PHNTM One: Your own AI assistant on your desk. Offline. No subscription

Hi Product Hunt, I'm Jacob, and I build PHNTM One by hand, one unit at a time. I wanted an assistant that was actually mine: on my desk, remembering what I told it last week, without sending a word of it to someone else's computer. So I built one. It's a Raspberry Pi 5 (8 GB) with a touchscreen, running Google's open-weights Gemma 3 4B plus local speech-to-text and text-to-speech. Unplug your router and it keeps working. I encourage you to try exactly that. What I'd like you to hold me to: - "Nothing leaves the box" comes with a method, not a promise: a tcpdump filter and a live on-device view of outbound connections. - The numbers are measured and public, including the bad ones: ~15-20 s per answer when warm, ~90-106 s for the first answer after a long idle. - The full parts list with what I paid ($409.95) is published. So are unedited transcripts, including the ones where the small model is confidently wrong. - If I disappear tomorrow, every unit keeps working. There are no servers. Who it's NOT for: if you want frontier-level reasoning or instant answers, use a cloud model. If you have a good GPU and like tinkering, run Ollama yourself; it's free and faster. This is for people who want the finished, private thing on a dedicated box. There's an optional bring-your-own-API-key mode for hard questions. It's off by default and every answer that left the device is labelled. I'll be here all day. Ask me anything, especially the hard questions.