Heiko is coming.
My father-in-law visits California for the first time this summer.
He grew up in East Germany, learnt Russian at school, and doesn't speak English.
I built this so he can go out alone.
Second goal: become the favourite son-in-law.
2026
Family software
Built alone, over a few weekends
Interaction Design, iOS Engineering,
Real-Time Speech Systems
Gemini Live API
Written with Claude Code
Heiko is coming.
My father-in-law visits California for the first time this summer.
He grew up in East Germany, learnt Russian at school, and doesn't speak English.
I built this so he can go out alone.
Second goal: become the favourite son-in-law.
One button. Waiting, listening, mid-conversation.
German is always the big text, whatever language comes in.
California is not only English. Six languages, any pairing.
The hard part was never the translation. It was knowing who spoke.
Software used to be too expensive
to build for one person.
That has changed.
He speaks and reads German and is not technical.
Every decision follows from one person.
37 logic tests. 56 replay assertions against the live API
Who built it?
I did, alone, on weekends, with Claude. I spend my working life handing designs to engineers. I wanted to know what happened if I didn’t hand this one over.
Why not just use Apple or Google Translate?
Often you should. They do far more: more languages, offline, a camera, a keyboard. They are translators.
This is an interpreter. It listens to two people and works out who just spoke — the part that breaks a conversation when it gets it wrong. One job, for two people standing face to face.
Does it record what I say?
Not on any server of mine. The transcript and the technical log stay on the iPhone and are deleted with the app. Your speech does go to Google’s Gemini API to be translated — without that there is no translation, and there is no offline mode. No analytics, no advertising, no accounts, no sign-in.
