I tried Local LLM Models with OpenClaw (unsuccessfully)

AI is apparently winning hearts and minds everywhere. Quick! Get an agent or you’ll be…{insert something drammatic}!

Right…

I’ll admit it, I’m on the skeptic side of things and AI is not an exception. Especially working in digital where you hear all the AI horror stories, but I finally caved and gave it a whirl… on my own terms.

I repurposed an old PC, slapped Ubuntu on it (didn’t have a USB big enough for a Windows install), and loaded up OpenClaw, the tool everyone seems to be raving about right now.

Instead of tying it to the cloud, I wanted to keep things secure and offline, so I downloaded smaller open models via Ollama.

The verdict? A total bust.

Even the most basic prompts took an agonizingly long time. I knew these models craved serious compute power, but even after continually downgrading to the smallest models available, it was completely unworkable. The local, offline concept is neat, but OpenClaw just couldn’t get anything done on this ancient hardware.

Until I get a significantly beefier machine, this project is firmly on the shelf. Maybe it’ll be useful for a local smart home setup someday? Who knows. Let’s keep digging at this.