Skip to content

Comment on LM Link: Use local models on remote devices, powered by Tailscale

Comments

The main thing i see here is that prompts never touch a third-party server. If you're in a regulated industry or just don't want proprietary context hitting an API, running inference on your own hardware with encrypted p2p from any device is really cool (and useful.)

(staying in userspace via tsnet without touching kernel sockets is a nice touch too.)

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.