dex.
Say it. It's handled.
1. Your voice never leaves your phone. Speech recognition runs on-device. Audio recordings are never uploaded, never backed up, never used to train anything.
2. Everything you capture lives on this phone. Your notes, tasks and memory are stored locally first. Cloud backup is optional, on by default, and you can pause or wipe it any time in Settings.
3. You can delete everything, any time. In Settings → Delete everything. No dark patterns, no retention games.
Backed up (only with your consent): transcripts, tasks, memories, places. Never backed up: audio. Location is captured only in the moment you speak, never tracked, and the precise coordinates are automatically deleted after 90 days — named places stay.
To turn your spoken notes into tasks and memories, the text of a note is sent to a language-model service (DeepSeek via OpenRouter, and Google Gemini) through our gateway. These providers receive the note text only — never audio, never your email, never your location. The text is processed to extract your tasks and memories, and is not used to train their models. If you prefer, Local-Only mode keeps everything on your phone with no AI processing.
Product analytics: we use PostHog (open source) to see which features are used — event names and screens only. It never includes note text, transcripts, audio, or your location. Session recording is disabled. Questions: support@builder-singh.com.
Builder Singh · support@builder-singh.com