Nightingale – open-source karaoke app that works with any song on your computer
🎤 Stem separation Vocals are isolated from instrumentals using the UVR Karaoke model or Demucs. Guide vocal volume is adjustable. 📝 Word-level lyrics WhisperX transcribes and aligns every word to the audio. Existing lyrics from LRCLIB are used when available. 🎯 Pitch scoring Sing into your mic and get scored in real-time. Star ratings and […]
Ndea (YC W26) is hiring a symbolic RL search guidance lead
Location: RemoteType: Full-time Ndea is building AGI systems that depend on search guidance. We’re hiring hands-on researchers/engineers to drive our deep learning effort in search guidance, training models to make structured search more efficient and reliable. This is a high-ownership role for someone who wants to operate at the leading edge of neuro-symbolic AI, designing, […]
Celebrating Tony Hoare’s mark on computer science
Tony Hoare at the LASER summer school, September 2007(All photographs in this article are by the author) Had they included just one of Tony Hoare’s major achievements, many scientific careers would be considered prestigious enough. His had a long list, which I am going to try to summarize, not pretending to get anywhere close to […]
SSH has no Host header
We have a challenge with ssh. Every VM has a standard URL that we use for both HTTPS and SSH, e.g. undefined-behavior.exe.xyz. Just as you can type the domain into a web browser (and have TLS and auth taken care of for you), you can run: ssh undefined-behavior.exe.xyz To get a shell in your VM. […]
Meta’s Omnilingual MT for 1,600 Languages
Abstract Advances made through No Language Left Behind (NLLB) have demonstrated that high-quality machine translation (MT) scale to 200 languages. Later Large Language Models (LLMs) have been adopted for MT, increasing in quality but not necessarily extending language coverage. Current systems remain constrained by limited coverage and a persistent generation bottleneck: while crosslingual transfer enables […]
Have a Fucking Website
Look, I get it. The foundations of the internet are broken and we’ve somehow gotten to a place where having a website is either expensive, complicated, or perceived as unnecessary, whereas social media platforms are cheap and easy. But still, please, if you are a business or an individual artist or creator, have a fucking […]
Launch an autonomous AI agent with sandboxed execution in 2 lines of code
This notebook demonstrates how to use the Agent pipeline from OnPrem.LLM to create autonomous agents that can execute complex tasks using a variety of tools. The pipeline works with any LiteLLM-supported model that supports tool-calling: Cloud: openai/gpt-5.2-codex, anthropic/claude-sonnet-4-5, gemini/gemini-1.5-pro Local: Ollama (ollama/llama3.1), vLLM (hosted_vllm/), llama.cpp (use OpenAI interface) For llama.cpp: Use openai/ (e.g., gpt-oss-120b) as […]
The Pleasures of Poor Product Design
(Photo © Katerina Kamprani — The Uncomfortable) Note: This post is not paywalled. Enjoy! — Paul As you all know, I’m a big fan of good design. But I recently learned of a project devoted to bad design, and I can’t get enough of it. Welcome to The Uncomfortable, the brilliantly named brainchild of a […]
Mamba-3
Since the release of Mamba-2 in mid-2024, most architectures have switched from Mamba-1. Why? Mamba-2 made the bet that training efficiency was the largest bottleneck for state space models (SSMs), and thus simplified the underlying SSM mechanism to deliver 2−8× faster training compared to its predecessor, leading to wider adoption. Since then, the LLM landscape […]
Mistral AI Releases Forge
Today, we’re introducing Forge, a system for enterprises to build frontier-grade AI models grounded in their proprietary knowledge. Most AI models available today are trained primarily on publicly available data. They are designed to perform well across a broad range of tasks. But enterprises operate using internal knowledge: engineering standards, compliance policies, codebases, operational processes, […]