Weekly recap
AI news this week, Oct 26 to Nov 1, 2026
The week in five lines
- Actor Patrick Stewart has died, according to the New York Times headline.
- Mistral released the files for its Large 4 model for others to use.
- Anthropic filed paperwork for an initial public stock offering, which still requires regulatory approval.
- WIRED reported a far-reaching surveillance operation by US homeland security investigators against left-wing activists.
- An expedition off northeastern Brazil recorded 31 marine species not yet described by science.
3,024 papers went into arXiv’s main AI categories this week.
Top Hacker News stories this week
-
Patrick Stewart has died
No summary: the source’s text could not be read.
-
HackerTube: YouTube built from videos recommended by HN
A creator collected videos recommended in Hacker News comments and launched a site organizing thousands of them across 150 topics.
-
Show HN: HackerBooks – HackerTube, but for books mentioned on HN
A website gathers books mentioned in Hacker News comments and ranks them by how many people recommend each one.
Quote
Browse thousands of books mentioned in Hacker News comments, 2007–2026. Ranked by how many different people recommend them.
-
Binance-hacker-1029 – $5.6B stolen from Binance
No summary: the source’s text could not be read.
-
We're releasing the weights of Mistral Large 4
Mistral has made the files for its Large 4 model available for others to use.
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.
-
Anthropic S-1
Anthropic filed paperwork for an initial public stock offering; the shares cannot be sold until regulators approve the filing.
-
Why is this fingerprinting script creating bioweapons?
A researcher found that an anti-bot script hid text designed to trigger AI refusals about biological weapons, though it appeared to collect ordinary signals.
-
Ghostty 1.4
Ghostty 1.4 improves speed and memory use, displays animated images, and adds compatibility updates and improvements for remote access.
Quote
Release notes for Ghostty 1.4.0, released on October 27, 2026.
-
Apple's git still trusts the compromised eMudhra certificates
A site warns that Apple’s macOS developer tools still accept compromised certificates, allowing secure connections to be intercepted.
Quote
Eleven days after Apple was notified of the eMudhra CA compromise, the git, curl and python3 that Apple ships with macOS still trust the compromised certificates.
-
Evidence of a spying operation by ICE against left-wing social movements
WIRED reports that US homeland security investigators are conducting a far-reaching surveillance operation against left-wing activists.
Quote
ICE’s Homeland Security Investigations is engaged in far-reaching spying against left-wing activists that rivals the FBI and CIA crackdowns on the Civil Rights and anti-war movements.
Top AI papers this week
-
PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations
A study presents a system that helps robots predict physical movement and choose actions without generating future videos during control.
Quote
Vision--Language--Action (VLA) models generate robot actions from visual observations and language instructions, but their understanding of physical dynamics remains limited.
-
The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents
A study presents a method that creates simulated worlds to train agents and reuse skills, improving results on familiar and new tasks.
Quote
Multimodal intelligence can be viewed as human-guided co-evolution between worlds and agents through interaction.
-
Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs
A method decides when images are needed while generating answers, reducing processing while retaining nearly all performance in tests.
Quote
Diffusion vision-language models (DVLMs) process their entire input at every denoising step.
-
Work Agents: Systems, Capabilities, and Optimization
A paper proposes a framework for understanding and evaluating digital agents that perform work tasks, and identifies capabilities needed for reliable results.
Quote
As foundation models evolve from information generation to digital work execution, agents are increasingly capable of completing complex tasks across tools, applications, and workflows.
-
Hy-Nomas: Towards Free-form Interactive World Generation
Hy-Nomas generates videos of interactive worlds from instructions and camera movements while retaining context from earlier interactions.
Quote
Interactive world generation requires models to follow successive commands while preserving relevant history and action consequences.
-
StreamVibe: Proactive Video Agents for Evolving User Intent
A study presents a system that follows videos and tries to infer what a person needs, even without explicit requests, adjusting responses as situations change.
Quote
Vision-language models are shifting from offline passive understanding toward online proactive interaction.
-
Proving and Improving the Platonic Representation Hypothesis with JEPAs
A study claims artificial intelligence systems learn more aligned representations with less data when trained on different views of the same scene.
Quote
Whether models trained on different modalities converge to the same representation of the world is central to AI.
-
A-RSI: Adaptive Task Refresh for Zero-Data Multimodal Self-Improvement
A study presents a method that creates new multimodal tasks when learning weakens, raising average performance across eleven tests.
Quote
Recursive self-improvement requires models not only to acquire useful training experience, but also to decide when their current experience has ceased to be useful.
-
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.
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.
-
ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning
A study presents a method that learns robotic movements from human and robot videos, helping robots handle new tasks and environments.
Quote
Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and coverage of action-labeled robot demonstrations.
What the AI labs announced
-
OpenAI appoints Dan Driscoll as Senior Counselor for National Security Modernization
OpenAI appointed Dan Driscoll to advise on responsible AI use in national security, defense against cyberattacks, and cooperation with U.S. allies.
Quote
OpenAI appoints Dan Driscoll to advise on responsible AI for national security modernization, cyber resilience, and collaboration with U.S. allies.
-
Making it easier for websites to express AI training preferences
OpenAI will honor a shared robots.txt signal so websites can tell participating providers their preferences for AI training.
Quote
OpenAI will honor a shared robots.txt signal to help website operators communicate their AI training preferences to participating providers.
-
Fair use benefits everyone
OpenAI shares excerpts from legal arguments saying fair use supports artificial intelligence training and innovation.
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.
-
Credit Saison achieves 954% ROI in a ChatGPT Enterprise pilot
Credit Saison says its ChatGPT Enterprise pilot returned 954% on investment and helped expand its use across the company.
Quote
Credit Saison’s ChatGPT Enterprise pilot delivered 954% ROI. See how leaders and employees helped turn the results into companywide adoption.
-
Retell uses OpenAI models to help power 100M monthly calls
Retell says an OpenAI model led its voice test, while a coding tool helped increase revenue per employee.
Quote
GPT-6 Astra tops Retell’s voice agent benchmark, while Codex helps it reach a $1.6M annual revenue run rate per employee.
-
How Check Point accelerates vulnerability research with Daybreak
Check Point says an OpenAI tool helps analysts turn information about security flaws into customer protections faster.
Quote
With OpenAI Daybreak, Check Point speeds up the research behind new security protections, enabling its analysts to turn vulnerability intelligence into customer defenses faster.
Science
-
The science behind that irritating twitch in your eye
A twitching eyelid can be irritating, but it is usually nothing to worry about.
Quote
Your twitchy eyelid might be irritating, but it's probably nothing to worry about. Here's why
-
Most solar systems look nothing like ours. Here’s why
New discoveries are helping scientists fill gaps in understanding how our Solar System formed and why it differs from many others.
Quote
Although key details are missing from the story of how our Solar System formed, recent discoveries are helping scientists fill in the blanks
-
A New Map of ‘Empty’ Space Challenges a Tranquil Picture of the Galaxy
New three-dimensional maps show that the space between stars is dynamic, although astronomers had treated it as a static backdrop.
Quote
For decades, astronomers treated the interstellar medium as a static backdrop to the goings-on of the universe. Now, a wave of 3D mapping is bringing the interstellar medium’s dynamic nature into focus.
-
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.
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.
-
How Spiders Spin ‘Silk Dope’ Into Nature’s Toughest Material
Biologists are investigating how shapeless proteins turn into strong, flexible silk fibers.
Quote
By probing its molecular structure, biologists are uncovering how a tangle of unstructured proteins turns into a strong, flexible fiber. The post How Spiders Spin ‘Silk Dope’ Into Nature’s Toughest Material first appeared on Quanta Magazine
-
How to change your perception of time (and make life seem longer)
BBC Science Focus asks whether changing how time feels can make life seem longer, but the text provides no explanation.
Quote
Can we feel like we have more time by bending our perception of it?
-
Experts now say you have far more than five senses. Here’s what you never noticed
Neuroscientist Lisa Feldman Barrett explains ways we perceive the world beyond sight, hearing, touch and smell.
Quote
Neuroscientist Dr Lisa Feldman Barrett discusses the different ways we’re able to perceive the world
-
The grossest thing you touch every day, according to science
You've likely already touched it about 47 times today.
Tech
-
Guarding the Gates of Knowledge
A test found publishers can keep pace with fake scientific papers generated by artificial intelligence, but more advanced tools make them harder to detect.
Quote
A recent scroll through social media got me thinking about trust.
-
When Electric Cars Were for Women
In the early 20th century, electric cars were marketed to women because they were easy to start and produced no exhaust.
Quote
At the turn of the last century, the automobile industry was still new and evolving rapidly. In 1898, there were about 50 carmakers in the United States. A decade later, that number had increased fivefold.
-
Multiphysics Modeling for Next-Generation Electromagnetic Manufacturing Technologies
A webinar will present simulations combining electromagnetic fields and heat to measure temperatures in metals and make ceramic parts using microwaves.
Quote
Advances in sensing, manufacturing, computational resources, and data-driven methods are enabling new approaches to electromagnetic sensing, non-destructive evaluation, and advanced materials processing.
-
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.
Quote
This article is brought to you by Qualcomm Technologies. For many consumers, much of the promise and appeal of AI is personal.
-
Systems Thinking Is Vital in the Age of AI
Artificial intelligence can speed up project work, but engineers must coordinate specialists, check results, and take responsibility for final decisions.
Quote
Modern engineering projects increasingly combine human judgment with machine capability.
-
Your Robot’s Safety Functions Already Work. What If the Input Lies?
Tests show that manipulated text, images, sounds, or messages can change robots’ behavior, and researchers urge teams to assess the resulting safety risks.
Quote
This article is brought to you by VicOne. A mobile robot slows for someone in a hallway. A collaborative arm eases its movement as a worker approaches. A humanoid pauses to let a person pass.
-
IEEE Xplore Digital Library Adds AI Assistant
IEEE launched a research assistant that answers questions using peer-reviewed scientific papers and includes links to check its sources.
Quote
IEEE recently released the IEEE Xplore AI Research Suite, built on the full text of millions of peer-reviewed publications in the IEEE Xplore Digital Library.
-
The October inflection point for vibe game development
Simon Willison says he made four enjoyable games with little guidance from AI and sees October as a turning point for creating games through conversation.
Quote
I've written before about how November 2025 was an inflection point for coding agents - the point where Opus 4.5 and GPT-5.1 crossed a tipping point that meant their respective coding agents became reliable daily drivers for software…
In Brazil
-
Por que uma desavença entre matemáticos alimenta temores de que a IA prejudica a ciência
Freezers in a former London clock factory store tubes that together contain 100 billion genes.
Quote
Em uma antiga fábrica de relógios no centro de Londres, uma fileira de freezers brancos mantém armazenadas dezenas de pequenas caixas com tubos de vidro identificados por códigos de barras. Juntos, esses tubos contêm 100 bilhões de genes.
-
Natureza profunda
An expedition off northeastern Brazil recorded 31 marine species that have not yet been described by science.
Quote
Expedição no litoral do Nordeste encontrou 31 espécies que podem ser novas. Veja as imagens
-
Duas gerações
Two awards recognized 49 recent doctoral graduates for their research and three social scientists for their academic work.
Quote
Capes e Anpocs premiam quem pensa o Brasil que vem pela frente. Veja ainda eventos e livros
-
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.
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
-
Como mapear a opinião de elites se elas não respondem a pesquisas?
Researchers combined scheduled meetings with hallway approaches to interview 170 lawmakers and compare their views on democracy with those of the public.
Quote
Cientistas desenvolvem estratégias para chegar a grupos pouco acessíveis e reduzir os efeitos da não resposta em estudos sobre elites políticas e sociais
-
O poder da matemática a serviço da indústria
CeMEAI brought universities and companies together in Brazil, applying mathematics to industrial problems and supporting research on proteins.
Quote
Centro de referência em matemática aplicada ajudou a aproximar academia e indústria no Brasil
New models on OpenRouter
-
Particle AI: Particle-1
Particle AI created an artificial intelligence model for coding, trained from DeepSeek V4.1 Flash and designed for software tasks.
Quote
Particle-1 is a token-efficient agentic coding model from Particle AI, post-trained on DeepSeek V4.1 Flash. It is suited for software engineering work such as terminal workflows, codebase changes, and data...
-
OpenAI: GPT-6.1 Terra Pro
On OpenRouter, a setting for the model aims to improve answers to complex tasks, but costs much more.
Quote
GPT-6.1 Terra Pro is the same underlying model as GPT-6.1 Terra, served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. **Cost note:** pro mode spends far more...
-
OpenAI: GPT-6.1 Terra
The description presents OpenAI’s Terra as cheaper than Sol and suited to everyday tasks, writing, coding, and understanding screenshots.
Quote
GPT-6.1 Terra is the mid-tier model in OpenAI's GPT-6.1 series, priced below GPT-6.1 Sol. It is suited for everyday agentic, coding, and writing work, is particularly strong at understanding screenshots...
Saved
Kept only in this browser.
Nothing saved yet. Tap Save on any link to read it later.
How this is made
A script builds this recap from the week’s editions. An AI writes the five lines and the summaries. Nobody reviews them before they go up.
