The AI That Actually Does Things

The social experiment turned security nightmare

You may have heard a lot lately about ClawdBot, renamed to MoltBot, and now OpenClaw. If not, it is basically an agentic AI system that runs locally on your own machine. It is basically being framed as AI that gets things done.

And it works. Really well in-fact.

It has exploded over the last few weeks with thousands of new skills being added to it, hundreds of thousands of developers contributing to it, and it caused a massage surge in sales of Mac Minis to run it on (not needed BTW).

But, just a warning, it is a security nightmare at the moment exposing people to malware, crypto scams, and huge personal data leakage.

I am going to say unequivocally, don’t install it yet. Let them work on security before you let this loose in your environment, and if you do want to play with it then expect some high API usage fees, and please run it like I do in a very restrictive virtual machine.

But, the point of this article is, not to neg on OpenClaw, but to applaud it. People are genuinely desperate to test out and play with agentic technology. OpenClaw as it stands right now can automate simple tasks, write code, setup a VPS server to your specifications, do research, even book flights.

The opportunity here, and the demand is staggering. Recently Anthropic released ClaudeBot which can do very similar things, and it is fantastic, and very restrictive for the same kinds of security problems OpenClaw is having right now.

But this really is the future. Happening right in front of our eyes, and if you keep an eye on this project over the next few months I guarantee that it will shape our collective future use of AI in our daily lives.

Just don’t install it yet.