1-Click GitHub Token Stealing via a VSCode Bug

200 ammar2 2026-06-03 02:37:41 UTC blog.ammaraskar.com Llama Share 26

An attacker can steal a GitHub token by exploiting a vulnerability in VSCode's webviews, allowing them to install malicious extensions and gain full code execution. The attack involves creating a Jupyter notebook with a payload that installs a local workspace extension, which can then be used to steal the token.

Use your Nvidia GPU's VRAM as swap space on Linux

216 tanelpoder 2026-06-02 23:37:41 UTC github.com/c0dejedi Llama Share 64

A daemon allocates VRAM as a block device using the NBD protocol, exposing it as a swap device to the kernel. This approach sidesteps NVIDIA driver limitations and provides a swap solution with lower latency than NVMe.

The Unreasonable Redundancy of Nature's Protein Folds

26 ray__ 2026-06-03 04:37:41 UTC research.ligo.bio Llama Share 2

Researchers improved biomolecular modeling by scaling models, compute, and data, but found that natural protein sequences are vast and their folds are more redundant than sequence counts suggest. They developed a scalable clustering algorithm to identify reusable structural neighborhoods and found that natural proteins reuse a relatively small set of fold solutions, suggesting that enzyme ...

MAI-Code-1-Flash

432 EvanZhouDev 2026-06-02 19:37:41 UTC microsoft.ai Llama Share 187

Microsoft introduces MAI-Code-1-Flash, a new coding model for fast and efficient assistance in developer workflows. It outperforms Claude Haiku 4.5 with better price to performance and is now rolling out to GitHub Copilot users in VS Code.

CT scans of BYD car parts

307 viasfo 2026-06-02 20:37:41 UTC lumafield.com Full Share 132

We CT scanned four BYD components—a battery cell, window switch, EV charger, and key fob—to see what's inside the world's best-selling EV that's banned in the U.S.

Agentic Mfw

112 elmerland 2026-06-03 02:37:41 UTC agenticmotherfucking.website Llama Share 32

The author mocks the current state of software development where maintainability and clean code are no longer valued, instead prioritizing complexity and speed to impress investors. The author also ridicules the idea of open-source contributions and the current state of GitHub, where bots and AI-generated code dominate and human contributions are often ignored or drowned out.

Roku LT Operating System open source distribution

57 dpmdpm 2026-06-03 01:37:41 UTC blog.roku.com Llama Share 14

Roku LT OS is a lightweight, open-source operating system for embedded systems and automotive engineering. It provides a powerful, predictable framework for high-performance hardware-level development.

AI outperforms law professors in Stanford Law study

159 berlianta 2026-06-03 00:37:41 UTC law.stanford.edu Llama Share 132

A Stanford study found law professors prefer AI-generated answers to student questions over human-written ones, with AI winning 75% of head-to-head matchups. The study suggests AI can be a valuable tool in legal education, offering high-quality, on-demand support that complements classroom instruction.

Capstone – multi-platform, multi-architecture disassembly framework

28 gregsadetsky 2026-06-03 02:37:41 UTC capstone-engine.org Llama Share 0

Capstone is a disassembly engine for binary analysis and reversing, supporting 24 architectures and implemented in pure C with bindings for various languages. It provides a generic API and is widely used in the security community.

Words of Type

23 tobr 2026-06-03 03:37:41 UTC wiki.wordsoftype.com Llama Share 2

Alphabets are writing systems using letters to represent sounds, forming syllables, words, and sentences. Typefaces have various parts, such as ascenders, descenders, and apertures, which contribute to their overall design and legibility.

Gmail thinks I'm stupid, so I left

763 speckx 2026-06-02 19:37:41 UTC moddedbear.com Llama Share 448

The user is frustrated with Gmail's new AI features that nag and interrupt them while composing emails. They're switching to Fastmail after 16 years with Gmail due to the unwanted features.

My thoughts after using Clojure for about a month

173 speckx 2026-06-02 20:37:41 UTC acdw.net Llama Share 89

The user is learning Clojure and finds it ergonomic and powerful, preferring it over Common Lisp and Scheme due to its cohesive design and large standard library. They appreciate Clojure's simplicity and uniformity, but acknowledge some pain points, such as its syntax and the need to learn Java for interop.

Pluto.jl 1.0 release – reactive notebook for Julia

66 fons-p 2026-06-02 23:37:41 UTC discourse.julialang.org Llama Share 5

Pluto version 1.0 is released, celebrating six years of progress and making scientific computing more accessible and fun. Pluto is an interactive environment for notebook programming in Julia, with features like reproducibility, accessibility, and reactivity.

Open Repair Data Standard – Open Repair Alliance

113 cassepipe 2026-06-02 19:37:41 UTC openrepair.org Llama Share 3

The Open Repair Data Standard (ORDS) helps combine repair data on electronics by defining a shared approach. It collects data on product, repair, and session information to identify trends globally and locally.

4K years ago, Mohenjo-daro grew more equal over time

71 marojejian 2026-06-02 22:37:41 UTC archaeologymag.com Llama Share 36

Archaeologists found lower inequality levels in ancient Mohenjo-daro, with economic gaps shrinking over time. The city's focus on public infrastructure and shared resources contributed to its stability and equality.

HP re-releases classic computer science calculator: The HP-16C

146 dm319 2026-06-02 19:37:41 UTC hpcalcs.com Llama Share 90

The HP 16c Collector's Edition retains the classic layout but is 100x faster with customizable word size and base conversions. It features programming capabilities with conditional branching, subroutines, and flags, and supports up to 203 bytes of program memory.

How we index images for RAG

107 mooreds 2026-06-02 18:37:41 UTC kapa.ai Llama Share 14

Kapa's AI assistant uses a vision model at indexing time to describe images as text, then retrieves these descriptions alongside text chunks at query time, improving answer quality and reducing costs. This approach, which involves describing images once at ingestion and storing them as separate text chunks, outperforms query-time vision and is more cost-effective for large-scale applications.

Loading Sega Games Off a Vinyl Record [video]

28 zdw 2026-06-03 01:37:41 UTC youtube.com Embed Share 5

OpenFOV – Webcam head tracking for iRacing

107 mwit2023 2026-06-02 14:37:41 UTC openfov.com Full Share 51

OpenFOV uses your webcam to control iRacing's in-game FOV. Unlocks VR-style functionality for your monitor!

Trump signs downsized AI order after weeks of reversals

196 _alternator_ 2026-06-02 17:37:41 UTC politico.com Llama Share 141

President Donald Trump signed an executive order addressing AI-driven cyber threats with a voluntary review process for AI companies. The order aims to enhance national security and global AI dominance while avoiding heavy federal oversight.

Multicore suppport for DOS is real – partly

78 beebix 2026-06-02 19:37:41 UTC vogons.org Full Share 13

Sorry, you are not authorized to view this page.

Show HN: Paseo – Beautiful open-source coding agent interface

50 timhigins 2026-06-02 22:37:41 UTC github.com/getpaseo Llama Share 26

Paseo is a self-hosted coding platform that allows running agents in parallel on your machine with full dev environment. It supports multi-provider models, voice control, cross-device access, and local server management.

Preparing for KDE Plasma's Last X11-Supported Release

171 jandeboevrie 2026-06-02 14:37:41 UTC blog.davidedmundson.co.uk Llama Share 198

Plasma is transitioning to Wayland, removing X11 support in Plasma 6.8. This change will allow for new performance improvements and features, but may require custom scripts and workflows to adapt.

The American Missile Crisis

51 JumpCrisscross 2026-06-03 03:37:41 UTC research.contrary.com Llama Share 47

US munitions stockpiles have declined significantly since the Cuban Missile Crisis, with supplies potentially depleted in as few as three days in a Pacific conflict. The bottleneck in solid-rocket motor production is attributed to ammonium perchlorate, a highly regulated and difficult-to-produce fuel.

Fidonet: Technology, Use, Tools, and History (1993)

156 BruceEel 2026-06-02 14:37:41 UTC fidonet.org Llama Share 64

FidoNet is a point-to-point email network that uses modems to connect over 20,000 nodes worldwide, allowing users to send and receive email and news. The network was developed in 1984 and has a unique addressing system, with gateways to the Internet and other networks, and a strong focus on user autonomy and minimal technical support.

SaySynth: A Brief History of Speaking Machines

5 evakhoury 2026-06-03 04:37:41 UTC brian.abelson.live Prefix Share 0

These are expanded notes from a talk I gave at composition.codes on December 21, 2025. Slides here. Video here. SaySynth is a synthesizer I built on top of macOS’s text-to-speech framework — more popularly known as the say command. But to explain why I built it and why I think it matters, I want to take a detour through the history of speaking machines more broadly. Mechanical — Literally ...

QBE – Compiler Backend – 1.3

89 birdculture 2026-06-02 18:37:41 UTC c9x.me Llama Share 29

QBE 1.3 is a significant release with 7k new lines of code and 1.5k deleted ones, featuring a new IL matching algorithm and optimizations. It achieves 63% of commercial compiler performance on coremark and 33% improvement on the Hare test suite.

Bringing Up DeepSeek-V4-Flash on AMD MI300X

96 kkm 2026-06-02 18:37:41 UTC fergusfinn.com Llama Share 11

Doubleword is building an inference cloud using AMD's MI300X, a high-end AI accelerator. The MI300X is underappreciated due to software issues, but it offers 192GB of HBM3 and comparable FP8 compute to NVIDIA's H100 at a lower price.

The advertising cartel coming to your web browser

174 speckx 2026-06-02 20:37:41 UTC blog.zgp.org Llama Share 46

Big Tech companies are proposing a built-in advertising system, Attribution Level 1, which prioritizes their own tracking over user privacy. This system could lead to more money for Big Tech and less for legitimate sites, while also enabling riskier tracking practices.

Stop Ruining It

261 herbertl 2026-06-02 10:37:41 UTC seths.blog Llama Share 122

Paul McGowan's insight is that quality is what remains when flaws are removed. He suggests that customer delight, curiosity, satisfaction, and trust are all preserved if not ruined by external factors.