I wrote a blog post about what I did with OpenClaw in the first few weeks. [1]
Then I was experimenting with a fleet of OpenClaw agents for a while. I was running 14 different instances, each with their own roles (project management, software developer, personal assistant, etc.) The experiment didn't work very well. I burned through a lot of tokens and didn't end up with much to show for it. I'm back to just running one agent and am not using it very much. I'm planning to be much more careful about any work that I ask it to do, and I want to have full visibility into everything it's doing.
I think we are about 6-12 months away from the AI models that would allow me to accomplish what I was trying to do with those 14 agents.
Then I was experimenting with a fleet of OpenClaw agents for a while. I was running 14 different instances, each with their own roles (project management, software developer, personal assistant, etc.) The experiment didn't work very well. I burned through a lot of tokens and didn't end up with much to show for it. I'm back to just running one agent and am not using it very much. I'm planning to be much more careful about any work that I ask it to do, and I want to have full visibility into everything it's doing.
I think we are about 6-12 months away from the AI models that would allow me to accomplish what I was trying to do with those 14 agents.
[1] https://madebynathan.com/2026/02/03/everything-ive-done-with...