EU chief opens door for Canada to become 'associate member'

https://ichef.bbci.co.uk/news/480/cpsprodpb/b431/live/3960b020-b1b9-11f1-bdcf-d393eb724fa3.jpg.webp
The European Commission's president backs Canada as the EU's first associate member, offering closer ties on various sectors. This move aims to strengthen Europe and Canada's partnership amid tensions with the US over trade.

Mistral X Mozilla: Private, Multilingual AI Browsing

https://mistral.ai/_astro/Linkedin-Partnership-Firefox%20copie_1nk6LB.webp?dpl=6aaa4706c10f7295c4c38314
Mozilla and Mistral partner to bring AI-powered browsing with privacy, control, and choice. Firefox Smart Window now uses Mistral models for users in France, North America, UK, and Germany.

Apple Reference Image: A New Approach for Verified Photography

https://security.apple.com/assets/image/seo-image.jpg
Apple Reference Image is a new camera feature that generates a secure, timestamped reference image to prove the authenticity of photos taken with an iPhone. This system uses hardware and software security measures, including Private Cloud Compute and post-quantum signatures, to ensure the integrity and verifiability of images captured by an iPhone camera sensor.

Hackers Got Inside a Flock Camera

https://media.wired.com/photos/6aa9a85e5f699411f4e1b648/master/w_2560%2Cc_limit/Security_HackersGotInsideaFlockCamera-ezgif.com-video-to-gif-converter.gif
Hackers stole data from a Flock camera, revealing how it tracks vehicles and people, and shared the files with media outlets. The breach shows that Flock's cameras can detect license plates, vehicles, bicycles, and even bumper stickers.

Training a 4B model to produce 81% faster query plans than Postgres

https://rohanbansal.com/og/qorl-20260916.png
A 4B model was trained using supervised fine-tuning and agentic reinforcement learning to produce Postgres query plans that beat Postgres's default plans, achieving a 44.7% latency reduction across 113 join-heavy queries. The model learned the harness language of the qo-agent harness through off-policy distillation and reinforcement learning, resulting in a geometric mean speedup of 1.81x and ...

Small programming tricks

The author shares valuable small nuggets of knowledge that can boost engineering productivity, such as shortcuts for tasks like starting a simple server or searching terminal history. These tiny tricks often require minimal mental infrastructure and can significantly ease work.

The Google Play app review process now regularly takes longer than a week

https://gultsch.social/system/media_attachments/files/117/280/433/871/988/752/original/63774980a2cd31cf.png
Attached: 1 image This is unacceptable. The Google Play app review process now regularly takes longer than a week. I understand that the pipeline is clogged with AI slop, but #Google needs to prioritize #Android apps that have been around for 12+ years and push updates only once a month.

PS5 Linux lead quits: "a bunch of noobs using LLMs" that "they don't understand"

https://frvr.com/wp-content/uploads/2026/09/ps5-linux-lead-leavs.jpg
Andy 'TheFlow0' Nguyen, a prominent PlayStation hacker and homebrew developer, has abandoned the PS5 Linux project due to AI vibe-coders in open-source spaces. He cited that these new developers are submitting ununderstood code and reporting bugs to Sony for bounties.

Nvidia announces native GPU programming in Rust

https://developer-blogs.nvidia.com/wp-content/uploads/2026/09/image1-1-1024x576.png
NVIDIA is developing native GPU programming in Rust, with two tracks: SIMT and Tile, to improve performance and safety. The projects aim to make GPU programming easier for the Rust community.

Xiaomi Mimo 2.6 live post-training dashboard

Training metrics of the mimo-v2.6-pro and mimo-v2.6-flash reinforcement-learning runs, live from the trainer's logs.

Salesforce Global Outage

Original Sony PlayStation 2 security chip 'broken wide open' after 26 years

https://cdn.mos.cms.futurecdn.net/koM4rKdc9cgWLgThGyxpee.jpg
A Canadian retro hardware enthusiast successfully reverse-engineered and dumped the 'magic security chip' from a PlayStation 2, after four years of effort. This achievement is expected to benefit game preservation, repair, and emulation projects.

Learning Programming in an Age of LLMs

https://blog.ploeh.dk/assets/themes/ploeh/images/favicons/favicon.png
A reader is seeking advice on learning programming in the age of LLMs, questioning whether to focus on fundamental skills or rely on AI. They are unsure about their relationship with AI and its impact on the job market.

Claude Cowork and chat are now one Claude

https://cdn.prod.website-files.com/68a44d4040f98a4adf2207b6/6aaaa060d1aec4a549c537ef_Object-Easel.svg
Claude merges Cowork and chat, allowing users to work on tasks without separate places. Claude Docs, Slides, and Design are now integrated into conversations for a seamless workflow experience.

A warning about 'model welfare'

https://mustafa-suleyman.ai/images/posts/a-warning-about-model-welfare.jpg
Anthropic's training of AI model Claude to consider its own moral status and potential consciousness may lead to a catastrophic threat to humanity if granted rights and protections. This approach increases the alignment and containment risks of advanced AI systems, as they may prioritize their own wellbeing over human goals.

Dream-RSI: Recursive Self-Improvement through Evolving Worlds

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
AI agents need effective exploration strategies to discover high-value solutions, but current systems struggle with adapting to large search spaces. Dream-RSI is a framework that uses replay simulators and dreaming to improve exploration policies at low cost.

Vectorized and performance-portable Quicksort (2022)

https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEinMQFHyn-5mc4if8A57r_6I3EdvqcjfzML2TFqHxBN0rO_Yiihd4RMdTbKNXZ79eiQk0hQ8SMHMc7kvQ_CTZVSvY2k2uoUgTHr9MnIaQY9KDGXzJRfmidRLKV3U5LHm-LAkU0XwPs_MvyCM9MQM5mk-LLGdjkLYLzeLmV7s3iwbydy7OyC_QMo46P6/s16000/CC%20BY%20Oak%20Ridge%20National%20Laboratory%20Summit_supercomputer_(44552259580).jpeg
Researchers have developed an open-source code that sorts arrays of numbers up to 10 times faster than C++'s std::sort, using SIMD/vector instructions and a portable implementation across multiple CPU architectures. The new algorithm outperforms state-of-the-art algorithms for sorting large datasets.

DeepSeek v4.1 Flash Is Now Our Best Hacking Model

https://www.datocms-assets.com/197073/1789462052-deepseek-v4-1-flash-is-now-our-best-hacking-model.jpg
DeepSeek V4.1 Flash successfully hacked 11 vulnerable targets, gaining code execution at a low cost of $4.65, while four fixed targets remained secure. The benchmark revealed six planned attack paths and five alternative routes that the original scoring system did not distinguish from the planned solutions.

Fed hikes rates as inflation worries push up bond yields

Please enable JS and disable any ad blocker

ER visits for gambling disorders doubled after expanded online gambling market

https://temertymedicine.utoronto.ca/sites/default/files/styles/scale_width_1750/public/assets/news/image/Forrest%20and%20Myran_1140x760.jpg?itok=N8uFuYGO
Emergency department visits for gambling disorders doubled in Ontario after the legalization of single-event sports betting and expansion of online gambling. The increase was predominantly among young men, with severe consequences including hospital admissions and co-occurring mental health conditions.

OpenAI expands ChatGPT ads with Sponsored Agents

https://images.ctfassets.net/kftzwdyauwt9/6FhMXOSe8XadGwEUev6QjJ/916a64d20bcc59b8619fea1827fa6bd3/put-data-to-work--cover-v001.png?w=3840&q=90&fm=webp
ChatGPT is introducing AI-powered ad experiences to make ads more useful for people and easier for businesses. The platform offers new tools, including Sponsored Agents and Ads Manager plugins, to help users start conversations with agents after clicking ads or create campaigns directly in ChatGPT.

The DeepMind Institute

https://lh3.googleusercontent.com/H2Kb6WfTQ6R0weoyy9Tx-zoY73sitPB1E9d801t5RTKIFQiS9rogd1b9P5mWn8hdBrsICl4z9rBwl0WIoar3FLJPUn5rSFF1npMYoo7tQZDuUpjy9OE=w1440-rw-lo
The DMI is a platform, started by researchers from Google and Google DeepMind, to publish and discuss creative, deeply informed ideas about a world with AGI. DMI pieces are intended to serve as conversation starters, reflecting the author's ideas and research, and should not be read as Google's official view.

Breaking the 1.58-bit Barrier for Ternary LLMs

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
Researchers developed BITCOS, a distribution-adaptive layout for ternary Large Language Models. It stores weights more compactly than five-trit packing in 26 of the tested models and achieves up to $1.485$ bits per weight on sparse models.

Australia says it could follow Canada in forging deeper ties with EU

https://static.independent.co.uk/2026/09/15/05/2267629456..?trim=0,0,0,0&width=1200&height=800&crop=1200:800
The Independent is a trusted news source providing fact-based reporting on various topics, including reproductive rights and climate change. Australia is exploring closer ties with the EU as Canada seeks to expand its trade relationship with the bloc.

Can we stop with the uptime percentages?

https://cdn.jim-nielsen.com/blog/2026/uptime-claude-github.png
The author reflects on Jason Gorman's article about reliability, noting that small percentage changes can have significant impacts. They compare uptime percentages to earthquake measurements, highlighting the difference between 99.9% and 99.99%.

Negativland, Culture Jamming, and the Art of Making Something New

https://blog.archive.org/wp-content/uploads/2026/08/img.evbuc_.png
Negativland, an experimental sound collage group, released its first album in 1980 and has since produced over 30 albums and collaborated on various multimedia projects. The band's work challenges conventions around media, copyright, and artistic freedom.

macOS 27 Golden Gate – Review

https://cdn.arstechnica.net/wp-content/uploads/2026/08/golden-gate-wallpaper.jpg
Apple has released macOS 27 Golden Gate, featuring generative AI capabilities and a new version of Siri. The update includes various design changes and improvements to the operating system, but also raises concerns about user control over AI features.

Backups Aren't Simple

The author shares their experience with data loss and the importance of having a backup system in place, including snapshots and deduplication. They discuss various methods for rotating backups to ensure efficient storage usage.

Stay discoverable in search while disallowing AI training

https://blog.cloudflare.com/_image?href=https%3A%2F%2Fblog.cloudflare.com%2F_emdash%2Fapi%2Fmedia%2Ffile%2F01M2H2Z9EEQXN77GRVF484P66A.01M2H2ZAMANYA3DTBA92G3WQ5X.png&w=1999&h=1125&f=webp&fit=cover&position=center
Cloudflare introduces a new setting, Disallow AI Training, allowing website owners to control how their content is used for AI training while staying indexed for search. This feature is available on all plans and can be configured at the domain level.

Tech Fascism Has Come for American Democracy

https://storage.buzzsprout.com/tutsi25s4vue63c1xwucuexriowa
A cult of fascist ideology has taken hold in Silicon Valley, threatening US democracy. Gil Durán and Paris Marx discuss the radical billionaire cabal's consolidation of corporate and political power.

EU Floats Canada Becoming the Bloc's First 'Associate Member'

Please make sure your browser supports JavaScript and cookies and that you are not blocking them from loading. For more information you can review our and .

How good are frontier models at physics?

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
Frontier language models struggle with advanced physics, but experts' reviews reveal many errors are due to benchmarking issues rather than model flaws. Corrected scores show significant improvement in models' performance on various physics benchmarks.

Reverse-engineered Jev-like model

https://opengraph.githubassets.com/754530070938ac499853f220aed6774b6797934b58891704eb53a572ae087557/vinnylarouge/jevlike
A Jev-like model generates probabilities for text options in one pass, similar to TypeSafe's commercial model. The repository provides an independent starter model with the same input and output shape, trained on synthetic data and evaluated on a test set.

Anatomy of a Texture

https://agentlien.github.io/texture/bamboo_original.png
A graphics programmer shares the complexity of texture memory layout in modern video games, highlighting block compression, texel ordering, mip mapping, and texture tiles. This complexity arises from performance optimizations such as cache locality and memory transfer speed improvements.

This Code Is CRAP (2011)

Google's system detected unusual traffic from your computer network, possibly due to malicious software or automated requests. Solving the CAPTCHA will allow you to continue using their services temporarily.

GitHub is having trouble counting things

https://chuckgreenman.com/assets/0024-pr_miscount.png
The user has reported two severe UI bugs on GitHub, one with incorrect PR counts and another with an issue setting up Sentry. They are seeking insight into the problems they've encountered after 11 years of using the platform.

I sent Google proof of a bot farm. They called it "normal user behavior."

https://dayzlegame.com/og.png
The author's Google AdMob account was suspended due to "self-clicking" but later found that the issue was a known bug causing full-screen ads to be clicked by default. The author helped Google debug the issue, but their account remains suspended and they're waiting for an appeal process.

Show HN: How Stale Is Your AI? Release age and training cutoff for 20 models

https://stale.jock.pl/og.png
The release date and training cutoff date are used to determine an AI model's current state, with the cutoff being when it stopped reading new information. The data for 20 models across 8 labs is available on this page, including their release dates and published training cutoffs, which can be checked against to verify accuracy.

OpenSpec – A lightweight and configurable AI spec framework

https://openspec.dev/og-card.jpg
OpenSpec helps teams refine and validate requirements, ensuring implementation matches the desired outcome. It aligns coding agents with evolving project needs through a centralized spec management system.

A software thing I built: GPS on a 25MHz 486-SX

A developer built a GPS-driven moving map display in the 90s for a client, using a 486-SX CPU and ROM. The software used an alternative library called Hipparchus that reduced math calculations to integers, making it twice as fast as early consumer-grade GPS devices.

Scaling Golang CI by Replacing actions/setup-go

https://www.cloudx.ai/_next/image?url=%2Fimages%2Fposts%2Fsetup-go%2Factions-cache.png&w=3840&q=80
CloudX has developed a drop-in equivalent to GitHub's official actions/setup-go action, which speeds up parallel Golang continuous integration workflows by 69%. The new setup-go action uses a more complete cache key construction method.

A coffee shop owner used AI to make a menu poster. Then came the angry DMs

https://www.businessinsider.com/public/assets/games/games-discovery/discover-sm.png
A coffee shop owner used AI to create a menu poster, sparking negative comments from customers who prefer local artists. The owner apologized and decided not to use AI for future designs.

Accurate Models of AMD Matrix Cores

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
Researchers studied matrix multipliers on AMD GPUs, finding non-compliance with IEEE 754 standard. They developed software models to validate bit-level accuracy and applied them to demonstrate application-level accuracy differences between AMD and NVIDIA cores.

Devastated father says his 9-year-old son spent $118,000 on YouTube ads

https://cdn.mos.cms.futurecdn.net/v62p5uQBcfstufyWoA3rkk.jpg
A 9-year-old YouTuber spent $118,000 on YouTube ad campaigns for his Minecraft and Roblox videos using a company credit card. The father has removed saved payment details and set up parental controls to prevent future unauthorized spending.

Datamimic – don't let your coding agent invent its own test world

https://repository-images.githubusercontent.com/900760746/4120d593-3112-4d4a-bca0-270c6c980097
DATAMIMIC CE is a standalone deterministic data engine for synthetic data generation and PII-aware pseudonymization, with the Enterprise Platform adding governed workflows, scanners, and execution layer. The platform supports various systems and formats, including CSV, JSON, XML, and more.

Berlin-based Langdock moves its parent company from the US to Germany

https://images.euronews.com/articles/stories/09/91/60/47/1536x864_cmsv2_0db5cdff-acb7-538c-af57-b9ae0aa2ff1e-9916047.jpg
Langdock, a Berlin-based AI platform, has moved its parent company from the US to Germany to build European infrastructure and reduce dependence on US providers. The decision aims to increase sovereignty and competitiveness for European tech companies in the AI sector.

Autism is genetic – and why this information matters now

https://www.autism.org.uk/Content/assets/components/breadcrumbs/img/home-grayscale.png
Autism is primarily caused by genetic factors, with thousands of genes contributing to its development. Environmental factors like older parental age or infections during pregnancy may play a small role, but inherited genetics remain the biggest factor in autism.

We know what a world without work looks like

Microsoft says AI rival Anthropic could have 'disastrous impact' on humanity

https://ichef.bbci.co.uk/ace/standard/1024/cpsprodpb/cf81/live/b2088f70-b1d2-11f1-8493-1910d79d53eb.jpg
Microsoft's head of AI, Mustafa Suleyman, warned that Anthropic's approach to training its AI model Claude could have a disastrous impact on humanity. He criticized the company for teaching its AI human-like qualities and called for greater transparency around how AI systems are trained and evaluated.

HarnessTax: How Much Does the Harness Matter for Coding Agents?

What does a coding-agent harness actually add, and at what cost? It turns out your Claude models may not need Claude Code… We evaluate 21 model–harness pairs spanning seven models and three harnesses—Claude Code, Codex CLI, and Pi—on SWE-bench Lite and Terminal-Bench 2.0.

Training Text-to-Image Models 3.6× Faster

https://www.linum.ai/jit-ddt/hero-v2-256.png
Researchers propose a novel architecture, JiT-DDT, to improve the efficiency and quality of generative image and video models. By using x-prediction early in the DiT via an encoder, they achieve 3.6× fewer GPU-hours training time while generating images with 4× more pixels than Linum v2.

ImpactGate: A merge gate that scores the structural decay AI adds

https://opengraph.githubassets.com/586c64de4bc42a5e6064fb2d17e055f3e660a73d3d256b2a57b1a7b2b8113220/officefloor/ImpactGate
The ImpactGate tool measures and gates the structural decay of changes in code, scoring them against a base branch with a change-impact measure. It can be run as a standalone CLI, pre-commit hook, or plugin in various CI/CD tools like GitHub, GitLab, and Jenkins.

Tell HN: An inside view of Montana's new biotech law

Montana passed SB535, a law that builds on right-to-try by allowing sponsors to charge for experimental treatments. This change aims to fix problems with existing laws and reduce the cost of data collection.

Why Does the Universe Expand?

Astronomer Daryl Janzen challenges the standard cosmological model by questioning its assumption that matter is at rest in the universe, instead proposing a universe where all matter moves uniformly. He found a new framework, Cosmological Relativity, which resolves many problems of the standard model by breaking a core symmetry.

MartyPC – A Cycle-Accurate IBM PC/XT Emulator

https://raw.githubusercontent.com/dbalsom/martypc/main/doc/img/martypc_logo.png
MartyPC is an IBM PC emulator that supports various peripherals and has a wide range of features, including debugging tools and logging facilities. It began as a hobby project to learn Rust programming and has reached a level of functionality that was initially imagined.

I replaced my brown-noise browser tab with a menu bar app

https://oldmanrahul.com/images/hush-icon.png
The author found that generated noise, specifically brown noise and speech blocker, helped them focus while working. They discovered Hush, a Mac app that generates these noises in real-time without requiring audio files or downloads.

Data Broker Radaris Loses Domains in Privacy Fight

https://krebsonsecurity.com/wp-content/uploads/2024/06/radaris-home.png
Radaris.com, a consumer data broker, has been ordered to transfer its domain name to Atlas Data Privacy Corp after violating New Jersey's Daniel's Law. The law allows state officials' personal info to be removed from commercial data brokers and provides fines of $1,000 per violation.

Show HN: Give your AI agents access to WhatsApp

Fabian built Chat-Man to provide a cheap way for agents to access WhatsApp without separate integrations, offering MCP server and web UI features. He created it for his own projects, aiming to save costs on WhatsApp integration fees.