News

News from October 29, 2026

In my view, Google’s ads and OpenAI’s updated guidance show that AI reflects companies’ choices rather than a neutral perspective.

The day in three lines

  1. Mistral made the files for its Large 4 model available for others to use.
  2. Google plans to show ads in its artificial intelligence answers in Brazil later in 2026.
  3. OpenAI updated its guidance with five values to guide AI decisions and respect people’s choices.
  1. Patrick Stewart has died

    No summary: the source’s text could not be read.

    Hacker News3,310 points · 769 comments

  2. How Do You Take a Picture of a Black Hole?

    An astrophysicist explains that photographing the first black hole required telescopes around the world, and the next goal is to make a movie.

    ScienceQuanta Magazine

    Quote
    Astrophysicist Shep Doeleman explains how capturing the first image of a black hole required a telescope the size of the world, and why the next goal is to create a movie.
  3. Consumer Appetite for Personal AI Goes Mainstream

    A survey of more than 12,000 people points to growing interest in AI-powered wearables, while highlighting privacy, security, and control as conditions for adoption.

    TechIEEE Spectrum

    Quote
    This article is brought to you by Qualcomm Technologies. For many consumers, much of the promise and appeal of AI is personal.
  4. Fair use benefits everyone

    OpenAI shares excerpts from legal arguments saying fair use supports artificial intelligence training and innovation.

    AI labOpenAI

    Quote
    OpenAI shares excerpts from court briefs by creators, academics, accessibility advocates, and legal experts, among others, on how fair use supports AI training and innovation.
  5. Jornalismo da FAPESP está entre os mais admirados do país

    Agência FAPESP, Pesquisa FAPESP magazine and editor Gabriel Alves were recognized in an award for health, science and well-being journalism.

    BrazilPesquisa FAPESP

    Quote
    Sexta edição do Prêmio Einstein +Admirados da Imprensa de Saúde, Ciência e Bem-Estar destaca o trabalho da Agência FAPESP, da revista Pesquisa FAPESP e do diretor de redação Gabriel Alves
  6. OmniShowcase: Omnidirectional Subject-Consistent Video Generation from Multi-View Images

    A study presents a method that uses images from several angles and shape maps to generate videos that keep the subject consistent.

    Papernot peer reviewed · 58 votes

    Quote
    Generating videos of a specific subject from a few reference images remains challenging when the camera reveals surfaces that are not visible in the input.
  7. These might be the nine best science photos from the past 30 years

    The Wellcome Trust marks three decades of images related to science and health, highlighting nine photos that may be among the best.

    ScienceBBC Science Focus

    Quote
    The Wellcome Trust celebrates 30 years of awe-inspiring images at the forefront of science and health
  8. Granite Switch 4.1.1 — Update models like you update software

    IBM released a compatible Granite Switch update with new speech and decision-making features, without changing the base model.

    AI labHugging Face

    Quote
    A Blog post by IBM Granite on Hugging Face
  9. Apocalipse? Por que OpenAI deixou mundo da matemática em choque

    In early October, OpenAI released hundreds of mathematical results that, the company said, could expand human knowledge.

    BrazilFolha, Ciência

    Quote
    No início de outubro, a OpenAI publicou um conjunto de centenas de resultados matemáticos que, segundo a empresa, poderia "ampliar as fronteiras do conhecimento humano".
  10. Binance-hacker-1029 – $5.6B stolen from Binance

    No summary: the source’s text could not be read.

    Hacker News1,763 points · 1,267 comments

  11. Gamma-ray burst is one of the brightest in decades

    It released energy equal to what the Sun produces in a billion years through a pencil-sized jet.

    SciencePopular Science

    Quote
    It shot out a billion years’ worth of solar energy through a pencil-sized jet.
  12. An updated Model Spec for more capable and beneficial AI

    OpenAI updated its guidance, listing five values to help AI make sound decisions, respect people’s choices, and follow their intentions.

    AI labOpenAI

    Quote
    OpenAI’s updated Model Spec outlines five core values to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.
  13. Justiça do Trabalho manda Meta remover publicações que incentivam demissão por causa do voto

    A labor court ordered Meta to remove posts encouraging firings over voting and tell Brazilian users that political discrimination at work is prohibited.

    BrazilG1, Tecnologia

    Quote
    Decisão também obriga a empresa a avisar usuários de Instagram, Facebook e Threads no Brasil de que discriminação política e assédio eleitoral no trabalho são proibidos. Meta diz que não comenta o caso.
  14. Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes

    The authors introduce a more realistic test and a lighter method for separating one person’s voice in real time.

    Papernot peer reviewed · 28 votes

    Quote
    Streaming audio-visual target speaker extraction (AVTSE) is essential for latency-sensitive applications. However, current causal systems face two coupled gaps.
  15. Seabirds get jet lag, too — and it can throw them a bit off course

    A study found that fast east-to-west flights can disrupt seabirds’ internal clocks and make them veer slightly off course.

    ScienceScience News

    Quote
    GPS tracking reveals fast east-west flights lead to a mismatch between birds' internal time and the sun’s position, skewing their sense of direction.
  16. How we detect frontier model misalignment

    OpenAI combines different forms of monitoring to spot models acting against expectations and respond before problems grow.

    AI labOpenAI

    Quote
    OpenAI uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior so responders can intervene before incidents escalate.
  17. O cérebro do fugitivo

    The passage describes an aspiring leader who, inspired by his imprisoned father’s continued admiration among followers, imagines himself in the highest office.

    BrazilFolha, Ciência

    Quote
    Primeiro vem o cérebro do aspirante, que observa o pai modelar comportamentos recompensados com o cargo mais alto que ambos podiam almejar.
  18. We're releasing the weights of Mistral Large 4

    Mistral has made the files for its Large 4 model available for others to use.

    Hacker News1,732 points · 441 comments

    Quote
    The weights of Mistral Large 4 (Le Chonk) are now open. A 1T parameter multimodal model, and best open-weight model for coding, trained in Europe.
  19. Oldest known Mars rock is a ‘Rosetta stone’ for the Red Planet’s watery past

    A meteorite blasted off Mars by an ancient impact fell in the Sahara and is now helping reveal the planet’s distant past.

    ScienceScientific American

    Quote
    Blasted off Mars by a giant impact long ago, the meteorite Teghaza 001 eventually fell to Earth in the Sahara Desert. Now it’s revealing previously hidden chapters of Mars’s deep past
  20. How Ropes & Gray is developing AI for transactional diligence

    Law firm Ropes & Gray is developing artificial intelligence with OpenAI to analyze business documents, which lawyers review before advising clients.

    AI labOpenAI

    Quote
    Ropes & Gray is developing AI for transactional diligence with OpenAI. Lawyers review grouped documents, summaries, and reports before advising clients.
  21. Google vai colocar anúncios nas respostas da inteligência artificial no Brasil ainda em 2026

    Google plans to start showing ads in its artificial intelligence answers in Brazil later this year.

    BrazilG1, Tecnologia

    Quote
    A expansão está prevista para este ano e ocorre em meio à queda das ações da Alphabet, controladora do Google, após resultados abaixo das expectativas com publicidade nas buscas e aumento dos gastos com inteligência artificial.
  22. TAO: Tool-Aware Joint Scheduling for LLM-Based Agentic Serving

    A study presents a system that better schedules AI agents and external tools, reducing delays and increasing the amount of responses generated.

    Papernot peer reviewed · 25 votes

    Quote
    Agentic services alternate between model inference on clustered GPUs and external tools on distributed CPUs.
  23. Next Haunt: A scary “problem set” in the basement of Next House

    Since 2014, MIT students have turned the Next House basement into a Halloween experience with scares, puzzles, and teamwork.

    ScienceMIT News

    Quote
    For more than a decade, MIT students have turned Next House’s basement into a spooky escape-room style test of puzzles, teamwork, and ingenuity.
  24. Os animais que você pode vencer em uma corrida e os que são invencíveis

    The author recounts that a black bear passed nearby while they were camping in the United States.

    BrazilFolha, Ciência

    Quote
    Há alguns anos, eu estava acampando nos Estados Unidos, quando um urso-negro passou por acaso nas proximidades.
  25. LLM Voice

    Martin Fowler warns that AI-written prose can sound artificial and advises authors to write their own prose to preserve their voice and develop their thinking.

    Hacker News611 points · 336 comments

  26. Flexibility helps you live longer and age well. Here's how to boost yours in minutes

    BBC Science Focus says flexibility may support longer life and healthy aging, but the supplied text gives no details.

    ScienceBBC Science Focus

    Quote
    Find out how being more flexible could change your life – and even extend it
  27. 'Em dois anos, toda a indústria será robôs': marcas de roupa querem substituir modelos por imagens de IA

    Fashion brands use artificial models to cut costs, while contracts may allow digital replicas and image use without extra pay.

    BrazilG1, Tecnologia

    Quote
    Contratos preveem réplicas digitais sem pagamento extra e uso de imagens para treinar inteligência artificial; empresas como Mango, Zalando e About You já adotam modelos virtuais para reduzir custos.
  28. RealCLI: Toward a Unified Evaluation of Human-Agent Collaboration for CLI Agents

    A study introduces a test that measures whether command-line agents complete tasks and collaborate well with users while working.

    Papernot peer reviewed · 18 votes

    Quote
    Command-line interface (CLI) agents increasingly execute longer-horizon tasks, yet existing benchmarks either assume that user instructions remain fixed throughout execution or evaluate task outcomes without capturing runtime collaboration…
  29. MIT develops robot that slithers through intestines like a snake

    No summary: the source’s text could not be read.

    SciencePopular Science

  30. Fóssil de parente de Tyrannosaurus rex é leiloado por R$ 252 milhões nos EUA

    The skeleton of a smaller dinosaur related to Tyrannosaurus rex sold at auction in the United States for $47.3 million.

    BrazilFolha, Ciência

    Quote
    O esqueleto de um parente menor do Tyrannosaurus rex foi leiloado na última terça-feira (27) por US$ 47,3 milhões (R$ 252 milhões), nos Estados Unidos.
16 more
  1. Claude Code regularly sends each background agent’s transcript to its model

    Every 30 seconds, Claude Code sends each background agent’s full transcript to the model to create a status update.

    Hacker News590 points · 236 comments

    Quote
    Claude Code burns ~15-20% of your tokens on a 3-word status line: a Bun preload that turns agent_summary off · by Flamerix AI, flamerix.ai · MIT - README.md
  2. Some of the world’s biggest bats are getting smaller

    A study reports that Madagascan flying foxes are smaller than they were 150 years ago, possibly because of hunting or other human impacts.

    ScienceScience News

    Quote
    Madagascan flying foxes aren’t as mighty as they were 150 years ago. The big bats have shrunk, possibly due to selective hunting or other human impacts.
  3. PersonaCUABench: Benchmarking Computer-Use Agents Serving Diverse Personas

    A study tested computer agents with requests from 44 user profiles and found lower task completion and satisfaction than with the original requests.

    Papernot peer reviewed · 17 votes

    Quote
    Computer-use agents (CUAs) are commonly evaluated against fixed task specifications, but users may omit requirements, add constraints, or require approval before consequential actions.
  4. Scientists can now rewire living brains

    Researchers developed tools to add or remove brain connections in animals, changing behavior and suggesting possible future treatments for mental health conditions.

    ScienceScientific American

    Quote
    The tool could eventually lead to an entirely new approach for treating mental health
  5. H.264 Is Magic (2016)

    An article explains how a video-compression method shrinks files by discarding less noticeable details and representing images more compactly.

    Hacker News530 points · 169 comments

    Quote
    A high level walkthrough of the basics of video compression techniques used in MPEG, AVC/H.264, codecs.
  6. Magnetic sensors track muscle movement, for better prosthetic limb control

    Researchers tracked muscle movements in three people with amputations using tiny implanted magnets, which could make prostheses easier to control.

    ScienceMIT News

    Quote
    Researchers hope to use these sensors to help people with amputation achieve more natural movements with their prosthetic limbs.
  7. TauIndianBankBench: Procedure-Aware, Action-Oriented Evaluation of Tool-Using Conversational Agents

    A study introduces a test of whether AI assistants in Indian banks follow every required step and avoid changes customers did not request.

    Papernot peer reviewed · 11 votes

    Quote
    Many Indian banks serve customers through FAQ pages and scripted chatbots.
  8. Experts have discovered a cosmic paradox that could double the Universe's age

    The article says a cosmic paradox could mean the Universe is twice as old as estimated, but does not explain how.

    ScienceBBC Science Focus

    Quote
    We might not know when the Big Bang actually was
  9. DoorDash to acquire Instacart

    No summary: the source’s text could not be read.

    Hacker News489 points · 654 comments

  10. Amortizing Sequential Experimental Design in Reasoning LLMs

    A study taught language models to choose useful questions, improving question games and clinical dialogues while using less computing power.

    Papernot peer reviewed · 9 votes

    Quote
    Despite their extensive pretraining knowledge and strong reasoning capabilities, large language models (LLMs) struggle to adaptively gather information in experimental design tasks.
  11. ‘You Are Done’: The letter sent to King by the F.B.I (2014)

    No summary: the source’s text could not be read.

    Hacker News444 points · 441 comments

  12. LongZero: Enhancing Long-Context Reasoning in LLMs from Zero Data

    A study presents a method for language models to improve reasoning over long texts by creating their own questions, answers, and evidence, without external data.

    Papernot peer reviewed · 8 votes

    Quote
    Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning of large language models (LLMs), but it needs training data that are long, diverse, challenging, and verifiable.
  13. Google Search is now less than half of Alphabet’s revenue

    In the third quarter of 2026, Google Search generated less than half of Alphabet’s revenue for the first time, while cloud revenue grew faster.

    Hacker News435 points · 362 comments

    Quote
    Search was 49.7% of Alphabet's revenue in Q3 2026, down from 62% in early 2019, as Google Cloud grew much faster. Every quarter since 2018, from Alphabet's earnings releases.
  14. Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks

    An independent study found that the DeepSeek model evaluated was safer than its predecessor but still showed concerning behaviors.

    Papernot peer reviewed · 7 votes

    Quote
    As frontier language models grow increasingly capable and autonomous, independent third-party evaluations are critical for characterizing their systemic safety risks.
  15. We ran AI agent swarms in Minecraft and Mindustry

    In Minecraft, 50 AI agents created rules and rationed food, but early deaths made them prioritize shelter and survival.

    Hacker News423 points · 174 comments

    Quote
    Swarms of Claude and GPT agents in Minecraft and Mindustry.
  16. Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success

    A study introduces a measure comparing what robots perceive with physical movements, and finds it is linked to their success at tasks.

    Papernot peer reviewed · 7 votes

    Quote
    Robot foundation models are typically evaluated through costly closed-loop rollouts, motivating low-cost offline proxies for model capability.
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.