"We become what we behold. We shape our tools and then our tools shape us"

-- Father John Culkin

Sometime in 2026, there was a schism in AI world.

Sure, there were plenty of people against all use of AI, some for good reason, but those ended up in the minority. Most jobs have some AI in it. Many places mandate its use.

The schism wasn't AI vs Non-AI. It was ultimately down to Citizen and Serf. Do you control your compute, or does it control you?

The Serf

The Serf isn't likely a technology sophisticate. They know how to open Zoom and Outlook on their work-issued ChromeBook and that's about it. They're happy like this. They don't want to peek behind the magic curtain. They're content with this. They don't want to think about heavy stuff.

So you can imagine that they're happy with using ChatGPT or Claude for everything. It does everything! Just enter text in box, hit enter, wait a few seconds, copy and paste. This is their professional existence now.

What they're not considering: where is that input going? Is the output factual or real? Can Anthropic decide that their action is now "harmful" and stop it? Where does the buck stop, is it you for sending company data to Anthropic or is it Anthropic for returning bad data?

This is the lot of the Serf. They don't really get a say in what happens. They're glorified assembly line workers, and while they may well be well compensated, their work could be automated. They know this. Their bosses know this.

So they're miserable. Just white noise in the background. They didn't learn their tools, now their tools are self-operating and they're left without an identity other than "Worker Drone". Less than dirt.

The Citizen

The Citizen is more sophisticated than the Serf. The Citizen likely knows how to use a terminal, even if they get stuck opening vim. While they're happy, there's always a scratch for them to itch. They want to do more than act as a ChatGPT jockey.

With this in mind, they start to explore their machinery. Fine, they want to bin Windows, it sucks, and they hear the horror stories of Microsoft spying on you. They install a Linux. Doesn't matter what kind, they're all kind of the same, but a Linux allows them to use the hardware paid for in a way that does not require permission for anything.

They see how much faster everything is. No more bloat. No more ads in the file explorer. So they sit down at Claude and start to think. Could Claude be better? Can I keep ownership of my data? Can I deny the Automation Machine fuel to automate me into a fine paste?

They start to look into ollama and download a Qwen. Sure enough, a Qwen does a lot. Not as much as Claude, but 75% of the scutwork tasks like converting a candidate's CV into a Markdown document for that clever dick CTO to look at, it seems to do well at.

But that's not it. They hear from Anons that they can get the machine to do everything for them. So they start to look into agents. They download pi and ask Qwen to build them a webpage that looks nice, and sure enough, it does well enough at that. But wait there's more: they hear of something called OpenClaw and Hermes, which can do more than code.

It can manage other things too. It can manage other agents. The machine works for the Citizen. And all of a sudden, the world spawns another portal.

Pick one

We bring ourselves back to today. These futures are fixed and here. Nobody can un-invent AI, but if you want to do more than be a serf, you can choose this path.

I don't hold any value judgment on the Serf, I'm sure that for someone who doesn't have the mental bandwidth to dig deeper, it's a cozy existence, but it's not a path I would choose.

Learn your tools, or your tools will learn how to use you.

Pick one.

Postscript

• Tokenmaxxers: sorry but you are a Serf who's being bounced by marketing into dumping your whole org's brain into an accountability sink. Mythos won't be for you. It'll replace you. The calculation has been made: AI slop is cheaper than human slop.

• I do not pay for AI services. I self-host. My Hermes agent does a lot for me. I'd skip OpenClaw. Hermes has a better agent system. My AI service runs on a RTX Ada under my desk.