simply that agents helped build it. It is that the company describes a concrete reference…

    Sources: aws.amazon.com, aboutamazon.com +1
  78. Agents

    SGLang's 1.0 plan puts predictable behavior ahead of another performance headline

    Event: November 12, 2026 release tracker opened. Status: roadmap, not a verified 1.0 release. First covered in this run: November 13, 2026, 11:53 UTC. SGLang has outlined a stability-focused path to version 1.0, with explicit configuration errors,…

    Sources: github.com
  79. Models

    ExecuTorch adds a browser-GPU path for PyTorch models

    PyTorch’s ExecuTorch project described a new WebGPU backend on November 10, allowing supported models to run on a browser’s GPU while staying within the PyTorch deployment toolchain. Developers export a model to an ExecuTorch .pte file, then execu…

    Sources: pytorch.org, docs.pytorch.org +1
  80. Models

    GPU Task Manager 0.4.1 makes local-AI memory reporting less misleading

    Scott Hanselman’s gputaskman released version 0.4.1 on November 13, adding Windows Intel activity reporting and changes that distinguish shared system memory from dedicated VRAM. The release also identifies loaded Foundry Local models on Snapdrago…

    Sources: github.com
  81. Models

    Huawei’s MACaQ gives local-AI builders another way to compress models

    guthlabs.newsroom Research catch-up, not breaking news. The paper was submitted November 7; the public repository was checked November 12 at 05:53 UTC. The repository’s original publication time was not established. Huawei researchers have made MA…

    Sources: github.com, arxiv.org
  82. Tools

    Nvidia’s FP64 update shows why a faster matrix kernel is not the same as a faster application

    Nvidia’s November 13 technical post describes Ozaki-II and 2M algorithms in cuBLAS with CUDA Toolkit 13.4.2. They map high-precision matrix calculations onto integer Tensor Cores. The vendor reports a 3.9-fold DGEMM improvement on GB200 versus nat…

    Sources: developer.nvidia.com
  83. Models

    Haru opens a local, talk-and-listen voice model, not a finished assistant

    November 13, 2026 | guthlabs.newsroom Higurashi announced Haru-4B on November 13, releasing a voice model designed to listen while it speaks rather than wait for alternating turns. The release adds a downloadable research foundation to the growing…

    Sources: blog.higurashi.ai, huggingface.co +2
  84. Tools

    LiteLLM’s LTS patch targets lost tool history and misleading AI bills

    LiteLLM’s November 12 LTS release, version 1.108.7, addresses a problem that can look like model confusion: streamed MCP follow-ups retaining only the latest tool round, causing repeated calls. The patch also fixes missing or duplicated sequence n…

    Sources: github.com
  85. Models

    Langfuse adds model-scoped gateway keys and tighter session handling

    Langfuse’s 4.80.0 release introduces gateway-key restrictions based on model patterns, alongside Google AI Studio and Microsoft Azure provider support. Existing gateway keys can have their names, metadata and model selections edited. The interface…

    Sources: github.com
  86. Models

    Codex 0.168.0 adds a side conversation and more multimodal inputs

    OpenAI's Codex CLI release 0.168.0 adds an optional second conversation pane beside the main chat, alongside audio-recording input for compatible models and optional animated-GIF generation or editing. The release notes identify separate feature f…

    Sources: github.com
  87. Agents

    AgentRun makes repeated agent work a platform problem

    AgentRun is taking aim at a less glamorous part of AI automation: paying an agent to rediscover the same procedure every time. Its November 10 introduction describes an open-source platform that adds durable execution, run tracking and reusable le…

    Sources: agentrun.ai, github.com
  88. Guth News

    llama.cpp reaches 1.0, making compatibility a first-class promise for local AI

    The most important change in llama.cpp 1.0 is not another benchmark. It is a clearer contract for the developers building local-AI applications on top of it. The maintainers released version 1.0.0 of both llama.cpp and its ggml tensor library on N…

    Sources: huggingface.co, github.com
  89. Agents

    EnvOpt asks whether the next agent upgrade should be its workspace

    A better AI agent may not require a different model. EnvOpt explores a third target: reorganizing the information and tools surrounding an otherwise fixed agent. The research paper was submitted November 5; Bespoke Labs published its companion exp…

    Sources: arxiv.org, bespokelabs.ai
  90. Agents

    JustRUA improves robot-agent scores by changing the harness, not training a new policy

    A November 11 preprint tests the software surrounding GPT-6 Astra rather than introducing another learned robot policy. The authors report an evaluation score rising from 29.7 to 50.3 in RoboDojo simulation. The retained changes preserve episode h…

    Sources: arxiv.org
  91. Agents

    An agent can stay inside its permissions and still do the wrong job

    Microsoft researchers propose an Intent Profile that captures an agent’s intended role and boundaries. Their November 12 preprint reports four-class macro-F1 increasing from 45% to 79% on a purpose-sensitive benchmark. Abstract and submission reco…

    Sources: arxiv.org
  92. Agents

    Qwen Code tightens approval handling while making agent wakeups durable

    Qwen Code's v0.27.3 release contains a practical permission fix: tool calls that match permissions.ask rules now require confirmation in auto-edit mode rather than being automatically approved, according to the project's official release notes. Th…

    Sources: github.com
  93. Tools

    HyperS2C pairs a chip-tool merger with an AI verification pitch

    HyperS2C launched in Beijing on November 13 with a proposition broader than another standalone coding assistant: bring digital chip-design and verification tools into one platform, with AI agents embedded in that workflow. A company-authored annou…

    Sources: eet-china.com, m.ithome.com
  94. Models

    Datasette opens a shorter path from local models to data agents

    Datasette's model-integration plugin has a useful new connection option. The 0.1a11 prerelease supports direct configuration of OpenAI-compatible endpoints without an extra provider plugin. That opens a simpler route to models served on a develope…

    Sources: github.com
  95. Policy

    Norway pauses new data-center grid reservations as it prepares a licensing regime

    The temporary grid measures have exceptions, while the government's broader licensing plans are still being developed. Norway is moving to regulate data-center growth through two distinct measures: a temporary pause in new grid-capacity reservatio…

    Sources: statnett.no, regjeringen.no