just enough time to be preoccupied

I've been playing with this thing lately. (more detail here) tl;dr version, it's a local agent that runs on top of the LLM of your choice that can access the host machine and also acts as something of an MCP. (thing that allows the agent to interact with other systems.. I think it's easier to think of it as a generic API interface) It does a lot of automation stuff too, but it's really all just cron jobs. 

Anyway, I haven't really found anything useful to do with it yet other than check auction sites and for-sale forums for things I want to buy. It's been interesting as a sysadmin buddy. I don't know that it's terribly good at it, but it has the "confidence" to just do stuff that I might not these days. (though this reminds me of being a teen in the era of windows 3.1 with no internet; just doing stuff to the OS was the only thing there was to do) I've built it up to be able to build me a precis on whatever topic, and that's kind of useful. I mostly interact with it through the telegram messaging platform, so I can get this stuff away from home. It can run stable-diffusion for image generation, but I have yet to make that work reasonably, or to even find a real use for it. 

I've also been using it to manage how I do RAG (Retrieval-augmented generation) where the material it's scanning is all on the local box. 

I think the biggest, and on-going disappointment is that I can't reliably get it to go scrape a website; something that's been automatable since the 90's. It's really attached to one solution that I can't get to work. It can try to access http headless, but that usually gets shut down. Typically, you can give such things a different browser signature, but it can't do that. It also is stuck on a lot of ethics and safety rules that may or may not actually apply. I thought this was a ghost of the host LLM, but even if I run a less commercial one locally, it complains, so it must be in openclaw itself. (if it's worth noting, one thing I was trying to do was scrape certain websites around UX career things and have it ingest it all for a UX thought partner agent. It'll do it if I manually collect the stuff, but that's half the effort. 

I find myself trying to figure out a way that I could run a business off of this.. afterall, that's the narrative driving the hypecycle on this thing. But it's the same problem I have without being ai-augmented, what's the business?