HN Explorer

Show HN: Minute – Offline meeting notes on macOS with Whisper and llama.cpp (github.com)

12 points by mraza007 ↗ HN
Hello HN,

I built Minute because I wanted searchable meeting notes without sending recordings or transcripts to a cloud service. It captures microphone and optional system audio, transcribes locally with Whisper, and generates summaries, decisions, and action items using a local LLM through llama.cpp. After the initial model download, everything runs on-device and in-process. There’s no account, server, or telemetry. The app is built with Tauri, Rust, React, and Metal acceleration. Notes are stored locally as plain text, JSON, and WAV files.

I built the project end-to-end with Codex as a design and engineering partner—including the UX, accessibility work, reliability tests, CI, and Apple Silicon/Intel release automation.

I guess software engineering isn’t ending. The leverage is getting ridiculous.

5 comments

[ 0.37 ms ] story [ 8.1 ms ] thread
tylerak
Clever, and I really like the design. The interface is how my brain works. Why the extra effort to support Intel Macs? What was it like building with Tauri? I only have experience with egui. Thanks!
udfalkso
This looks really nice! Gonna try it out. Thanks
dgsl007
Did the same thing for a different use case, and the VAD ate way more of my time than the model did. Are you doing diarization ?
k-thimmaraju
Worth noting the business model here. Minute ships the actual model, Whisper plus Qwen or Gemma, fully offline. Works because for a meeting-notes app the model is a commodity, the moat is the UX and the fact that nothing leaves the machine. Different bet than a product where the model itself is the product, evaluation or classification work, where giving the model away gives the business away.