AI research focused on privacy-first, local-only inference. No cloud required. Everything runs on your machine.
A pure Rust audio engine that runs Moonshine speech recognition (Metal-accelerated via Candle), voice activity detection, and local LLM summarization entirely on your Mac. Nothing leaves your machine.
Every decision optimized for privacy and performance.
No API keys, no accounts, no telemetry. Your audio never touches a server.
Metal-accelerated inference tuned for M-series chips. Realtime transcription on consumer hardware.
No Python runtime, no Electron shell. A single statically-linked binary under 1 MB.
On-device language models generate structured summaries. Intelligence stays on your hardware.
Get notified when we ship new research and releases.
No spam. Rare updates only. Unsubscribe anytime.