Glimps
PrototypeA personal memory search engine. Capture moments as voice notes or photos, then ask for them back in plain language.
Last updated Apr 8, 2026
Early prototype
This project is a working prototype — the core concept is proven but it is not yet production-ready.
The problem
The moments worth keeping end up as undated voice memos and unlabelled photos. You can search a photo library for a face or a place, but not for “that thing someone said at the lake in August.”
What’s being built
Capture by voice or photo, then retrieve by natural-language query, with every answer grounded in stored memories — the model never fabricates an event or infers an experience that wasn't recorded. React Native (Expo) client over a Node, Express, and TypeScript backend, PostgreSQL with pgvector for embedding search, transcription and vision models in the ingestion pipeline, and a memory-resurfacing layer on top.
Latest update
The core promise held through the build: every answer is grounded in a stored memory, so the system will say it doesn't know rather than invent one.
Current milestone
All phases complete — ingestion, retrieval, answer generation, mobile app, and resurfacing
Tech stack
A personal archive. Answers are grounded strictly in what you recorded — Glimps never invents a memory you didn't capture.
Follow this project
Get notified when there's a meaningful update — no noise, no marketing.
Confirm via email. Unsubscribe any time. No spam — ever.