DeepGEMM is a unified, high-performance Tensor Core kernel library that brings together key computation primitives of modern large language models, such as FP8, FP4, BF16 GEMMs, fused MoE with overlapped communication, MQA scoring, and HyperConnection, into a single, cohesive CUDA codebase. All kernels are compiled at runtime via a lightweight Just-In-Time (JIT) module, requiring no CUDA compilation during installation. Despite its lightweight design, DeepGEMM's performance matches or exceeds expert-tuned libraries across various matrix shapes.
Voicebox is a local-first, open-source AI voice studio by jamiepine, offering a private alternative to ElevenLabs and WisprFlow. It features high-fidelity voice cloning, speech generation across 23 languages via 7 different engines (like Qwen3-TTS and Kokoro), and global dictation via STT. With its native MCP server, it enables AI agents such as Claude and Cursor to speak using custom-cloned voices, ensuring all voice data and models remain securely on the user's local machine.
