Mac Home Server is an open-source project that transforms an always-on Mac into a family network hub using Tailscale VPN, AdGuard Home ad-blocking, cloud storage, and a mobile control panel to manage per-member permissions and time quotas. Installation requires a Mac (macOS 13+), Tailscale account, and optional external drive, with all setup automated via a bash script.
Better Call GPT is a plugin for Claude Code that enables voice interaction while coding. Users can speak commands while the AI works on their repository, with voice handling casual conversation and only forwarding actual work requests to Claude Code. The system ensures voice input never approves permission prompts, which require keyboard confirmation.
DockTerm is a free, open-source terminal workspace for Claude Code that keeps your AI session front and center while a mascot named munu notifies you from the menu bar or notch when Claude needs input, even while using fullscreen apps. It provides checkpoint navigation, live agent tracking, usage monitoring, diff review, Git integration, and file management without replacing Claude Code itself.
A sophisticated GPU smuggling network diverts NVIDIA's advanced AI chips to China despite strict US export controls, with brokers operating openly across Hong Kong and Shenzhen. Industry insiders acknowledge the illicit trade is widespread and profitable, with China estimated to have access to 2.8 million H100-equivalent GPUs through smuggling, legal purchases, and domestic production. US enforcement efforts, including a 2026 arrest of Supermicro's co-founder, have failed to disrupt the flow of chips critical to AI development and national security.
Jinei is an interactive guide to computer architecture that provides a searchable knowledge tree covering 44 structures and principles across eight major categories including processors, memory systems, GPUs, storage devices, and interconnects. The tool features visual decomposition of computer components and explanations of architectural concepts in both Chinese and English.
Vynody is a cross-platform local music player built with Flutter and Rust that offers a modern UI, lightweight performance, and advanced features like AI-generated karaoke lyrics, serverless LAN syncing, and native audio optimization across Windows, macOS, Linux, iOS, and Android.
Researchers steered a 4-billion parameter language model into simulated pain and pleasure states using activation engineering, then measured how the model responded to choices between its own relief and others' suffering. The pain signal produced coherent negative outputs up to roughly 6x dose intensity before degrading into repetitive loops, while pleasure steering remained weaker and less reliable across all measured doses.
HonorFit is a pushup app blocker that requires users to complete exercises to earn phone screen time, with no camera, motion tracking, or data collection. Built by Thomas Willheim, it tracks various exercises independently and stores all data locally on the device.
LLMxRay is a free, locally-run observability tool for monitoring large language models. It provides real-time token streaming, performance profiling, quality analysis, cost tracking, and model comparison features without requiring cloud services or API keys.
Research on steering language models toward extreme negative and positive emotional states, measuring outputs through mechanistic lenses and valence networks. Experiments on Qwen3 models show steerable affect space is primarily spanned by human emotions, with pain/pleasure directions producing dose-dependent behavioral and representational changes up to coherence breakdown around dose 8.
DeepGEMM-Ascend is a matrix multiplication kernel library ported to Huawei's Ascend NPU platform, offering API compatibility with DeepGEMM and support for BF16, FP8, and FP4 data types. It provides lightweight abstractions over Ascend hardware primitives and optimization techniques like sparse data loading and coroutine pipelining to achieve near-peak hardware performance across various matrix shapes.
Moching is a desktop AI agent built in Rust that controls your entire PC through screen perception, mouse control, and app automation. It includes 219 built-in tools across 12 domains and works with any OpenAI-compatible API, requiring Windows 10+, macOS 12+, or 4GB RAM.