Harvard study predicts most suicide attempts a week in advance

23 jodacola 2026-09-09 02:08:57 UTC current.fas.harvard.edu Llama Share 3

Researcher Matthew K. Nock developed a new method to predict suicide attempts, using real-time surveys and data analysis, with 87% accuracy in predicting events before they occurred. The study's findings aim to improve the clinical-care system by detecting fleeting suicidal thoughts and providing timely interventions to keep people safe.

Muse – Meta’s personal AI agent

333 yks 2026-09-08 20:25:57 UTC ai.meta.com Full Share 334

Sorry, something went wrong. We're working on getting this fixed as soon as we can.

Large language models develop novel social biases through adaptive exploration

101 paimapi 2026-09-08 22:25:57 UTC openreview.net Full Share 49

Complete the check below to continue to OpenReview

How to build a printer

155 cat-whisperer 2026-09-08 21:25:57 UTC nishantjosh.dev Llama Share 37

The author built a custom firmware for the Xteink e-ink display to enable wireless printing from their MacBook using the Internet Printing Protocol. They successfully implemented this feature, allowing them to print documents directly onto the device's screen.

Navier-Stokes – Tristan Buckmaster [pdf]

1298 procedurecall 2026-09-08 06:25:57 UTC cims.nyu.edu Llama Share 560

A mathematician collaborated with a Large Language Model (LLM) to solve the Navier-Stokes problem, but OpenAI claims to have solved it first. The LLM was used to push the collaboration's work from rough forcing to smooth forcing and incompressible Euler equations.

AlphaGenome Atlas: a high-resolution map of human DNA

498 utiiiD 2026-09-08 15:25:57 UTC blog.google Llama Share 115

Scientists have developed the AlphaGenome Atlas, a database predicting effects of single nucleotide variants in the human genome. The Atlas helps researchers quickly prioritize promising research areas by providing an easy-to-use score for coding and non-coding regions.

DaVinci Resolve 21.1

359 tosh 2026-09-08 14:25:57 UTC blackmagicdesign.com Full Share 157

Media

On the Navier–Stokes Millennium Prize Problem

1138 tedsanders 2026-09-08 17:25:57 UTC openai.com Llama Share 980

Researchers at OpenAI have developed a proof that the Navier-Stokes equations for fluid motion can develop a singularity in finite time, resolving one of the Millennium Prize Problems. The solution involves an internal model significantly more capable than GPT-6 Astra and demonstrates the potential for AI to advance research and technology benefiting humanity.

Tao: Open math problems being non-renewably mined by AI

163 _alternator_ 2026-09-08 21:25:57 UTC mathstodon.xyz Llama Share 107

Mathematicians are collecting open problems in a non-renewable way, potentially leading to scarcity. Most generated problems are trivial or unproductive, while few hold the potential for breakthroughs and connections.

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

218 stared 2026-09-08 15:25:57 UTC quesma.com Llama Share 109

The author tested various GPU RAM requirements for running Qwen3.8 27B without sacrificing quality, finding that a 24 GB card like RTX 4090 can handle the full BF16 model with some context reduction.

The Microeconomics of Artificial Intelligence (2025)

32 neehao 2026-09-08 22:25:57 UTC direct.mit.edu Full Share 11

I-have-ADHD: A skill to stop coding agents from burying the answer

326 domhudson 2026-09-08 14:25:57 UTC github.com/ayghri Llama Share 256

User wants to update jsonwebtoken API in auth.ts and run tests, also reviews claude plugin versions.

27.5KB language-agnostic WebGPU syntax highlighter

9 bpierre 2026-09-09 01:25:57 UTC gpu-lexer.vercel.app LocalLLM Share 2

gpu-lexer uses a WebGPU model to label code parts without relying on language-specific grammars, achieving 87.43% agreement with Shiki on held-out files. It supports multiple languages and mixed-code regions, with results varying based on corpus size and language support.

Kimi K3 (2.8T) at 1 token/s on a MacBook Pro, streamed from four SSDs

214 Argonautlabs 2026-09-08 20:25:57 UTC github.com/argonautlabsai Llama Share 113

Deltafin is a fork of gavamedia/deltafin, running Kimi K3 from SSDs on Apple Silicon with the ARGODRIVE storage work. It aims to squeeze out efficiency from large models like K3 without reducing quality.

A Topological Picture Book, Rendered

57 mathgenius 2026-09-08 23:25:57 UTC e-infinity.space Llama Share 7

This text discusses various techniques for creating non-photorealistic renderings of 3D models, including contour drawing, hatching, and line rendering. Researchers have developed methods to stylize silhouettes and outlines using algorithms such as chaining, visibility, and principal curvature directions.

Getting phpBB 1.4.4 working in Docker

25 HeckFeck 2026-09-08 23:25:57 UTC thran.uk Llama Share 12

The PHPBB 1.4.4 script can be run on modern systems with patches, including changes to MySQL and PHP settings. The installation requires config.php to be writable during setup, which is set by the Dockerfile for compatibility reasons.

An Accidental Blackboard

6 saikatsg 2026-09-09 01:25:57 UTC martinfowler.com LocalLLM Share 1

Thoughtworks engineers built an airline IROps system in four days using agentic engineering. They accidentally discovered a blackboard pattern for agent coordination through shared repo updates.

Animation in Bevy: The Big Picture

47 ibobev 2026-09-08 20:25:57 UTC glocq.com Llama Share 4

The author is explaining how to play an animated 3D model in a Bevy application, starting with loading the animation data from a .glb file and setting up the necessary components. They break down the process into steps, including extracting animation data, adding it to an entity's AnimationToPlay component and then finding the entity with that component.

Ask HN: 3.5 inch diskette read errors, would a period correct drive do better?

11 rietta 2026-09-09 00:25:57 UTC news.ycombinator.com Llama Share 16

You're trying to restore a 1991 Learning Company Spellbound for DOS game on an old system, but your current setup is causing read errors. Using a period-correct IDE floppy drive with the Leading Edge WinPro 486 might improve data accuracy.

Show HN: LLM Attention Visualization

132 ifz 2026-09-08 17:25:57 UTC ishamf.dev Full Share 22

Mercury 2.5

126 Topfi 2026-09-08 20:25:57 UTC inceptionlabs.ai Llama Share 16

Mercury 2.5 is the latest production model from Inception, offering a 40% increase in intelligence and comparable cost to GPT-5.6 Luna, with prices starting at $0.04 per million input and output requests.

Implementation of GCC's Nested Functions (vs. C++ Lambdas)

60 uecker 2026-09-08 18:25:57 UTC uecker.codeberg.page Llama Share 9

GCC's nested functions are implemented by collecting parent function variables accessed by the nested function into a synthetic structure and passing its pointer to the nested function. This approach decouples implementation from rest of compiler, allowing for optimization without affecting other variables or frame structure.

Replacing a Rust Enum with a 64-Bit Word Made My Interpreter 17% Faster

89 metrofun 2026-09-08 17:25:57 UTC pointersgonewild.com Llama Share 41

The author refactored the Plush language interpreter's Value type to use a low-bit tagging scheme, reducing memory usage by up to 37% in some benchmarks. The new representation is faster than the old one, with every single benchmark ending up faster.

The Helicopter with Radioactive Blades

149 zdw 2026-09-08 15:25:57 UTC hackaday.com Llama Share 46

The CH-53 Sea Stallion helicopter uses a simple mechanical system to detect rotor blade cracks, involving ionizing radiation from Strontium-90. This method has been effective for decades and is still used today in newer models with fiber optic sensors.

We built our house for LAN parties (2024)

426 fittingopposite 2026-09-08 07:25:57 UTC lanparty.house Llama Share 313

Kenton Varda and Jade Wang, a married couple with two kids, designed and built their house from scratch in Austin, Texas. The house features custom-built gaming stations with 22 game machines that cost around $75,000.

Tracing np.add, all the way down

40 luu 2026-09-08 20:25:57 UTC blog.veitheller.de Llama Share 2

The author revisits the NumPy implementation of np.add, tracing its path from Python to SIMD kernel. They explore how NumPy resolves input types and selects an iteration strategy before running a CPU-specific C function that is either a plain for loop or vectorized SIMD code.

Show HN: Copperhead – Cursor for circuit boards

209 animeshchouhan 2026-09-08 14:25:57 UTC copperhead.sh Llama Share 82

Copperhead is an open-source AI platform for designing and verifying printed circuit boards. It helps hardware teams by automating the design process, ensuring consistency and accuracy.

Reverse Engineering an ASIC

36 burekqueen 2026-09-08 21:25:57 UTC kjartanvandriel.github.io Llama Share 4

The puzzle involves analyzing a GDS file to determine the behavior of a digital circuit. The chip is designed using the SkyWater 130 nm process and consists of several logic gates, flip-flops, and buffers. By following the wires between components and grouping them by location, it can be determined that the chip implements an 11x11 grid counter with additional checks for neighboring cells.

The 92-Year-Old Mathematician and the Teenage Apprentice

145 robinhouston 2026-09-08 15:25:57 UTC nytimes.com Full Share 11

Please enable JS and disable any ad blocker

C*: Unifying Programming and Verification in C (2025)

71 rramadass 2026-09-08 16:25:57 UTC arxiv.org Llama Share 41

Researchers focus on formal verification of systems software due to its safety-critical nature. A new language, C*, integrates proof capabilities into C programming, enabling real-time verification and unifying implementation and proof code development.