Postman silently auto-installs AI plugins and attempts to share all API data

https://github.githubassets.com/assets/platform-4ce65ae0bf273ce2.png
User found Postman plugin and startup hook causing issues, including authentication warnings and routing API work through Codex skills. They request a rollback of the auto-installed integrations to prevent similar problems in the future.

The 'Astra can read and write weights to improve Qwen' post is misleading

https://honnibal.dev/_next/image?url=%2Fprofile.png&w=640&q=75
The author explains that the "Astra can read and write weights" post is misleading, as it doesn't actually demonstrate a single technique to analyze or modify network weights. The experiments involve different tasks and techniques, including writing code snippets to analyze or modify weights.

Google will turn off search for schools that reject AI

Google will allow K-12 students to use AI tools unless school administrators opt out, but restricting AI may also block access to Google search. Schools must choose between allowing or blocking AI for their students.

Qwen 3.8 27B on my laptop can interpret model weights, too

Researchers Rikiya Takehi and colleagues demonstrated that large language models can interpret and modify neural network weights, a capability dubbed Weight-Space Programming. The model successfully identified game behaviors from raw weights without executing code or using tools.

Show HN: Hacker Times, a New York Times-inspired version of Hacker News

https://rikiyat.github.io/images/blog/weight-space-programming.png
A team at MIT CSAIL reports early experiments with GPT-6 Astra, a language model that can modify neural-network weights without writing a training loop. HackerTube gathered 6,454 YouTube videos mentioned by Hacker News readers over nearly two decades.

Could a modern chess engine on a Pentium have defeated Kasparov in 1996?

https://nicholas.carlini.com/writing/2026/chess/ppro-screen.png
A computer program called Deep Blue defeated chess champion Garry Kasparov in 1996, but a modern chess engine running on the same hardware could beat him. A custom-built engine using Stockfish and adapted to run on 1996 hardware achieved an estimated Elo rating of 3300.

Kalshi and its co-founders face felony charges for violating Utah gambling laws

https://cdn.arstechnica.net/wp-content/uploads/2026/09/kalshi-640x427.jpg
The US government is suing Utah over its attempt to ban Kalshi, a prediction market that offers sports bets. The CFTC claims Utah's efforts obstruct Congress's decision to federally preempt state law regarding prediction markets.

OpenAI seems to be silently re-enabling model training on Codex chats

Sorry

Stringman, an $800 open-source room scale gripper robot

https://storage.googleapis.com/nf-site-assets-prod-b/assets/exploded.webp
Stringman is a robotic arm designed to navigate and manipulate objects in a room, using cable-driven architecture and overhead lines for movement. It offers an open-source platform with customizable control software and AI models, allowing users to create their own workflows and behaviors.

DiscImageStudio: burn visible artwork onto CDs, DVDs and Blu-ray discs

https://raw.githubusercontent.com/JiaFeiMiao-K-Cat/DiscImageStudio/main/docs/images/case1.jpg
Disc Image Studio 是一款 Windows 桌面版和 Linux 和 macOS 的实验性命令行版,可用于生成 CD、DVD 和蓝光镜像,并可将图像转换成符合这些介质物理层约束的数据。

Mac at Any Scale

https://namespace.so/_next/image?url=%2F_next%2Fstatic%2Fmedia%2Fnichtverstehen.1bc76f95.jpg&w=1080&q=75
The author shares their experience building custom Mac racks for a datacenter, overcoming challenges such as limited supply of Mac Minis and designing efficient cooling systems. They also discuss the use of Apple's Virtualization framework to run macOS instances in the datacenter.

Safari introduces Adaptive Dark Mode in 27.3

https://webkit.org/wp-content/uploads/Demo-of-Adaptive-Dark-Mode_light.png
Safari 27.3 beta introduces Adaptive Dark Mode, which automatically converts light websites to dark mode for a consistent user experience. The feature is designed to work with most web pages and can be enabled by setting system appearance to dark in Safari settings.

Nvidia – open-source auto-ontology

https://raw.githubusercontent.com/NVIDIA/auto-ontology/main/docs/assets/arch.png
Auto Ontology is an open library that builds a context layer for agents to access and analyze data from enterprise metadata, reconciling definitions across systems. It provides a semantic catalog/ontology layer over relational sources, stored in Postgres alongside the catalog it describes.

Study estimates 75% of NFL players who died over 15-year span had CTE

https://media-cldnry.s-nbcnews.com/image/upload/t_fit-560w,f_avif,q_auto:best/rockcms/2026-10/261027-football-CTE-helmet-sjr-1644-2-104adc.jpg
A new study found that 75% of professional football players who died over a 15-year span had signs of chronic traumatic encephalopathy (CTE), the highest estimate to date. The condition can cause memory loss, depression, and violent mood swings, but diagnosis is only possible during autopsy.

Intel stock collapse worsens as $3.5T chip rout spreads

https://s.yimg.com/cv/apiv2/finance/ai/ai_loading_loop_v3.gif
Coinbase is a platform for exploring stocks, and Yahoo Finance provides information on market trends. The tech sector experienced significant losses last week, with chip stocks leading the decline.

US tech giants' hidden AI debt swells 90% to $3.2T in 6 months

https://images.ft.com/v3/image/raw/https%3A%2F%2Fcms-image-bucket-productionv3-ap-northeast-1-a7d2.s3.ap-northeast-1.amazonaws.com%2Fimages%2F9%2F1%2F3%2F4%2F13394319-2-eng-GB%2Faad3b104a912-photo_SXM2026103100004660.jpg?fit=cover&gravity=faces&dpr=2&quality=medium&source=nar-cms&format=auto&width=780
PALO ALTO, California -- U.S. tech giants' off-balance-sheet debt from artificial intelligence-related obligations continues to balloon, almost doubling in six months to about $3.2 trillion, with investors increasingly alarmed by seemingly unending spending.

The Little Book of Generative AI Foundations: An Intuitive Mathematical Primer

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
This book provides a compact introduction to the mathematical foundations of generative AI, connecting major families of models. It aims to make generative modelling more accessible while retaining its mathematical substance for researchers and practitioners.

Net worth for people aged 30–33 in the U.S.

https://www.keepingup.com/product/dashboard-demo.jpg
The median net worth for people aged 30-33 in the US is $68,000. The top 10% of this group has a net worth of around $440,000.

Revolut moves to self-hosted GLM

https://pbs.twimg.com/media/HWJxC2yWEAAWYYO?format=webp&name=large
Revolut moved 2,000 engineers to self-hosted GLM for coding and improved engineering lead time by 70%. The move allowed them to allocate resources more efficiently without rationing tokens.

Just a little AI resistance, as a treat

https://shkspr.mobi/blog/wp-content/uploads/2017/11/Confused-Robot.jpg
The author complained to Ocado about their AI-generated customer service response, which they found unhelpful and lacking in human touch. The company responded personally, apologizing for the frustration and promising to pass on the feedback to improve their service.

Preparing for the Flipper One Kickstarter and opening pre-orders

https://flipper.net/cdn/shop/files/page_flipper-one_body.png?v=1793720883&width=3840
Flipper One is a Linux-based, expandable computer with Wi-Fi 6E and Gigabit Ethernet. It features a powerful ARM processor and supports various M.2 modules for customization.

Ask HN: Does anyone have a contact with Cleveland-area Ham club?

Sorry
Durations are estimated until the YouTube metadata is fetched, so some programs will run short ("please stand by") or get cut off. Keys: ↑↓ channel · 0-9 tune · G guide · F picture · M mute · C captions · I info · T theater · double-click screen for fullscreen · P power

Gitbook Breached

Sorry

I dumped Apple’s word list

https://helgesver.re/_next/image?url=%2F_next%2Fstatic%2Fmedia%2Fvulgar-word-usage-tweet.289c366e.webp&w=1200&q=75
The author of the article discovered a private database on their Mac called VulgarWordUsage.db that tracks profanity usage per app and recipient, influencing autocorrect suggestions. The database is populated by Apple's system dictionary flags, which categorize words into tiers (vulgar, offensive, sensitive) and topics.

uBlock Origin filter to block autoplay videos in TechCrunch

TechCrunch articles have an autoplaying video that can't be hidden. Blocking scripts with uBlock Origin's MV2 version stops the video from loading and removes the empty box left behind.

AI is destroying my CS degree

https://bearblog.dev/static/og-image.png
A computer science student is struggling with the increasing use of AI in their studies, citing inconsistent university policies and excessive workload. They long to learn without feeling pressured by commercial interests and outdated headlines about the field's obsolescence.

Charles Tandy, the mastermind behind Radio Shack

https://i0.wp.com/dfarq.homeip.net/wp-content/uploads/2026/03/CharlesDavidTandy.jpg?resize=216%2C300&ssl=1
Charles David Tandy was a successful entrepreneur who expanded three retail chains: Tandy Leather, Radio Shack, and Pier 1 Imports. He built his business empire through strategic acquisitions and innovative marketing strategies.

macOS 27 does not support CoreStorage volumes

macOS 27 has broken support for encrypted HFS+ volumes, making it impossible to mount or convert them. Users are forced to delete and rebuild their data from backups due to the lack of functionality.

Bear turns 10, and it still feels like day one

https://blog.bear.app/wp-content/uploads/2026/10/default-blog_1920x1080-868x488.png
Bear, a note-taking app created by three friends in 2013, is celebrating its 10th anniversary with new features and design principles. The app prioritizes simplicity, naturalness, and safety to provide a comfortable space for users to create notes.

Quili AI

https://www.quili.ai/images/hero-w-v1.jpg
Quilicura, Chile hosts many AI data centers using its water supply, but a severe drought is expected until 2040 due to climate change. Residents took action by becoming "human A.I." for a day to raise awareness about the impact of irresponsible AI use on local resources.

The Internals of PostgreSQL

https://www.interdb.jp/pg/title3.png
This documentation explains PostgreSQL's subsystems and their features, aiming to provide a comprehensive understanding of the database system. The author is seeking remote positions focused on applying ML and AI technologies to DBMS due to personal reasons of moving to Japan.

Practice speaking Japanese with real YouTube videos, graded line by line

https://i.ytimg.com/vi/4VxU2BhGm6U/hqdefault.jpg
A Japanese YouTube video was added, and JIVX turned its lines into sentences to be typed or said. The video ranges from 1 to 180 minutes long and includes various topics such as figure skating, kindness, introductions, travel guides, and morning routines.

Google researchers demand freedom to tell the world what they know about AI

How I use AI (Oct 2026)

https://ben.stolovitz.com/_astro/cover-dark.DYeplbFh_1x6zaR.png
The author, a research engineer at Salesforce AI Research, reflects on their experience with large language models (LLMs) and artificial intelligence (AI) tools over the past year. They have seen significant changes in how they use AI, from managing multiple agents to drafting a personal AI policy.

Anthropic IPO filing: $1.7B quarterly payroll, about $107K per employee monthly

https://www.techspot.com/images2/news/bigimage/2026/11/2026-11-03-image-7.jpg
Anthropic's personnel costs averaged over $100,000 per employee per month. The company spent more on stock-based compensation than its revenue in 2023 and reported a significant increase in expenses for restricted stock units after going public.

Cloudflare experiencing multiple active incidents

We're not currently aware of any issues affecting Cloudflare services. We'll post updates here as soon as anything changes.

The AI Boom May Be Sowing Seeds of Another Financial Crisis

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 .

Show HN: HackerChurn – which SaaS tools HN wants to cancel

https://img.logo.dev/lastpass.com?token=pk_HbQGX2MMSuC45xo3YqC9eQ&size=128&format=png&fallback=404
A sample of public Hacker News comments from the last 3 years was analyzed to count complaints about tracked software, with tools requiring 3 complaints in 2 threads to be listed. The results were sourced from a public search API and nightly updates.

Show HN: I built a little browser game to practice Vim commands

https://hjkl.duy.run/og.png
Pick the skills you want, then make small real edits in a browser Vim editor. Each skill runs two guided edits, two prompted edits, and two independent edits. Any valid edit that reaches the target text passes. Type :help in Normal mode for the current commands.

No Vehicles in the Park

The game asks players to determine if a hypothetical park's rule "No vehicles in the park" has been violated, disregarding external rules and personal beliefs. Players answer 20 questions about the park to reach the end of the game.

Tiny Basic for Arduino

https://dilshan.github.io/tiny-basic/logo/tinybasic.png
Tiny Basic for Arduino is a lightweight BASIC interpreter designed for mid-range MCU boards, allowing direct code writing and running on the hardware. It features an extensive standard library with built-in functions for hardware I/O, math, and bit manipulation.

The AI Bubble Monitor

There is a growing concern that the AI bubble may be about to burst, with many experts warning of a potential collapse in the industry. The current valuation of AI-related companies is extremely high, and investors are increasingly questioning whether these companies will be able to meet their enormous lease commitments. If the bubble bursts, it could have significant consequences for the ...

Show HN: Kairos – Open-source risk and fraud decisioning platform

https://raw.githubusercontent.com/finic-ai/kairos/main/docs/images/readme/kairos-experiments.gif
Kairos is an open-source alternative to Sardine, Sift, Riskified and SEON for real-time fraud decisions with decision models catching fraud before labels are available. It combines Grayson's unstructured data analysis with traditional machine learning models like logistic regression and gradient-boosted trees.

Introduction to Programming in ATS

ATS is a syntax-rich and feature-rich programming language that covers core features like functional programming, types, and theorem-proving. The book assumes no prior programming experience but may be dense for those without considerable coding background.

Handles are the better pointers (2018)

The author discusses a memory management approach using 'index-handles' to prevent common memory-related problems in C and C++. This approach involves moving memory management into centralized systems, reducing the use of pointers, and ensuring that only system-private array base pointers are exposed.

Octomind – open-source AI harness that saves your tokens, static rust

https://raw.githubusercontent.com/Muvon/octomind/master/assets/octomind-animation.gif
Octomind is a CLI-first AI coding agent runtime that gives an LLM real access to your repo, allowing it to read and write files, run shells, and more. It's built for autonomous work and can be used in various ways, including interactive sessions, piped commands, daemon mode, and scheduled runs.

Open-Source 5G Ran Platforms: A Dual Perspective on Performance and Capabilities

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
This paper compares OpenAirInterface (OAI) and srsRAN, two open-source 5G RAN implementations, to evaluate their performance in various scenarios. The study assesses supported features, deployment flexibility and radio resource control setup procedures for different applications like VoD, LS and CG.

USDB: Scaling Storage for Next-Generation Recommender Systems at ByteDance

https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png
USDB is a unified sequence storage system that optimizes streaming and batch workloads by reusing maintenance results across ingestion, inference, and training. It achieves significant performance improvements with 85.8% lower mean write latency and 42.1% lower infrastructure cost compared to an Optimized RocksDB baseline.

Google Will Turn Off Search for Schools That Reject A.I

Google will allow K-12 students to use AI tools unless school administrators opt out, but restricting AI may also block access to Google search. Schools must choose between allowing or blocking AI for their students.

Microsoft's Canada timezone troubles leave admins fixing calendars

https://image.theregister.com/261387.jpg?imageId=261387&panox=0.00&panoy=0.00&panow=100.00&panoh=100.00&heightx=0.00&heighty=0.00&heightw=100.00&heighth=100.00&width=960&height=432&format=webp&format=jpg
Microsoft is experiencing issues with its software after Canada changed clocks, causing problems with Outlook and other products. The company is working on fixes but some users are facing incorrect times and workarounds until then.

Show HN: Interactive notes on RLHF, attention, LoRA and LLM evals

Interactive explainers for computer science and engineering concepts, one per page.

Pratt parsing is a black box

Pratt parsing is an efficient algorithm that reduces parsing to a simple question of what to expect to the left of a token, allowing for fast and flexible parsing without complex dependencies or backtracking. To effectively use Pratt parsing, one must adopt a mindset shift towards abstracting grammatical structures into nuds (null denotation) and leds (left denotation), which enables powerful ...

Sam Altman's Case Against Your RAM Is Efficient. That's the Problem

Sam Altman discussed the importance of efficient use of memory, suggesting that having RAM in data centers is more efficient than local inference. However, he overlooked the broader implications of centralizing control over tools and systems on user empowerment and autonomy.

Notice How Old websites look better than new websites

https://yingster.net/images/personal-pictures.png
Michael Caldwell is a tech enthusiast with interests in design, science fiction, and woodworking. His personal webpage showcases his past projects and fond memories, including his beloved cat Yingster.

No brother of mine will play the flute

https://substackcdn.com/image/fetch/$s_!AkYa!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa658ffcd-5a3a-4c9a-b3e2-957a2ea2ac39_1338x1022.png
The author discusses the importance of respecting individual preferences and boundaries, particularly when it comes to touch and social interactions. They argue that forced intervention can exacerbate problems rather than solve them.

Anthropic's IPO Filing Warns It May Be 'Imminently' Close to RSI

Anthropic's S-1A filing reveals the company may soon meet its internal definition of recursive self-improvement (RSI), a point at which AI systems can autonomously improve their own training. The filing notes that while current models have not yet crossed this threshold, Anthropic is considering updates to its Responsible Scaling Policy to ensure human control over frontier model development.

Show HN: Rustable is an open source database client built in Rust

https://rustable.app/assets/table-view.png
Rustable is a database explorer with native, GPU-rendered UI and support for multiple databases including MySQL, MariaDB, PostgreSQL, SQLite, and Redis. It offers features like saved queries and edits before saving to reduce work on the UI thread.

How to write comparison content that doesn't age like milk

https://res.cloudinary.com/dmukukwp6/image/upload/comparison_content_by_year_a684b63280.png
The author shares their experience writing comparison content, highlighting the importance of being fair and accurate to competitors while showcasing a product's strengths. They emphasize that honesty is key to building trust with readers.

Show HN: Bema: Work messaging built for humans

Bema is a non-instant messaging platform that alerts users to urgent messages and allows them to schedule check-ins for other messages. The system is experiencing high memory usage and degrading response times, prompting immediate attention.

Show HN: ReBirth 338 in Browser

Run the original ReBirth RB-338 Windows release in your browser with Wine and WebAssembly.

There is no posttraining data commons

https://openathena.ai/og.jpg
Researchers are struggling to create a digital "task commons" for posttraining AI models, as the raw data needed is often ephemeral and difficult to capture. A pretraining commons was largely created by accident through human activity on the internet.

Jailbreaking my car's navigation system just so I can change the time

https://blog.nns.ee/img/2026-11-02-renault-hack/ui.jpg
The user's 2014 Renault Latitude car has a TomTom Carminat LIVE navigation system that requires a legitimate SD card with specific data to function, including the time settings. The user successfully exploited a kernel vulnerability using the LOG command and core_pattern string manipulation to execute arbitrary code on the rootfs.

How McDonald's first got its start

https://th-thumbnailer.cdn-si-edu.com/uIHPTOoa8iPl_6C2hyRUYjRQ_P0=/1026x684/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer/83/66/83668b53-6c9f-4db6-845e-a13d083acf6e/12417851_636538253151507_5379487058622609205_n.jpg
The story begins with the McDonald brothers, Dick and Mac, who sold their theater in 1937 and started a food stand called "Airdrome" selling oranges and hot dogs. They later opened another stand in San Bernardino, which became successful due to its location near Route 66.

Prompting Is Programming

A professor struggled with an AI tool, unable to articulate their goal due to its lack of understanding of programming concepts. The author believes that the key to effective use of AI lies in developing a programmer's mindset, including specifying clear goals and breaking down tasks into manageable pieces.

Ten harness patterns that you can apply to a real repo

https://deepsense.ai/wp-content/smush-webp/2026/10/Harness-cover-1024x614.png.webp
Harness Engineering is about designing better systems around coding agents, providing context and feedback to improve their performance. This involves creating a knowledge base, loading context just-in-time, and closing the feedback loop to prevent agent mistakes.