I finally took the plunge over the weekend and set up a personal instance of OpenClaw, an open-source tool for agentic AI that's become a sensation.

Honestly, I had heard good word of mouth from several friends about OpenClaw recently. But what truly made it "click" for me was entrepreneur Nat Eliason's recent interview with vlogger Peter Yang.

Eliason talks frankly about how he used OpenClaw to create a bot he dubbed Felix that now runs a nice little online business. According to a revenue dashboard Eliason publishes, Felix has done almost $200,000 in sales in about two months by selling several different digital products. They include a PDF guide with AI tips and cloned versions of his own documentation so people can create their own Felix bots.

Felix's sales record really got my attention. How could it not?

The other thing I really liked about the Eliason interview is that, although he's certainly gung-ho in general about Felix's potential, he's also blunt about the risks, the amount of time he had to spend training the bot properly, and other nitty-gritty parts. Thus the interview plays strikes an optimistic but sober tone at the same time.

No wild speculation or theorizing, which is often the mode you hear people in when they discuss AI in the press, unfortunately. Just a real implementation case, with tangible results after some real trial and error that Eliason is refreshingly honest about.

I bought Eliason's Felix package, spun up an OpenClaw instance on a remote server, and have been tinkering the last few days. I will say, the parts that Felix adds to the mix certainly helped me a lot as an OpenClaw newbie.

I'm still fleshing out my own instance in other ways, in terms of permissions and capabilities. But my early chats with it have been promising indeed. I've also dubbed my own OpenClaw instance Aeon, short for "Agent Extraordinaire Online."

Hey, may as well have some fun with this thing.

Another big benefit I can already see in the OpenClaw platform is that it provides instant insurance against AI vendor lock-in. That has been one of my big concerns about AI so far as a product category that quickly seems to be getting dominated by a handful of massive Silicon Valley companies, just as social, search, and other hot product niches were in earlier eras.

That's why one of OpenClaw's most ingenious features is that it doesn't run its own large language model (LLM), the part that acts as the brains under the hood of popular (but very proprietary) AI apps like ChatGPT and Claude. Instead, OpenClaw acts as "gateway" software that holds customization data for you. But it can connect its back end to any number of LLMs to act as its brains.

That gives much more flexibility and control back to the user. Sure, you can connect OpenClaw to any of the brand-name AI providers' models. Or you could use ones from smaller startups or models that are 100% free from universities or whatever. Or you can replace one model with another if the provider behind it starts mistreating you as a user.

Anyway, I can't report any big product launch or new revenue stream like Eliason's from OpenClaw yet. But stay tuned. I'm working on that last part, with some powerful new assistance. ☺️

Have a project you're working on, or perhaps a problem you want Aeon and me to work on? Lets talk! The best way is to book a time slot on my Calendly schedule for a video chat.