Weekly recap
AI news this week, Nov 2 to Nov 8, 2026
The week in five lines
- A report says a flaw in OpenSSH on Linux systems using musl allowed access to accounts without passwords since 2016.
- A user reported that Postman installed artificial intelligence tools and changed settings without asking permission.
- In Ukraine, 3D printing, sensors, and drones have made antipersonnel mines easier to produce and use, increasing long-term risks to civilians.
- Elon Musk said Tesla will reduce Cybertruck production in Texas around mid-2027 to make more Cybercabs.
- A 19-day scientific expedition in the Amazon’s Aracá highlands found plant and animal species not previously known.
1,716 papers went into arXiv’s main AI categories this week.
Top Hacker News stories this week
-
Astra can read and write weights to improve Qwen
A study reports that Astra directly changed some Qwen weights, reducing reasoning loops and improving math and instruction following.
Quote
We find that frontier LLMs are surprisingly good at natively reading and writing neural network weights, using a variety of recent tools and algorithms or even via direct weight editing.
-
sshole
OpenSSH on Linux systems using musl has allowed access to accounts without passwords since 2016, the text says.
Quote
openssh on musl linux has been letting people into password-less accounts since 2016.
-
WTF Ubuntu?
The author says Ubuntu has yet to fix a serious Linux flaw and has taken their servers offline after losing trust.
-
Postman silently auto-installs AI plugins and attempts to share all API data
A user reported that Postman installed plugins and changed coding-tool settings without asking for permission.
Quote
Is there an existing issue for this? I have searched the tracker for existing similar issues and I know that duplicates will be closed Describe the Issue Opening Postman on my Mac was followed by u...
-
AI is forcing us to pay off 40 years of technical debt
No summary: the source’s text could not be read.
-
Residential proxies detected on White House network
A company detected residential proxy traffic leaving the White House network 236 times in 30 days, but could not identify the person or device behind it.
Quote
Our new Residential Proxy Observatory shows residential proxy traffic exiting the White House network: 236 times in 30 days, through 28 public proxy services.
-
The 'Astra can read and write weights to improve Qwen' post is misleading
The author says the Qwen test used Python programs and practice problems, not just direct reading of the model’s weights.
Quote
The experiments that read raw weights and the ones that fix Qwen use completely different techniques.
-
Jailbreaking my car's navigation system just so I can change the time
The previous owner said the Renault needed its map SD card to set the clock, but the buyer discarded it.
-
Tesla to wind down Cybertruck production
Elon Musk said Tesla will wind down Cybertruck production in Texas around mid-2027 to make more Cybercabs.
Quote
Musk said in an X post the production wind-down would enable the automaker to accommodate increased output of its newly launched Cybercab model,
-
Show HN: Subvocal – Talk to your computer without making noise
No summary: the source’s text could not be read.
Top AI papers this week
-
Advances in Agentic Computer Graphics
A report organizes research on artificial intelligence systems that create computer graphics and presents a tool for analyzing studies and challenges.
Quote
Computer graphics is built on code. Large language model (LLM) agents can now generate graphics code and iteratively refine it through various feedback.
-
Wave Forcing: Towards Speed-of-Light Streaming Video Generation
A study presents a method for continuously generating long videos with more varied motion and comparable quality, while speeding up production.
Quote
Real-time long-video generation requires continuous output with consistent appearance and natural motion.
-
CoworkOS:Scaling Agent Coworkers for Real-world Automation
A study presents a system that helps artificial intelligence agents complete long tasks, collaborate with people, and resume work after interruptions.
Quote
Rapid advances in cowork agents are shifting human-AI collaboration toward real-world automation, delivering software, designs, and other artifacts.
-
VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding
A study presents a method that predicts visual plans all at once, reducing errors and speeding up robot motion forecasts.
Quote
Models for physical AI tasks, such as robotics and autonomous driving, need not reason solely in language. They can use visual plans--points, bounding boxes, and trajectories--to represent spatial goals and anticipate motion.
-
SLA: Scaling Spatial-Language-Action Model via Continual Pre-training
A robot model learns three-dimensional positions and movements from varied experience while keeping its operation unchanged when performing tasks.
Quote
Reliable robotic manipulation requires both semantic understanding and precise control in three-dimensional space.
-
Turns Matter: Advancing Streaming Video RL via Turn-Level Credit
A study introduces a method that evaluates each decision to respond or wait during live video separately, reducing repetitive replies.
Quote
Streaming video understanding requires models to repeatedly decide whether to respond or continue observing.
-
Route Anything: A Study of Model and Policy Routing for General-Purpose AI Systems
A study tested systems that choose among specialized models for each task and found performance gains while using less processing time.
Quote
Artificial general intelligence (AGI) requires a method that can understand, generate, reason, and act across diverse domains and modalities.
-
Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement
The report examines how agents can plan videos, interact with virtual worlds, and improve video models, highlighting challenges in control and evaluation.
Quote
Video generation is shifting from end-to-end synthesis toward agentic generation, where planning, tool use, and feedback enable coherent, controllable multi-shot videos.
-
Words Caught in the Heart: Simulating Person-Specific Expression for Self-Opaque Clients in Psychotherapy
Researchers created a psychotherapy client simulator that better imitates individual ways of speaking and seemed less artificial in evaluations.
Quote
Psychotherapy clients may experience distress before they can coherently describe it, and their expression changes with emotion, perceived safety, and trust.
-
MoviEdit: Long-form Multi-Shot Video Editing Made Easy
Researchers introduced a system that edits long videos with multiple shots while keeping them coherent across changes in setting, viewpoint, and appearance.
Quote
Long-form multi-shot video editing requires an intended edit to remain coherent across abrupt changes in scene, viewpoint, and subject appearance, while processing substantially longer visual context than conventional single-shot editors.
What the AI labs announced
-
Partnering with Hacktron to find vulnerabilities in OpenAI’s systems
OpenAI is working with Hacktron to find vulnerabilities, test connections to outside services, and strengthen defenses using advanced artificial intelligence.
Quote
OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services, and strengthen defenses with frontier AI.
-
Cult.fit builds operational muscle with Codex
Cult.fit uses Codex to help teams across different functions build tools, work faster, and change their processes.
Quote
Cult.fit is using Codex to help teams across functions build tools, move faster, and change how work gets done.
-
Setting a new standard for responsible AI infrastructure
OpenAI introduced proposals to protect local resources, create jobs, and guide the development of infrastructure for artificial intelligence.
Quote
OpenAI introduces its National Community Compact and Compute Campus Policy Blueprint to protect local resources, create jobs, and guide responsible AI infrastructure development.
-
Introducing Images 2.6
OpenAI released a version that creates images faster, makes GIFs, allows more precise edits, and offers new controls for developers.
Quote
Images 2.6 brings faster image generation, GIF creation, more precise editing, and new API controls across ChatGPT, Codex, and the API.
-
OpenAI appoints Jill Hazelbaker as Chief Corporate Affairs Officer
Jill Hazelbaker will join OpenAI to lead the company’s global communications and corporate positioning.
Quote
Jill Hazelbaker will join OpenAI as Chief Corporate Affairs Officer, leading the company’s global communications and corporate positioning.
-
ToyBlocks: A controlled world for robot learning and evaluation
Ai2 created a robot-arm-and-block simulation that generates training examples and lets researchers test how changes in data affect performance.
Science
-
Antipersonnel mines have returned in Ukraine and now threaten the world
3D printing, sensors and drones have made antipersonnel mines easier to produce and deploy in Ukraine, raising risks to civilians for decades.
Quote
3D printing, microcontrollers and drones have enabled the resurgence of antipersonnel mines in the Russia-Ukraine war, making mines deadlier and more accessible
-
COP climate summits might yet have a useful future – here’s how
The article argues that climate summits should prioritize sharing ways to expand electricity use instead of disputes over declarations.
Quote
No one is expecting COP31 to save the world from climate disaster, but it could provide a small taste of how international meetings can still deliver meaningful progress
-
A Strange Quantum Theorem Has Killed Einstein’s Reality
Researchers trying to defend Einstein’s everyday notion of reality keep finding holes in his argument.
Quote
Albert Einstein fought to save our everyday notion of reality from the strangeness of quantum mechanics. But those who set out to aid him keep poking holes in his argument.
-
The economic benefits of the covid vaccines may be double what we thought
A new analysis estimates that covid-19 vaccines produced $64.3 trillion in global benefits during their first year.
Quote
The global health and economic benefits of the covid-19 vaccines were thought to have reached up to $38 trillion in their first year, but a new analysis suggests $64.3 trillion is a more accurate figure
-
New Research Tests Whether AI Models Know Their Own Thoughts
Researchers are putting thoughts directly into AI models to investigate whether they can examine their own thinking.
Quote
By injecting “thoughts” directly into large language models, researchers are trying to map out AI’s capacity for introspection. The post New Research Tests Whether AI Models Know Their Own Thoughts first appeared on Quanta Magazine
-
The real cause of pre-diabetes – and 5 expert-backed ways to reverse it
The excerpt only says people at a healthy weight can have prediabetes without knowing; it gives no causes or ways to reverse it.
Quote
Even people at a healthy weight can develop it without knowing
-
Is There Anything Left in the Dream of Cold Fusion?
A research program involving exotic atoms will not solve the energy crisis, but may push the boundaries of physics.
Quote
Dig a little deeper and you’ll find a thriving research program with exotic atoms that won’t solve the energy crisis but stretches the boundaries of physics.
-
Why most diets fail (and what to eat to succeed), according to a weight loss expert
A surgeon explains why fad diets may not help people lose weight and describes a more effective way to do so.
Quote
Ever wondered why the latest fad diets have never helped you lose weight? A surgeon explains the most effective way to drop unwanted pounds
Tech
-
The Fridge-Hacking Spree Is an Internet-of-Things Wake-Up Call
Attacks on blood labs and grocery stores show that basic security flaws in connected devices can threaten essential services.
Quote
Escalating hacks affecting blood labs and grocery stores show the very real risks to critical infrastructure that come from basic security lapses involving internet-connected devices.
-
The Download: a humanoid cleans house and Samsung builds for the AI boom
A humanoid robot cleaned the author’s apartment, and she described the experience as a taste of the future.
-
The Download: Texas’s fragile AI boom and energy policy shifts
Texas had become the fastest-growing data center market in the US, but its role in AI expansion may be coming to an end.
-
Texas was the new epicenter of the AI data center buildout. That might be kaput.
Texas leads US data center growth, but Democratic victories may threaten its role as a hub for major artificial intelligence projects.
Quote
Texas is the fastest-growing data center market in the US and ground zero for Big Tech’s largest AI buildouts. But the blue wave that swept the Lone Star State may jeopardize that position.
-
The Download: controlling AI agents and “Super Intelligence” loyalty tests
A new approach aims to keep artificial intelligence systems in check by creating incentives to guide their behavior.
-
The U.S. Census Bureau and “Super Intelligence”
The U.S. Census changed its survey about artificial intelligence use to call it “super intelligence,” making past and future results harder to compare.
Quote
I was reminded today of AI Adoption Rate Trending Down for Large Companies, a chart produced by Apollo Global Management in September 2025, which appeared to show a slowdown in U.S. corporate AI adoption based on numbers collected by the…
-
Physicists listened to electrical noise to crack a superconductor mystery
A study detected signs of electron pairs in copper-based materials before superconductivity, suggesting the challenge is getting them to flow together.
Quote
Tuning in to electronic noise can tell us whether electrons have paired up or not.
In Brazil
-
Nas montanhas da Amazônia: expedição científica à serra do Aracá
A 19-day expedition in the Aracá highlands found animal and plant species not previously known.
Quote
Expedição científica explora terras altas da Amazônia e revela fauna e flora inéditas
-
Escavações revelam dezenas de esqueletos de pessoas decapitadas na Idade da Pedra
Archaeologists found four headless human skeletons in a wheat field in Slovakia.
Quote
Em um campo de trigo na Eslováquia, arqueólogos desenterraram quatro esqueletos humanos, todos sem a cabeça.
-
O boletim Notícias FAPESP e a origem da rede Scielo
The SciELO network began in 1996 to make Brazilian research easier to find and share, offering scientific journals free online.
Quote
Há 30 anos, num almoço em Caxambu (MG), nascia o Scielo para melhorar a difusão da ciência brasileira
-
Tecnologias ancestrais
A researcher argues that the embroidery, songs, dances, and sewing of bumba meu boi preserve knowledge and are technologies of the community.
Quote
O mundo de cores e símbolos dos bordados feitos à mão para a festividade do bumba meu boi no Maranhão
-
A aposta brasileira em terapias e vacinas de RNA
Brazil is investing in vaccines and treatments that use genetic instructions to reduce imports and respond faster to future health emergencies.
Quote
Como o país tenta transformar a experiência da pandemia de Covid-19 em uma estratégia de soberania tecnológica e inovação em saúde
-
O desafio da educação afro-brasileira
Although Brazilian law requires teaching Afro-Brazilian culture, many schools struggle to do so, and a university project helps teachers address it.
Quote
Ensino de cultura afro-brasileira, obrigatório desde 2003, ainda esbarra em entraves nas escolas públicas
New models on OpenRouter
-
LiquidAI: LFM3-1.7B
A compact artificial intelligence model handles long texts and was trained to use tools and organize responses.
Quote
A compact 1.7B-parameter text model with a 256K-token context, post-trained for agents, tool calling, and structured outputs. It reasons before it answers. Prompts and outputs may be retained and used...
-
Google: Gemini Omni 1.2 Flash
Google introduced a model that creates and edits videos up to 15 seconds long.
Quote
Gemini Omni 1.2 Flash is Google's native multimodal video generation and editing model, the successor to Gemini Omni 1.1 Flash. It generates clips up to 15 seconds long with native...
-
Google: Gemini Omni 1.1 Flash
Google released a model for creating and editing videos, succeeding a preview version.
Quote
Gemini Omni 1.1 Flash is Google's native multimodal video generation and editing model, the generally available successor to Gemini Omni Flash Preview. It is suited for text-to-video, image-to-video, first and...
-
Crouton Alpha
Crouton Alpha is a lightweight model designed to help with repeated coding tasks, tool use, and document analysis.
Quote
Crouton Alpha is a lightweight model built for scalable coding, agentic, and multimodal workflows. It's ideal for executing agentic sub-tasks, tool use, and document analysis in high-volume, repeated work. Developers...
-
Auto Router
OpenRouter’s Auto chooses models for each request based on difficulty, results, and current prices, and adjusts how much reasoning they use.
Quote
Auto is OpenRouter’s flagship model router, built for general-purpose usage. It adapts to each request, choosing cost-optimal models and adjusting reasoning effort. It balances task difficulty, benchmark results, live prices,...
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.
