Luthen
Decentralized video archiving with on-device AI
Organizations doing on-the-ground documentation — legal observers, community accountability teams, protest documentation projects — need to record and preserve video evidence. The problem with centralized storage is not just privacy; it’s reliability. A server can be subpoenaed, seized, or taken offline at exactly the moment the documentation matters most.
On-device ML for video processing is increasingly feasible. Distributed storage is a solved problem in isolation. The integration challenge is building a system that holds up under adversarial conditions — device seizure, network surveillance, hostile infrastructure — while remaining usable for people who are not trained technologists. These requirements are in tension in ways that make this a genuinely hard design problem.
Luthen keeps recordings distributed across participants’ devices, processes video and audio locally, and never routes sensitive data through a central server. The name is from the Andor series: the logistics organizer who keeps the network running without anyone knowing the full picture.