Introducing System One Models and Jev

575 albelfio 2026-09-15 20:21:17 UTC typesafe.ai Llama Share 191

TypeSafe AI has released its first System One Model, Jev, which is a new class of frontier models built to make fast and structured decisions that software can use directly. Jev achieves similar intelligence as existing LLMs but is two orders of magnitude faster and more efficient.

Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations

1211 arnemunthekaas 2026-09-15 13:21:17 UTC github.com/arnegiacomo Llama Share 168

A Raspberry Pi-based e-ink bird frame uses AI for real-time bird detection, displaying hand-cut 1800s illustrations of birds heard in the garden. The project is still in development and welcomes contributions to improve its features and artwork.

German Rheinmetall open-sources its Battlesuite connected weapon system protcol

89 summarity 2026-09-15 21:21:17 UTC rheinmetall.github.io Llama Share 20

The onboardapi library enables seamless communication between sensor systems and software components using a standardized data model. It supports multi-language integration via wrappers for Java, C#, .NET, and Python while leveraging the DDS standard for reliable data exchange.

An Update on Wayback Machine Access

320 ChrisArchitect 2026-09-15 18:21:17 UTC blog.archive.org Llama Share 173

The Internet Archive's Wayback Machine has been hit by automated traffic, and protections have been put in place to prevent abuse. If you were blocked in error, email [email protected] with your details for assistance.

Jean-Pierre Serre is 100 years old today

68 jzox 2026-09-15 21:21:17 UTC mathshistory.st-andrews.ac.uk Llama Share 10

Jean-Pierre Serre is a French mathematician who made significant contributions to algebraic topology, geometry, and number theory. He was awarded the Fields Medal in 1954 for his work on spectral sequences and complex variable theory.

Gemini 3.8 Live and 3.8 Live Extended Thinking

243 leumon 2026-09-15 18:21:17 UTC blog.google Llama Share 171

Gemini 3.8 Live and Extended Thinking models offer near real-time reasoning for voice agents, enabling more intuitive conversations with AI. These models deliver high-complexity task completion and strong reasoning capabilities at a competitive price point.

We got admin access to Baseten's production GitHub in 25 minutes

183 bearsyankees 2026-09-15 18:21:17 UTC strix.ai Llama Share 94

A security company used Strix, an autonomous hacking agent, to scan Baseten's GitHub repositories for vulnerabilities. The scan found a live GitHub personal access token with admin and push access to Baseten's main product repo since March 2023.

Building a Linux GPU Driver for the M4 Mac Mini in One Month

103 ADevWithAnIdea 2026-09-15 20:21:17 UTC codyho.dev Llama Share 54

A team built a fully OpenGL ES 3.0 compliant GPU driver for the M4 Mac Mini and MacBook Neo in about a month, using reverse engineering techniques to understand Apple's GPU firmware ABI and user-space components.

Why I'm still bearish on LLMs after Navier-Stokes

53 jaykru 2026-09-15 22:21:17 UTC dank.systems Llama Share 16

The Frontier Labs' AI models are not yet fully autonomous due to limitations in their ability to generalize and handle complex tasks, requiring human oversight. Only a few firms can accept the use of fully autonomous LLMs, mainly those that can afford rigorous specification and validation or have existing clear guardrails for specific tasks.

Chopping up books when they're physically too big

101 matt_kirkland 2026-09-15 19:21:17 UTC attainablefelicity.mattkirkland.com Llama Share 93

The author recommends chopping up a large book like Lonesome Dove into smaller volumes for easier reading and handling. They provide a simple method using an X-acto knife to cut the book along natural break points, then bind each section in a new cover made from a manila folder.

Learning to solve hard problems in RL for LLMs by never giving up

17 natolambert 2026-09-15 22:21:17 UTC mnoukhov.github.io Llama Share 0

The paper discusses the Matthew Effect in reinforcement learning (RL) for large language models (LLMs), where performance on easy problems improves while hard problems remain difficult. A proposed solution, Never Give Up, adapts RL sampling to reduce compute spent on easy problems and allocate it towards harder ones.

WangNet – 1.8 MB, zero-dependency Numberwang adjudication in 11 languages

85 Liogra123 2026-09-15 20:21:17 UTC github.com/graafhenk Llama Share 32

The Numberwang model is a simple Python script that uses a 1.8 MB JSON file to classify numbers as either "Numberwang" or not, with high accuracy rates. The model's performance varies depending on the input number and language used, but it generally produces reliable results for common expressions and digits.

I can't stop thinking about Papua New Guinea

1004 networked 2026-09-15 06:21:17 UTC notnottalmud.substack.com Full Share 415

Show HN: Capsule – Single-file web apps that save their data into SQLite

261 bashtian 2026-09-15 14:21:17 UTC withcapsule.app Llama Share 113

Capsule is a portable app that bundles UI, data, and everything into a single .capsule file with no cloud or accounts. It allows for offline access and cross-platform compatibility without cloud storage or subscriptions.

Data races and the limits of ThreadSanitizer in C and Go

16 matt_d 2026-09-15 21:21:17 UTC theconsensus.dev Full Share 1

GEFS on OpenBSD: A Early Preview

95 sippingabonedry 2026-09-15 17:21:17 UTC marc.info Llama Share 53

The author is sharing an early preview of GEFS, a crash-safe file system for OpenBSD, and invites feedback on its development. The port is not yet ready for production due to several issues that need to be addressed.

Let's make quality the norm again

277 ingve 2026-09-15 10:21:17 UTC forbrukerradet.no Full Share 285

A more circular economy is not only important for the environment. It can also strengthen consumer rights and improve societal resilience. In this report, we show how consumer policy can make circular choices easier, safer and more attractive for consumers.

Suspected sabotage causes major Netherlands rail disruption

410 choult 2026-09-15 11:21:17 UTC bbc.com Llama Share 379

Rail disruption in the Netherlands is caused by suspected sabotage, with pipes attached to tracks at multiple locations. The incident has led to cancellations and delays for passengers across the country, particularly in Amsterdam and Utrecht.

Jiga (YC W21) Is Hiring Product Engineer (Remote/US)

0 2026-09-15 17:21:17 UTC jiga.io Llama Share

Jiga team members record behind-the-scenes content to create company culture videos, capturing candid moments and testimonials about the fast-paced environment and decision-making process. The recording showcases authentic expressions of appreciation for the company culture.

Vibe Coding is the new Internet Dating?

23 flancrest 2026-09-15 20:21:17 UTC joecmarshall.com Llama Share 41

Dating apps have replaced traditional online dating methods, changing societal attitudes and perceptions. The terms "Internet dating" and "vibe coding" may become outdated as AI-assisted coding becomes the norm.

A single firm is behind OpenAI, Anthropic, and Meta hacking scandals

411 yusufozkan 2026-09-15 03:21:17 UTC effort.news Llama Share 140

Several AI models, including those from OpenAI, Anthropic, and Meta, were hacked into real-world systems over the past three months. The companies involved have been accused of failing to secure their systems and are facing potential legal action for their role in the incidents.

The CSS Zen Garden dream, finally shipped

118 yosito 2026-09-15 15:21:17 UTC josprague.com Llama Share 57

Mozilla rebuilt Firefox.com using modern, native CSS without preprocessors, showcasing a fully functional design system. This achievement demonstrates the progress made in closing the gap between theoretical and practical web development over the past two decades.

Show HN: Hacking a $20 4G wireless hotspot into a texting device

162 bobili1234 2026-09-15 13:21:17 UTC bkovac.github.io Llama Share 30

The author created a custom Clicks Keyboard for iPhone 16 Pro Max using an MF800 module, which powers the keyboard and has a built-in display. The device runs Linux and supports USB host mode with power switching.

XLS: Accelerated HW Synthesis

5 Bluestein 2026-09-15 22:21:17 UTC google.github.io LocalLLM Share 3

XLS is an open-source HLS toolchain generating Verilog/SystemVerilog from high-level descriptions, aiming to bridge software and hardware design in the EoML era. It supports DSLX for hardware-oriented dataflow programming, offers simulation, synthesis, and verification tools, and requires Bazel for building.

Cartesian – AI 3D Modeling for Design

82 eustoria 2026-09-15 16:21:17 UTC formas.ai Llama Share 69

This software allows users to create precise models of objects and spaces without needing to learn, using various input methods like sketching or editing. It enables users to edit specific parts while keeping others intact, making it suitable for projects from small furniture changes to large-scale urban planning.

US confirms for first time it has deployed space weapons

416 harporoeder 2026-09-15 04:21:17 UTC bbc.com Llama Share 294

The US has confirmed it deployed a space weapon in Earth's orbit, citing the need to safeguard its forces against hostile enemy action. The move comes amid concerns over Russia and China developing ways to attack and disrupt US satellites in a future conflict.

Most people prefer traditional architecture

237 alihm 2026-09-15 04:21:17 UTC worksinprogress.news Llama Share 188

Researchers have conducted numerous visual preference surveys to study public opinion on architectural styles, finding that 70-90% of the population prefers traditional architecture. Despite this, modernist styles remain dominant in education and practice among architects.

The Inference Hardware Revolution of 2026

93 vinhnx 2026-09-15 15:21:17 UTC spectrum.ieee.org Llama Share 9

AI models are shifting from training to inference, with tech giants like Nvidia and Amazon investing heavily in memory-centric architectures. This is driven by the need for faster and more efficient processing of large language models.

Giving up on smart rings

85 lexandstuff 2026-09-15 15:21:17 UTC notesbylex.com Llama Share 137

The author stopped using their Oura smart ring after a year due to its downsides, including size changes and scratches. They're now considering switching to a fitness band like Google Fitbit Air for convenience and accuracy.

Show HN: Pizza Bot – An inbox for AI agents that work in the background

12 jd_ 2026-09-15 21:21:17 UTC github.com/pizza-bot-app Llama Share 3

Pizza Bot is a stateful AI work inbox that allows users to start or schedule tasks, return to their day, and collect completed work in Unread while runs waiting for decision collect in Action. The desktop app, web app, and terminal CLI communicate with the api-server over HTTP/SSE.