I run two autonomous AI agents on a $80 computer

My Raspberry Pi 5 cost about $80 on Amazon. I bought it to block ads on my home network. It didn't do a great job at that. Then I turned it into a remote controller for a 3D printer. It did worse at that. Now it sits on a shelf in a basement closet, connected to the internet, running two AI agents around the clock.

One agent handles growth and marketing across everything I'm building. It monitors performance data, executes campaigns, queues decisions that need a human. The other handles venture operations: pulling financial data, writing daily briefs, running projections, flagging risks.

They run on Node.js processes managed by PM2. The only way to reach them from outside my home network is through Tailscale, a private mesh network that doesn't expose any public ports.

Total infrastructure cost for both agents: maybe $8/month in electricity and $5/month for the VPN layer. The rest runs on cloud APIs.

The lesson isn't that you should use a Pi specifically. It's that infrastructure at early stage should match the actual load. Two agents sending emails, pulling API data, and writing briefs do not need a Kubernetes cluster. They need a computer that stays on (unless your adventurous children find it and press the green button).

Right-size the machine to the problem. You wouldn't buy a Ferrari to charge your cell phone, don't invest in a tech stack you don't need. The problem is almost always smaller than you think.