I built a voice-first AI assistant that runs behind a half mirror. It’s fully self-hosted (local LLM supported) and designed to feel ambient rather than chat-based.

I’m experimenting with what “voice-native interfaces” look like outside phones and speakers. Source code: https://github.com/orangekame3/mirrormate