Latest in Discovery

Human & AI Reasoning
  • When Artificial Touch Improves Grasp Control

    When Artificial Touch Improves Grasp Control

    17 Nov 2026

    When a hand closes around a fragile object, the person controlling it needs to know when contact occurs and how firmly to hold it. A bidirectional brain–computer interface can return information from a sensor to the person while a decoder translates intended movement into action. Human experiments have tested how…

View collection →
Open Models
  • Dory1.0-4B and the value of a second pass

    Dory1.0-4B and the value of a second pass

    17 Nov 2026

    Can a language model solve a harder problem by applying the same learned transformation a second time? NVIDIA’s Dory1.0-4B makes this question experimentally tractable. Its central block can process the same sequence at different depths while the model’s four billion parameters remain fixed. NVIDIA reports a substantial gain from the…

View collection →
Biographies
  • Claude Shannon

    Claude Shannon

    16 Nov 2026

    As a boy in Michigan, Claude Shannon built a telegraph with a friend so that they could send messages between their homes. At Bell Laboratories he later asked how reliably any message could travel over a noisy line. His answer gave engineers a way to measure information and established a…

View collection →
Horizons of Electronics
  • When does cooling an AI processor from inside the silicon improve efficiency?

    When does cooling an AI processor from inside the silicon improve efficiency?

    18 Nov 2026

    Engineers can now etch cooling channels into the back of an AI processor and circulate liquid close to its hottest regions. Tests on a working CPU and an NVIDIA GH200 show what this changes for sustained clock speed, GPU temperature and nearby high-bandwidth memory. The electricity question requires a wider…

View collection →
AI’s Human Mirror
  • Walking After a Stroke: From the Clinic to a Gravel Path

    Walking After a Stroke: From the Clinic to a Gravel Path

    17 Nov 2026

    One person chose a gravel path in a park. Another chose a sidewalk. Both had survived a stroke, and both could already walk. What they wanted to try was a soft robotic support that gave the weaker ankle a push as the foot left the ground. First came a test…

View collection →
The AI Oligarchy and Its Suppliers
  • Zhang Peng, Co-founder and CEO of Z.ai

    What does Z.ai still control after it releases the weights of GLM-5.3?

    17 Nov 2026

    A cloud provider can download Z.ai’s GLM-5.3 and run it on its own computers. For some providers, using it commercially takes one more step. The difference between GLM-5.2 and GLM-5.3 reveals who decides, which companies face a review, and what choice they retain.

View collection →
New Programming Languages versus Rust / C++
  • Slang versus Rust and CUDA C++ in neural rendering

    Slang versus Rust and CUDA C++ in neural rendering

    17 Nov 2026

    A piece of leather looks different as the light moves. NVIDIA uses Slang to learn that changing appearance and evaluate it while drawing an image. GPU measurements show what automatic derivatives cost alongside handwritten CUDA C++. Dimforge chose Rust for its physics engine for a different reason.

View collection →
Dead Languages and the Past
  • What can AI find in a lidar map of the Maya landscape?

    What can AI find in a lidar map of the Maya landscape?

    17 Nov 2026

    On a map of the Maya lowlands, a small rise may be the surviving mass of an ancient building. It may also be a natural formation. Lidar makes the shape of the ground easier to examine beneath vegetation; artificial intelligence can help select shapes that deserve a closer archaeological look.…

View collection →
Articles by the Director — read the editorials

RSS feed: Google AI Google AI

  • Genie 4 is now available November 18, 2026
  • Introducing Gemini 4 Hydrogen November 12, 2026
  • AlphaGenome 2: Deciphering how cells read DNA to better understand disease November 11, 2026
    Predicting how DNA sequences and their variations impact specific human cell types and disease.

RSS feed: OpenAI OpenAI

  • Le Groupe Figaro and OpenAI partner to bring Le Figaro journalism to ChatGPT November 18, 2026
    Le Groupe Figaro and OpenAI partner to bring Le Figaro’s journalism to ChatGPT through summaries and excerpts, with links to original stories.
  • Introducing GPT-6.2 Astra and Sol November 17, 2026
    Introducing GPT‑6.2 Astra and Sol, with better professional outputs, coding, computer use, and efficiency at lower cost across ChatGPT Work, Codex, and the API.

RSS feed: MIT News – AI MIT News – AI

  • Helping AI text detectors keep up with changing conditions November 18, 2026
    A more flexible training method can make AI text detectors more robust when new models emerge or AI-generated text is disguised to appear human.
    Adam Zewe | MIT News
  • Putting a spotlight on AI, society, and humanity November 17, 2026
    Supported by the MIT Human Insight Collaborative, scholars are exploring how AI is impacting the human experience and reshaping higher education.
    Danna Lorch | School of Humanities, Arts, and Social Sciences

RSS feed: Berkeley AI Research Berkeley AI Research

  • BAIR at CoRL 2026 November 9, 2026
    The Conference on Robot Learning (CoRL) takes place November 9–12, 2026, in Austin, Texas. BAIR faculty, postdocs and students will be presenting research at various sessions, including an Early Career Keynote from Professor Negar Mehr. See below for the full schedule of sessions with BAIR members, including papers co-authored with collaborators at other universities and […]
  • From CUDA to MLX: How K-Search Brings Decades of Kernel Expertise to Apple Silicon July 29, 2026
    Figure 1: CUDA-to-MLX optimization translation map. CUDA optimization knowledge can be translated into architecture-native MLX strategies rather than copied instruction-for-instruction. We face a new epoch in computing. Hardware is changing rapidly — not just faster GPUs, but a growing range of chips from different vendors, each with its own architecture and often tailored to specific […]

RSS feed: METR — AI Capabilities & Risks METR — AI Capabilities & Risks

  • Independent Assessment of AI R&D Acceleration at Anthropic in October 2026 November 5, 2026
    METR's high-access assessment of the degree to which AI is accelerating AI research and development at Anthropic.
  • Review of the "Risks from automated R&D" section in Anthropic's Risk Report (August 2026) November 5, 2026
    External review from METR of the "Risks from automated R&D" section in Anthropic's August 2026 Risk Report

RSS feed: Apple — AI Research Apple — AI Research

  • Split-and-Scale Latent 3D Representations November 16, 2026
    Latent 3D representations have significantly expanded the capabilities of modern shape modeling, enabling compact encoding of 3D scenes and supporting powerful 3D generative models. Despite this progress, current representations struggle to capture the fine details of complex objects, fundamentally limited by the capacity of their latents. However, scaling latent capacity at training time is computationally […]
  • Theoretical Limits of Language Model Alignment November 12, 2026
    Language model (LM) alignment improves model outputs to reflect human preferences while preserving the capabilities of the base model. The most common alignment approaches are (i) reinforcement learning, which maximizes the expected reward under a KL-divergence constraint, and (ii) best-of-NNN alignment, which selects the highest-reward output among NNN independent samples. Despite their widespread use, the […]
  • Toward a Theory of Evaluability November 10, 2026
    Machine learning models, particularly large foundation models, are increasingly evaluated along multiple axes that are relevant for their usage. In addition to accuracy measures appropriate for the specific task of interest, measures capturing safety, calibration and fairness properties are often of interest. The classical approach to these problems assumes that the test set has been […]

AI on Apple Silicon

Run and study AI on Apple Silicon. MLX is Apple’s open-source framework for machine-learning research, model inference and fine-tuning. Its design takes advantage of unified memory on Apple devices.

Explore MLX and its examples ↗

RSS feed: NVIDIA — Technical Blog NVIDIA — Technical Blog

  • Build Training-Ready Physical AI Datasets with Two Prompts in NVIDIA Cosmos November 18, 2026
    World model accuracy starts with high-quality, diverse training datasets. Yet capturing rare defects, nighttime incidents, and other challenging scenarios is...
  • Tackling GPU Bottlenecks in Plane-Wave DFT with K-Point Batching and Mixed Precision November 18, 2026
    Training a next-generation machine learning interatomic potential (MLIP) requires ground truth: tens to hundreds of millions of density functional theory (DFT)...
  • Co-Designing Sparse Attention for Efficient Long-Context Inference November 17, 2026
    Agentic workloads, long documents, and large codebases are pushing context lengths from tens of thousands of tokens to hundreds of thousands. At those lengths,...
  • Integrating QPUs into Supercomputing Centers Using Slurm HRES Plugins November 17, 2026
    Quantum processors are being integrated into supercomputing centers worldwide. National programs in North America, Europe, and Asia are driving the installation...

Groq — AI inference

Specialised processors for AI inference. Groq develops inference technology and cloud infrastructure. NVIDIA entered a non-exclusive licensing agreement with Groq in December 2025; Groq joined the NVIDIA Cloud Partner programme in August 2026.

Technology agreement · 24 Dec 2025 ↗
Cloud partnership · 12 Aug 2026 ↗

Latest official Groq updates ↗

RSS feed: AMD — ROCm Blogs AMD — ROCm Blogs

  • An AITER Pin Keyed by Kernel Text: Faster Glue Than the Inductor Default November 17, 2026
    GEMM and attention kernels already have offline tuning paths. The kernels around them often do not. A compiled inference graph still launches a long tail of small Triton kernels for casts, concatenations, normalizations, activations, gathers, and histogram-style updates. TorchInductor generates those kernels. With autotune on, it times a candidate list and writes the winner into […]
  • Keep the Agent Warm: How ATOM Serves AgentX Workloads on AMD Instinct MI355X GPUs November 17, 2026
    A coding agent makes many model calls in one session, and each one re-sends almost everything before it: one recorded session makes 198 calls, and call #48 sends 195.5K tokens, only 1.3K of them new. Agentic workloads combine long shared histories and short outputs, interrupted by tool calls and parallel sub-agents. Serving them well takes […]

RSS feed: Intel OpenVINO — AI inference Intel OpenVINO — AI inference

  • Accelerating LLM Inference with Multi-Token Prediction (MTP) and OpenVINO™ November 6, 2026
    A step-by-step guide to Multi-Token Prediction in OpenVINO™ GenAI 2026.4Local LLM inference has gotten faster every release, but the basic shape of decoding hasn’t changed much: one token out, one forward pass in, repeat. Speculative decoding broke that rule by letting a second, smaller model guess ahead while the main model checks its work. OpenVINO™ GenAI […]
  • Precision Document AI, Day 0: MiniCPM-OCR Is Now Available on Intel Hardware with OpenVINO™ October 27, 2026
    Most OCR pipelines answer one question: what does this document say? MiniCPM-OCR answers a second one too: can you trust that answer? Today, Intel® announces day 0 support for MiniCPM-OCR, a new two-model series from ModelBest* and Tsinghua University* that pairs a compact 1.3B document parsing model with a reliability judge that automatically evaluates OCR […]

RSS feed: AWS Trainium — AI chips AWS Trainium — AI chips

  • Deploy frontier AI models on AWS Trainium with SGLang-Neuron November 12, 2026
    SGLang-Neuron is now publicly available through an AWS Deep Learning Container. With SGLang-Neuron, you can serve models and generate reinforcement learning rollouts with SGLang on AWS Trainium using the same SGLang interface you already know, all running on native PyTorch.
  • Train Qwen3 on AWS Trainium with torchtitan October 21, 2026
    Train Qwen3 on AWS Trainium with torchtitan three ways on a single trn2.48xlarge: the Hugging Face Transformers v5 modeling backend, torchtitan's native Qwen3, and NKI mixture-of-experts kernels. The same torchrun launcher and config structure work unchanged across all three paths, including loading and saving published Hugging Face safetensors.

RSS feed: Qwen Code — Release notes Qwen Code — Release notes

  • Release v0.28.1-preview.0 November 17, 2026
    What's Changed test(managed-agent): observe a real scheduled /loop fire in the full-stack E2E by @wenshao in #17506 fix(verify-pr): harden the watch-state shape guard, container nonce, and VM prose by @wenshao in #17440 fix(external-context): deflake cumulative-deadline deletion timing by @qwen-code-dev-bot in #17342 docs(design): propose idle Session suspension toward a stateless Hosted Harness by @doudouOUC in #17527 […]

RSS feed: Moonshot AI · Kimi Moonshot AI · Kimi

  • Admin API is now available on Kimi Open Platform November 2, 2026
    【Admin API is now available on Kimi Open Platform.】When multiple teams in your organization use Kimi API, managing accounts, access, costs, and security becomes increasingly complex. Admins handle onboarding and offboarding, finance teams reconcile bills and usage, and security teams review audit logs—often through repetitive manual work in the console. Kimi Open Platform now offers […]
  • 📢 Kimi K3.1 Pro has launched! Start exploring now! October 23, 2026
    We’re introducing Kimi K3.1 Pro, the first model in our new K3.1 series and our most capable 3T-class open-weight model to date, with a 1-million-token context window. It delivers substantial improvements across coding, 3D creation, financial research, and other agentic capabilities. We are also releasing Kimi K3.1 Pro Highspeed, which delivers 2-3× the speed of […]

RSS feed: OpenMined open-source community OpenMined open-source community

  • How to scale the impact of embedded AI evaluators September 19, 2026
    AI labs are promising independent evaluators deep access to their models. Doing that at scale means removing the friction of sharing weights, data, and logs.

RSS feed: Montreal AI Ethics Institute Montreal AI Ethics Institute

  • The AI Ethics Brief #203: Power, Fracture, Resistance November 10, 2026
    Introducing the State of AI Ethics Report, Volume 8 (2026) Source

RSS feed: IEEE Technology and Society IEEE Technology and Society

  • Greg Adamson Receives 2026 Brian M. O’Connell SSIT Distinguished Service Award November 6, 2026
    Greg Adamson was presented with the 2026 Brian M. O’Connell SSIT Distinguished Service Award on November 5, 2026. The post Greg Adamson Receives 2026 Brian M. O’Connell SSIT Distinguished Service Award appeared first on IEEE Technology and Society.
    terribookman

RSS feed: European Digital SME Alliance European Digital SME Alliance

  • Europe’s Digital Networks Act must do more than strengthen telecoms. It must strengthen Europe’s digital economy. November 10, 2026
    The post Europe’s Digital Networks Act must do more than strengthen telecoms. It must strengthen Europe’s digital economy. appeared first on European DIGITAL SME Alliance.
    Sofia Gottarelli

RSS feed: natura natura

  • At-home microbiome tests are taking off. Can they be trusted? November 18, 2026
    Nature, Published online: 18 November 2026; doi:10.1038/d41586-026-03710-yOur reporter mailed off her poo to find out what two companies say about her microbiome. Here’s what she found.
    Amber Dance

RSS feed: MedRxiv MedRxiv

  • Microcirculation Mirror: A zero-install browser workstation for hand microvascular anatomy and nailfold capillaroscopy November 17, 2026
    Background Hand microvasculature is conventionally presented through diagrams, cadaveric atlases and specialized optical instruments, while conventional clinical radiography cannot resolve individual capillary loops (5-15 m lumen) in vivo, which dominate cutaneous oxygen exchange. Methods We present Microcirculation Mirror, a zero-install browser workstation that registers layered hand anatomy to a live 21-landmark pose and couples it […]
    Wang, Y., He, J., Liu, W., Xu, X., Li, B., Li, Y., Wang, B., Liu, M.

RSS feed: Plos Medicine Plos Medicine

  • Economic value of gonorrhea resistance testing in the United States: A multi-perspective modeling study November 17, 2026
    by Brooke E. Nichols, Betsy Wonderly Trainor, Grace Hampson, Yonatan H. Grad, Jeffrey D. Klausner Background Rising antimicrobial resistance in Neisseria gonorrhoeae threatens the effectiveness of existing therapies. Ceftriaxone is the only recommended first-line (empiric) treatment for uncomplicated gonorrhea in the United States. Ciprofloxacin is no longer recommended for empiric use because of widespread resistance, […]
    Brooke E. Nichols

RSS feed: IEEE Engineering in Medicine & Biology Society IEEE Engineering in Medicine & Biology Society

  • Call for Student/YP Deputy Editor September 29, 2026
    The post Call for Student/YP Deputy Editor appeared first on IEEE EMBS.
    Deidre Artis

RSS feed: Towards Data Science Towards Data Science

  • When Qini Rewards the Wrong Uplift Model, Use Doubly Robust Scores November 18, 2026
    On 1 million simulated customers, calibrated Qini scored a Linear DR learner 0.0856 against 0.0826 for the true-CATE ranking that should have won. The post When Qini Rewards the Wrong Uplift Model, Use Doubly Robust Scores appeared first on Towards Data Science.
    Ivan Gorban

RSS feed: Machine Learning Mastery Machine Learning Mastery

  • Graph-RAG for Enterprise Documents: Building a Knowledge Graph from PDFs, Wikis, and Databases November 18, 2026
    Prerequisites You'll need Python 3.
    Shittu Olumide

RSS feed: Rbloggers Rbloggers

  • "You get a free drink if you order a Burger or a Salad and a Dessert." November 17, 2026
          A restaurant advertises, "You get a free drink if you order a Burger or a Salad and a Dessert." Suppose you only order a Burger. Should the restaurant give you a free drink?     &nb... Continue reading: "You get a free drink if you order a Burger or a Salad and a Dessert."
    Jerry Tuttle

RSS feed: Ars Technica Ars Technica

  • Vulcan failure caused by inadequate fuel supply to engines, ULA says November 18, 2026
    "ULA is a valued partner, and we look forward to Vulcan's return to flight."
    Eric Berger

RSS feed: TechCrunch TechCrunch

  • China dominates battery graphite, but this startup could break the monopoly November 18, 2026
    Chinese companies maintain near total dominance of the graphite industry, but Molten Industries says its new factory in Oakland could spell the end of the monopoly.
    Tim De Chant

RSS feed: Verge Science Verge Science

  • Nvidia isn’t selling any more AI chips to China November 18, 2026
    Nvidia has stopped sending its advanced AI chips to China, despite the approval it received from the White House to do so in December last year. The announcement was made by interim Nvidia CEO Colette Kress during an earnings call on Tuesday, marking a notable shift for the company that, under former CEO Jensen Huang, […]
    Jess Weatherbed

RSS feed: Claude — Service status Claude — Service status

  • Elevated errors affecting the Google Docs connector November 18, 2026
    Nov 18, 01:44 UTC Resolved - The issue affecting editing documents with the Google Docs connector on claude.ai, Claude Code, and Claude Cowork has been resolved. Impact occurred from 16:19 PT on November 17 / 00:19 UTC on November 18 to 17:33 PT on November 17 / 01:33 UTC on November 18. Nov 18, 01:40 […]
  • Claude unable to join meetings November 18, 2026
    Nov 17, 23:39 UTC Update - Claude Meetings (beta) is joining calls more reliably, and we are monitoring it. It remains enabled for all Enterprise customers. Team customers can again add Claude to new calls manually, and Auto-join is being restored gradually for Team customers. Nov 17, 02:02 UTC Identified - Claude Meetings (beta) saw […]
  • Some users unable to sign in with Apple on claude.ai and the Claude iOS app November 17, 2026
    Nov 17, 21:51 UTC Resolved - The issue affecting Sign in with Apple on claude.ai and the Claude iOS app has been resolved. Impact occurred from 08:40 PT / 16:40 UTC to 12:20 PT / 20:20 UTC. Nov 17, 18:56 UTC Update - We have identified the cause of the issue affecting claude.ai and are […]

RSS feed: Grok / xAI — Service status Grok / xAI — Service status

  • [Grok Bot] Elevated errors for some requests November 12, 2026
    Status: RESOLVEDSeverity: disruptionResolved: Thu, 12 Nov 2026 21:39:26 GMTUpdates:Thu, 12 Nov 2026 21:39:26 GMTResolvedThis incident has been resolved.Thu, 12 Nov 2026 20:43:56 GMTMonitoringA fix has been implemented and we are monitoring for service restoration.Thu, 12 Nov 2026 20:33:19 GMTInvestigatingWe've identified the cause and are rolling out a fix. We'll share another update soon.Thu, 12 Nov […]

RSS feed: DeepSeek — Service status DeepSeek — Service status

  • DeepSeek 网页/Chat 性能下降(DeepSeek Web/Chat Degraded Performance) November 14, 2026
    Status: resolved本次问题已解决,服务已恢复。 The incident has been resolved and service has been restored.Affected components: 对话服务(Chatservice)
  • DeepSeek 网页/API 部分中断(DeepSeek Web/API Partially Unavailable) November 7, 2026
    Status: resolved本次问题已解决,服务已恢复。 The incident has been resolved and service has been restored.Affected components: DeepSeek V4 Pro API服务(API Service), DeepSeek V4.1 Flash API服务(API Service), 对话服务(Chatservice), 搜索服务(Search Service)
  • DeepSeek 网页/API 部分中断(DeepSeek Web/API Partially Unavailable) November 3, 2026
    Status: resolved本次问题已解决,服务已恢复。 The incident has been resolved and service has been restored.Affected components: DeepSeek V4.1 Flash API服务(API Service), 对话服务(Chatservice), DeepSeek V4 Pro API服务(API Service)
follow us in feedly Flipboard