BrainDump
Dump a rough thought by voice or text in one tap, and Gemini Nano, already on your phone, turns it into a clean Markdown note that sounds like you.
- Year
- 2026
- Role
- Solo
- Status
- In progress
- Context
- KORE Tech (my startup)
Results
0
network permissions
Speech and LLM both run on the phone
3
dumps to learn your voice
Persona re-learned every 5 dumps
Capture is one tap from anywhere: a Quick Settings tile, a home-screen widget that resizes from a single mic button to a 2×2 card, a launcher shortcut, or the in-app buttons. Voice goes through the phone’s on-device speech recognizer and fills the text box live.
Each dump is saved to Room, then sent to Gemini Nano through the ML Kit GenAI Prompt API, which rewrites it as a polished Markdown note with a title and a category. If the active Nano build returns nothing, the app tries the other builds AICore has installed and keeps the one that works.
After 3 dumps, Gemini Nano reads your raw words (never its own notes) and sums up how you write: voice, tone, mood, structure and focus. That short persona card travels with each new dump so notes sound like you, is re-learned every 5 dumps, and can be turned off or forgotten.
Notes are grouped by day; you can chat with Gemini Nano to change one or share it as Markdown. Nano only answers foreground apps, so an interrupted dump is kept and polished next time, with a Retry for anything that failed. The app has no Internet permission, and GitHub Actions builds an R8-optimized APK on every push.