One-tap voice recorder on iPhone/Apple Watch; recordings sync via iCloud to the Mac. It’s the capture front-end of the dictation pipeline — everything downstream (Whisper large-v3 transcription, dash-command parsing, task extraction, journal routing to Notebooks) is capture-agnostic.

• Target for recordings: ~/Sync/ED/dictation/inbox/ (the field-dictation path) and ~/Sync/farm/emlid-exports/voice-memos/ for GPS-survey voice memos.
• Status note: the dedicated field recorder was lost in June 2026 and isn’t being replaced — JPR on Watch/iPhone (or AirDrop of .mp3s into ~/Downloads, which Hazel routes to the inbox) is the capture path. Diary volume has been low since; the Ed dashboard diary section reflects that honestly rather than looking broken.
• Transcription details: whisper-transcribe.py (mlx-whisper, auto-chunks recordings >10 min; hallucination filters). See dictation pipeline docs for the full flow.