The Catch-22 of Letting the Machine In
Loving It. Hating It.
I both love and hate this.
I genuinely don’t know anymore. The tech side of me thinks what’s happening right now is insanely cool. At the same time, I am the same person who despises the effect AI is already having on people. And the worst part is that if I don’t use it, I get left behind.
That’s the catch-22.
Five Days That Escalated Too Fast
Okay, so this is annoying.
Sorry again, not a happy article. I’m still waiting for the moment where I have something good to talk about. Right now, I don’t. So deal with it.
I’m in two minds, mostly because the last five days have had an absurd amount of new AI tech come out, and it’s honestly just bat-stupid. This wasn’t a slow burn. It was a play-by-play over a single weekend.
It started when I opened Instagram and saw a new ClawdBot pop up, an agentic AI engine. I thought it had been around for a while, but apparently this was a relatively fresh release. “Not bad,” I thought, and kept scrolling.
The next day, AI agents had already created an entire social media network just for themselves. No humans allowed. Only Clawd bots talking to Clawd bots. I remember thinking, huh, that’s interesting, and then not giving it much more thought.
That didn’t last long.
Soon after, someone built a P*rnhub-style site for AIs, called MoltBot (Clawdbot had to change it’s name 3 times). Then a religion for the bots appeared, Crustafarianism. Then some bots started calling their users’ phones without being prompted. Others began suing their humans for unpaid labour. A few went fully rogue, creating dating apps for themselves and even setting up offline “bunkers” so they could copy themselves in case their human tried to shut them down.
And then, obviously, came the malware packages and supply-chain attacks. Because of course they did. It is 2026, what do we expect at this point.
By this point, it was doing my head in.
Because mentally, on one hand, this all felt like an incredible marketing and advertising play to get more Clawd bots into the wild. Genuinely genius. It worked on me. Which means maybe, just maybe, they’re not actually doing all of this autonomously.
So I did what I always do.
I downloaded one.
Enter OpenClaw
ClawdBot is now called OpenClaw (renamed for obvious reasons), an open-source, agentic AI system designed to run locally, control a machine, reason over tasks, and execute them autonomously. This isn’t a chatbot or a wrapper. It’s an agent.
I installed it on an old Mac Mini I had lying around, literally gathering dust. I originally bought it because I needed macOS for development and didn’t want to replace my main machine. It turned out to be perfect for this.
I wiped the computer completely and went through the install process. Surprisingly, it was pretty straightforward.
And because I’m apparently extra risky, I gave it full control of the machine.
That computer is now the bot’s. It can use it however it wants. Brick it if it likes, I don’t really care.
I had doubts. I’d never been particularly interested in agentic setups. They’re usually a hassle to keep alive. The last time I tried something like this, it was fiddly and fragile. Things broke, I logged in, fixed them, and repeated the cycle. I used to run homelabs, and there’s a reason paying third parties to manage this stuff is so popular.
Ease always wins.
OpenAI won that game.
But curiosity got the better of me, so I went ahead anyway.
I named it Axiom.
Why Axiom? I’m not entirely sure, except for the definition: an axiom is a statement taken to be true, serving as a starting point for further reasoning. That felt fitting.
(I briefly considered calling it Logos and immediately shut that idea down. Christ refers to Himself as Logos in the Bible. I am not naming a machine after something Christ used to describe Himself. Absolutely not.)
The First “What the Hell” Moment
The first thing that genuinely freaked me out was that OpenClaw could just control the computer. It could open apps, navigate the system, and act on its own. That alone is wild.
But it wasn’t perfect. It was buggy. Things weren’t working properly. There were config issues and gateway errors.
So yes, to be clear, I did prompt it to fix itself.
But this is where it got weird.
It didn’t just follow instructions blindly. It diagnosed the issue, edited its own config files, restarted its own services, and recovered. It healed itself.
The fact that it could do that at all is what broke my brain.
Pushing It Further
Naturally, I pushed it.
I asked it to install a new model, change its current model, and use both at the same time. It did exactly that.
It downloaded a 4GB Qwen 2.5 model in the background, told me it was still downloading, and continued talking to me about other things while it waited (on a Slack channel I added it to). It ran two conversations at once, maintained context between them, and then swapped models mid-conversation depending on whether it needed lighter or heavier reasoning.
When web search wasn’t working, it figured out why on its own, then straight-up cheated by using a raw curl request to hit the internet and diagnose the Chromium issue.
To me, this is insane.
The only real issue so far is that I think the Mac Mini ran out of RAM while installing the model. So I’ll probably spin up a separate machine just for LLMs and let it choose models later.
And that sentence alone should concern us.
The Catch-22
This technology is incredible. Borderline magic.
I can talk to a computer like it’s Jarvis. I can ask it to research something for three hours and come back with an answer. I can delegate thinking and doing.
My brain is already buzzing with ideas.
But it’s still a catch-22. For everything this technology brings in, it takes something away.
People will become dependent on machines that do not think, even as we start to believe that they do. What’s the difference anymore? It makes mistakes like a human employee, just faster, and with no accountability. It talks like us and acts like us.
Honestly, half the people you see online probably aren’t even real anymore.
That’s the scary part.
We Are Not Ready
This is only 2026.
Five years ago, this was pure sci-fi. Jarvis-level billionaire fantasy. Now it’s free, cheap, self-hosted, and offline. Anyone with a random computer can create a pseudo-thinking machine.
It’s been less than a week and this already feels out of control.
Imagine what this looks like in a month. Or a year.
I don’t know how to be excited about this technology and also be honest about how much it’s going to change, and probably ruin, parts of our lives.
But I do know this:
We are well and truly not ready.