Homelab
I've always wanted a homelab, but I just didn't have the time to set one up. Instead, I'm working with AI agents to design and implement an effective and secure AI-managed homelab.
A–AThe Hardware
The hardware I'm running this on is just my old desktop PC. It's not much, but I don't really plan to use this for compute. It's got an AMD Ryzen 5 2600X, a GTX 1060 6GB, and 32GB of RAM. Between SSDs and HDDs, I have roughly 5TB of storage.
B–BThe Vision
This whole project started when OpenClaw was released. There was a lot of hype around it, and I thought it'd be so sick to have a homelab that was entirely managed by AI. It'd be so easy to do things like host services, store files, watch movies, and host game servers.
C–CThe (Main) Problem
The whole problem with OpenClaw, though, is that it's pretty unsafe. Giving an agent near unlimited access to your computer opens you up to a lot of risk. An agent could make a mistake and damage your computer or expose your data, and if you let an AI model read things like API keys and passwords, your credentials get stored in the provider's servers and could be ingested into future training data. AI also makes your server susceptible to a brand new type of attack, where an attacker could leverage your own AI to attack you.
D–DThe Idea
Frankly, I'm only worried about the security side---I doubt the AI would brick my server.