News

News from November 10, 2026

In my view, the Gemini 4 Carbon release and results reported by Google DeepMind and Amazon show why we need to evaluate capabilities and costs before promising AI benefits.

The day in three lines

  1. Google began releasing Gemini 4 Carbon, with a tool that splits tasks among artificial intelligence agents.
  2. Google DeepMind researchers say AI agents mathematically proved that parts of Signal’s code protect messages.
  3. Two Amazon engineers report cutting internal ticket labeling costs by 98% using another AI model while maintaining similar results.
  1. Smart home devices should work without the internet

    The author keeps essential home controls working without internet by using equipment he owns instead of relying on companies’ servers.

    Hacker News728 points · 259 comments

    Quote
    Rule one is that nothing in my house needs a vendor's server to work. The rest of the rules I follow after ten years of home automation, plus both talks I gave on it, slide by slide.
  2. The Case for a Bubble at Life’s Start

    The lipid-first theory proposes that a cell membrane gathered chemicals and thereby made the emergence of life possible.

    ScienceQuanta Magazine

    Quote
    Is self-enclosure life’s greatest trick? According to the lipid-first theory, a cell membrane enabled life by giving chemistry a place to gather. The post The Case for a Bubble at Life’s Start first appeared on Quanta Magazine
  3. The Download: Pelosi’s successor targets AI and data centers face power woes

    Nancy Pelosi’s successor, Scott Wiener, wrote a bill aimed at reducing the risks of advanced artificial intelligence.

    TechMIT Technology Review

  4. Helping caregivers use AI in everyday life

    OpenAI Academy will bring free, hands-on ChatGPT workshops to caregivers in five U.S. cities.

    AI labOpenAI

    Quote
    OpenAI Academy is bringing free, hands-on ChatGPT workshops to caregivers in five U.S. cities, helping them build practical AI skills for everyday life.
  5. Qual é o segredo de Xangai no ensino de matemática?

    The international assessment tests 15-year-olds in mathematics, science, and reading, and publishes average scores for participating countries.

    BrazilFolha, Ciência

    Quote
    O Programa Internacional de Avaliação de Estudantes ( Pisa, na sigla em inglês) é realizado a cada três anos pela Organização para a Cooperação e o Desenvolvimento Econômico ( OCDE ).
  6. Learning Heterogeneous Tactile Representation for General Dexterous Manipulation

    A study presents a model that combines data from different touch sensors and improves robots’ performance on manipulation tasks.

    Papernot peer reviewed · 92 votes

    Quote
    Tactile sensing is an indispensable modality for contact-rich dexterous manipulation, a capability that hints at the ultimate frontiers of embodied intelligence.
  7. Correcting quantum computer errors may be more complicated than expected

    Studies suggest errors in quantum computers can retain memory and influence one another, complicating the methods used to correct them.

    ScienceNew Scientist

    Quote
    Correcting quantum computer errors is key for making the devices practical, but some types of errors, and the complications they could introduce, may be more likely to occur than we had thought
  8. ‘We Have to Take Over’: Far-Right Extremists Want to Seize Control of the GOP

    Far-right extremists say the MAGA movement failed because it did not go far enough, and see election losses as a chance to take over the Republican Party.

    TechWIRED

    Quote
    MAGA failed because it didn’t go far enough, right-wing extremists say—and following devastating midterm losses, they see an opportunity to take over.
  9. Gemini 4 Carbon is now available

    Google DeepMind’s new model is now available to developers, businesses, scientists, and consumers.

    AI labGoogle DeepMind

    Quote
    Today, we’re making Gemini 4 Carbon available to developers, enterprises, scientists, and consumers.
  10. Espanha pede suspensão do TikTok após denúncias de abuso sexual infantil

    Spain asked TikTok to suspend operations until it removes child sexual abuse content, after an investigation found 1,183 posts.

    BrazilG1, Tecnologia

    Quote
    Investigação identificou 1.183 publicações explícitas na plataforma, com pelo menos 48 milhões de visualizações; governo acionou Ministério Público e órgão de proteção de dados.
  11. .name Grandfathered

    Verisign dropped its plan to terminate 22,000 existing domains and email redirects, but will stop accepting new registrations for third-level .name domains.

    Hacker News695 points · 162 comments

  12. These two simple numbers can predict your heart disease risk

    A new calculation using two commonly tracked measures may help predict long-term heart disease risk.

    ScienceBBC Science Focus

    Quote
    Could a new calculation based on two commonly tracked metrics offer a cheat code for your long-term health?
  13. Speech Processing Pioneer Bishnu Saroop Atal Dies at 93

    Bishnu Saroop Atal died at 93; his techniques helped transmit clear voice calls while using less space on cellular networks.

    TechIEEE Spectrum

    Quote
    Bishnu Saroop Atal Acoustics and signal processing pioneer Life Fellow, 93; died 19 August Atal was an acoustics and signal processing researcher at Bell Labs in Murray Hill, N.J., and later at AT&T ’s Shannon Laboratory, in Florham Park…
  14. Gemini Intelligence: New features, on more devices

    Gemini Intelligence is coming to more devices with expanded features.

    AI labGoogle AI

    Quote
    Gemini Intelligence is rolling out to more devices. See how expanded capabilities across these devices help you do more.
  15. Disney e Adobe anunciam parceria para desenvolvimento de IA

    Disney and Adobe partnered to provide artificial intelligence models for films, theme parks, and other Disney creations.

    BrazilTecnoblog

    Quote
    Modelos proprietários serão voltados para produções audiovisuais e até parques temáticos da gigante do entretenimento.
  16. WorldFoundry: Towards Unified Infrastructure for World Intelligence

    Researchers introduced an open platform for running and comparing systems that simulate environments, track changes, and respond to actions.

    Papernot peer reviewed · 54 votes

    Quote
    Visual generative models increasingly blur the line between synthesizing observations and simulating environments.
  17. College student develops LEGO-like, 3D-printed moon bricks

    Prototypes made from artificial lunar soil hold together without glue.

    SciencePopular Science

    Quote
    The prototypes made from artificial lunar regolith don’t require any binding.
  18. Seagate's new hybrid drive raises some old questions

    Seagate launched drives that combine a hard drive with a separately purchased SSD, but has not explained how they protect data if either drive fails.

    TechArs Technica

    Quote
    The FireCuda Forge revisits an old idea, with some familiar questions.
  19. Introducing Businesses in ChatGPT

    Businesses can manage their information, see how they appear in ChatGPT, and turn users’ interest into action.

    AI labOpenAI

    Quote
    Manage your business information, get insights about how your business shows up, and turn interest into action in ChatGPT.
  20. Como identificar teorias de conspiração e não acreditar nelas

    In 2017, American student Peter McIndoe brought a sign claiming birds do not exist to a rally for Donald Trump supporters.

    BrazilFolha, Ciência

    Quote
    "Birds aren't real" (os pássaros não são reais). Foi isso que o estudante americano Peter McIndoe escreveu em um pedaço de papelão ao se misturar, em 2017, a uma manifestação de apoiadores do presidente dos Estados Unidos, Donald Trump.
  21. Amazon saves millions by replacing Sonnet 4.6 with Strands Decider

    Two Amazon engineers switched to another model for labeling internal tickets, achieving similar results while cutting costs by 98%.

    Hacker News689 points · 242 comments

    Quote
    Two Amazon engineers swapped Claude Sonnet 4.6 for a decision model in one of the world's biggest internal ticketing systems.
  22. Science to restore speech is moving fast. Here’s what’s next

    Scientists are developing technologies to help people who cannot speak communicate, while users point to improvements they would still like to see.

    ScienceScience News

    Quote
    As neurotechnology aims to restore seamless speech to those who can no longer talk, what’s next for the science, and for the people it promises to help?
  23. Microsoft’s DLSS competitor is out for the Xbox Ally X and works in handheld mode

    Microsoft is rolling out technology that improves game resolution on the Xbox Ally X, Xbox Ally X20, and some Windows PCs.

    TechThe Verge

    Quote
    Microsoft is now rolling out Automatic Super Resolution (Auto SR), an upscaling technology that competes with Nvidia's DLSS, for the Xbox Ally X, Xbox Ally X20, and select Windows Copilot Plus PCs, the company announced today.
  24. We’re bringing Adobe’s creative and productivity tools closer to your everyday Google apps.

    Google and Adobe are integrating tools to edit PDFs in Drive and use Adobe’s artificial intelligence features in Gemini.

    AI labGoogle AI

    Quote
    Google and Adobe are integrating creative and productivity tools. Edit PDFs in Drive and use Adobe AI in Gemini. See what’s new in this partnership.
  25. Anthropic, dona do Claude, tem IPO estimado em US$ 1,4 trilhão, diz agência

    Anthropic plans to go public at a valuation of up to $1.4 trillion and raise up to $90.4 billion through two offerings.

    BrazilG1, Tecnologia

    Quote
    Oferta pode movimentar até US$ 90,4 bilhões, em meio à desaceleração do entusiasmo dos investidores com empresas de inteligência artificial.
  26. Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery

    An AI system that adjusts research ideas based on evidence and open questions helped guide experiments that improved the yields of two chemical reactions.

    Papernot peer reviewed · 53 votes

    Quote
    We introduce Sage-Taiyi, a metacognition-inspired architecture that uses accumulated scientific understanding to guide what to investigate next and how to investigate it.
  27. Was Isaac Newton’s contribution to calculus stolen?

    Newton accused Leibniz of stealing his ideas about calculus, but historians conclude that both probably made their discoveries independently.

    ScienceScientific American

    Quote
    Newton accused Gottfried Wilhelm Leibniz, another mathematical giant of the time, of stealing the concept of infinitesimals. Here’s the real story
  28. X launches its own clipping feature to spur influencer engagement

    X launched a tool for trimming videos and posting clips on the platform, with links to the original and possible rewards for views.

    TechTechCrunch

    Quote
    X is jumping into clipping, the popular practice of cutting long videos into short, shareable clips that can go viral.
  29. Baidu: ERNIE-5.5-Flash-Preview

    The latest ERNIE model brings improvements for writing code, handling tasks, and researching information.

    AI labOpenRouter

    Quote
    ERNIE 5.5 Flash Preview is the latest foundation model in the ERNIE family, with comprehensive improvements in coding, agentic capabilities, knowledge work, and deep research. It has 280 billion total...
  30. Google libera Gemini 4 Carbon ao público com recurso para equipes de agentes de IA

    Google began releasing Gemini 4 Carbon to subscribers and developers, with a tool that divides tasks among artificial intelligence agents.

    BrazilTecnoblog

    Quote
    Modelo chega a assinantes do Google AI Ultra e desenvolvedores, com foco em tarefas complexas de programação.
27 more
  1. AlphaProof Code

    Google DeepMind researchers say artificial intelligence agents mathematically proved in four weeks that parts of Signal’s code protect messages.

    Hacker News582 points · 178 comments

  2. MIT Schwarzman College of Computing welcomes seven new faculty for 2026

    MIT welcomed seven new computing faculty members whose research spans artificial intelligence, the environment, work, finance, and quantum computing.

    ScienceMIT News

    Quote
    Faculty members serve in core computing and shared roles, bringing a wide range of experiences and areas of expertise to the MIT community.
  3. Anthropic tiptoes into MS Office territory

    Anthropic launched tools for spreadsheets, whiteboards, and data in Claude, expanding its work features and bringing it closer to office suites.

    TechThe Register

    Quote
    Sheets, Whiteboards, and Data artifact types bring Claude closer to a productivity suite
  4. Baidu: ERNIE-Light-DT-1.0

    The model combines open-source models with research from Baidu Search and a method that focuses processing on selected parts of the information.

    AI labOpenRouter

    Quote
    ERNIE-Light-DT-1.0 is based on open-source GLM models and includes research findings from Baidu Search's Lightning large model. It features Baidu Sparse Attention (BSA), a unique sparse attention mechanism developed by...
  5. A fascinante descoberta de uma vitória-régia gigante na América do Sul

    The discovery came at the last minute, when Carlos Magdalena and Mauricio Bernal were about to give up.

    BrazilFolha, Ciência

    Quote
    A descoberta ocorreu no último minuto, quando os pesquisadores Carlos Magdalena e Mauricio Bernal se davam por vencidos.
  6. Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning

    A study introduces a method that trains visual agents to make better use of tool observations and reports stronger results on image and video tests.

    Papernot peer reviewed · 38 votes

    Quote
    Agentic reinforcement learning (RL) and multi-teacher on-policy distillation (MOPD) offer complementary routes to post-training. For visual agents, however, tool use presents two related challenges.
  7. NASA Invites Media, Public to Join Voyager 1 One-Light Day Activities

    On November 18, Voyager 1 will be so far away that light from Earth will take 24 hours to reach it.

    ScienceNASA

    Quote
    Launched nearly 50 years ago in 1977, NASA’s Voyager 1 on Wednesday, Nov. 18 will have traveled one light-day from Earth, which is more than 16 billion miles (25.9 billion kilometers).
  8. Evolving metal extraction processes

    Most metal extraction methods still rely on techniques familiar to metallurgists from earlier generations.

    TechMIT Technology Review

    Quote
    The metals industry has experienced transformational moments before: the Bessemer process, flash smelting, electric arc furnaces, and modern continuous casting all changed what was possible.
  9. Google: Gemini 4 Carbon

    Google’s Gemini 4 Carbon is a model for coding and complex reasoning that accepts text, images, audio, video, and files.

    AI labOpenRouter

    Quote
    Gemini 4 Carbon is a frontier Gemini 4 model from Google for coding, agentic and complex reasoning work. It takes text, image, audio, video and file input, supports long decode...
  10. Fabricante dos chips da Nvidia vai investir US$ 1 bilhão na OpenAI e na Anthropic

    TSMC plans to invest up to $500 million in each of OpenAI and Anthropic, deepening its bet on artificial intelligence.

    BrazilG1, Tecnologia

    Quote
    Fabricante taiwanesa destinará até US$ 500 milhões (R$ 2,76 bilhões) a cada empresa e reforçará aposta no crescimento da inteligência artificial.
  11. Blender 5.3 released with Gaussian Splat support

    Blender 5.3 can now import, edit, and render 3D scenes reconstructed from photos or videos using small transparent shapes called Gaussian splats.

    Hacker News439 points · 68 comments

    Quote
    Gaussian splats, DLSS denoising, light dispersion, EEVEE fisheye, Outliner sorting, ripple video editing, and so much more.
  12. Scientists discover why some mosquitoes love you and others don't

    A study found that different mosquito species are attracted to distinct combinations of odors, chemicals, and bacteria on human skin.

    ScienceScienceDaily

    Quote
    A surprising new study reveals that mosquitoes have very different tastes in humans, with each species attracted to its own combination of skin odors, chemicals, and bacteria.
  13. Nex AGI: Nex-N2.6

    Nex AGI built an artificial intelligence model to use computers, write code, and handle long tasks with tools.

    AI labOpenRouter

    Quote
    Nex-N2.6 is an agentic model from Nex AGI, built on GLM-5.3-Flash with about 321B parameters. It is suited for computer use, coding, and long-horizon tasks that involve tools. Compared with...
  14. Flide: Accelerating Decision Inference with Exact State Forking

    A study presents a system that speeds up branching answers from AI agents while preserving the results of separate calculations.

    Papernot peer reviewed · 32 votes

    Quote
    Decision models such as Jev have emerged as a highly promising paradigm for LLM agent systems, providing lightweight and reliable categorical reasoning for rapid System One execution.
  15. What terrifying sleep paralysis demons reveal about the mind-body connection

    Neuroscientist Baland Jalal studied sleep paralysis after waking with the feeling of a shadowy presence pressing on his chest.

    ScienceNew Scientist

    Quote
    After waking with the feeling of a shadowy presence pressing on his chest, neuroscientist Baland Jalal went on to study sleep paralysis. He explains what this frightening condition might reveal about consciousness
  16. AI usage generates a fragile codebase

    A developer argues that AI-generated code can be hard to test, understand, and maintain, increasing failures and future problems.

    Hacker News395 points · 417 comments

  17. This bear’s big butt is so beautiful, it just won an award

    Volverine, an iconic mother bear, won the 2026 Big & Beautiful Award.

    SciencePopular Science

    Quote
    The iconic mama bear named Volverine wins the 2026 Big & Beautiful Award.
  18. ReproBridge: Building Reproducibility into ML Research Artifacts

    A study finds that organized code, setup, and instructions help agents repeat machine-learning experiments more reliably and at lower cost.

    Papernot peer reviewed · 30 votes

    Quote
    Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, released code, and supporting artifacts.
  19. Toxic secrets are hiding in Arctic ice

    A study links toxic metals found in Arctic ice to emissions from a Montana mining and smelting company more than a century ago.

    ScienceScience News

    Quote
    Toxic metals can harm human health. Arctic ice cores trace arsenic, lead and other heavy metals to historic mining and smelting in Montana.
  20. Scion is back and is going to sell a kei truck in America

    Toyota is reviving Scion to sell a small truck in the U.S., limited to low-speed roads and initially available only to fleets from 2027.

    Hacker News344 points · 256 comments

    Quote
    Toyota's Scion is back and this time it's weird. The new Scion is going to sell a Kei truck in America, plus a bike and a hybrid side-by-side
  21. Longtail-SWE: High-Quality SWE Tasks for Underrepresented Languages

    Researchers created a set of 202 programming tasks in 13 underrepresented languages, revising tests and descriptions to evaluate solutions accurately.

    Papernot peer reviewed · 20 votes

    Quote
    Challenging repository-level benchmarks remain limited for many programming language ecosystems, where filtering scarce candidates with ambiguous descriptions or defective tests can discard valuable engineering problems.
  22. Val Town is winding down

    Val Town is shutting down because it could not compete in cloud computing and has paused customer billing.

    Hacker News313 points · 82 comments

  23. InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure

    A study presents a system that adjusts AI training settings and code, reporting faster results across four tasks.

    Papernot peer reviewed · 17 votes

    Quote
    While Recursive Self-Improvement (RSI) accelerates isolated algorithm and kernel synthesis, distributed LLM training depends on tightly coupled launch settings, operator kernels, and communication schedules, where single edits often hit…
  24. Synaptics patent claims stop work on DisplayLink drivers

    Synaptics threatened legal action against a developer for creating an independent Linux driver for a DisplayLink dock.

    Hacker News305 points · 166 comments

    Quote
    Mike Lothian owns a docking station (or dock) that uses a DisplayLink chipset that had, at leas
  25. SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research

    A study tested whether language models can tell when simulator flaws change which systems design should be chosen and found frequent mistakes.

    Papernot peer reviewed · 12 votes

    Quote
    LLM agents increasingly drive systems research and rely on simulators to evaluate candidate designs at scale.
  26. Ableton Live 13

    Ableton announced Live 13, planned for 2027, with new tools for making beats, editing audio, automating controls, and mixing music.

    Hacker News278 points · 301 comments

    Quote
    See what’s coming in Ableton Live 13: experimental sound design tools, precision mixing effects and workflow improvements to bring your ideas into focus.
  27. How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate?

    A study finds models learning from their own guided responses tolerate execution differences, but lose quality when these grow large.

    Papernot peer reviewed · 11 votes

    Quote
    On-policy distillation (OPD) trains a student language model on its own responses, guided by a fixed teacher. Generating these responses is expensive, so systems use quantized inference and asynchronous rollouts to reduce the cost.
This week’s papers

How this is made

A script picks the links by votes or by date. An AI summarizes each one using only the source’s text, and writes the day’s three lines from those summaries. Nobody reviews them before they go up, so check the original before citing.

Data in JSON · RSS

Sources: Hacker News, arXiv, Hugging Face, OpenRouter, Quanta Magazine, New Scientist, BBC Science Focus, Popular Science, Nature, Science News, Scientific American, MIT News, NASA, ScienceDaily, Agência FAPESP, Pesquisa FAPESP, Folha, Ciência, G1, Tecnologia, Tecnoblog, MIT Technology Review, WIRED, IEEE Spectrum, Ars Technica, The Verge, TechCrunch, The Register, Simon Willison, OpenAI, Google DeepMind, Google AI, Hugging Face.