A finished research run on an example topic, shown in full — the findings, the sources behind them and the product versions are all the run's own output, except that a product name may have been changed where it matched an existing brand.
A local-first meeting notetaker that records calls on-device, runs Whisper transcription and diarization, and generates summaries without sending audio to the cloud.
01
What to build
MVP
A macOS/Windows desktop app that captures system audio, runs Whisper + pyannote locally, and outputs a searchable transcript with an LLM summary via Ollama.
Why it can be better
Matches the privacy positioning of Meetily and Hyprnote but with a cleaner UI and cross-OS support from day one.
Revenue model
One-time license fee or optional paid Pro tier for advanced summarization models, paid by individual professionals.
Target user
Privacy-conscious knowledge workers and consultants who record their own meetings.
02
Context
The opportunity this answers
Build a local-first meeting-memory layer for professionals that unifies transcripts from any notetaker or capture mode into one private, searchable history.
Why now
A wave of open-source local-first notetakers built on Rust/Swift/Tauri has already proven demand for on-device processing, with diarization and Whisper models now standard, high-download building blocks anyone can bundle, while search interest for generic 'AI meeting notes' has plateaued — signaling the crowded cloud tier is saturated even as privacy-driven local alternatives keep multiplying, and none of them yet unify capture contexts, ingest rival platforms' exports, or sync back out privately.
