Hi Product Hunt — solo developer here. 👋 Amoli started with a simple observation: my photo library already holds my much of my life — eight years, multiple countries, travel — and I couldn't ask it anything. Search found dogs and beaches. It couldn't tell me what that winter in Milan felt like. So I built a journal that reads. Amoli runs Gemma 4 E4B — a 3.7 GB vision-language model — entirely on your iPhone, and works through your library one photograph at a time, writing what it finds: who's there, where it seems to be, what was happening. Faces you name once become people it knows. Themes become threads it follows across years. And then you can ask — "when were we in Kobe?", "what did we do that spring in 2024?" — and it answers in prose, from your own photos, with the photos on the page. The defining constraint I made is that there is no server. No account, no upload, no analytics — not a privacy policy, an architecture. The model lives on the phone; your photos, your words, and everything Amoli writes never leave it. Deleting the app deletes everything. The requirements are high: it needs a recent iPhone (the model is large by design), the first launch downloads ~3.7 GB over Wi-Fi (a few minutes on decent internet), and reading a big library takes some evenings on the charger — mine read 14,000 photos at about one every six seconds. After a 14-day full trial, the ongoing AI work is a subscription — but nothing already made is ever taken back: every reading, thread, and conversation stays open, search works forever, filing new photos stays free. The principle is "your memories are always yours; only new intelligence pauses." I spent launch week chasing a GPU bug through my own family's phones (the full war story is will be on Hacker News today), so ask me anything — about on-device inference on iOS, the privacy architecture, or why a photo journal should write prose. I'll be here all day.

Amoli

Your photos become a living journal — all on your phone

Your photos become a living journal — all on your phone