Thesis · Control

Who sets the rules AI runs on?

The fight that decides how AI shapes your life is not about how capable the models get. It is about who writes the rules underneath them, and whether you are ever allowed to see those rules at all.

Most of the AI conversation is about capability: bigger models, higher scores, what the systems can now do. I think that is the less interesting half. Capability tells you how far a system can go. It does not tell you where it is pointed. The thing that decides where it is pointed is a set of rules, chosen by a small number of people, that the system then runs on for everyone else. That choice, not the raw power, is the one I would watch.

Why the rules matter more than the capability

Conway's Game of Life is the cleanest demonstration I know. Four tiny rules on a grid, and patterns emerge that nobody drew: shapes that walk, structures that pulse, whole worlds you did not design. Change one rule and the world changes with it. AI is the same shape at an enormous scale. A few people choose the objective, the data, the constraints, the refusals, and a system grows behaviour from those rules that reaches millions who never saw them. Capability is how big the grid is. The rules are what grows on it.

Is AI capability or AI governance the real contest?

Governance, by a distance, though the word undersells it. Capability compounds and spreads, and it is hard to hoard for long. The rules are different. Whoever sets them sets the defaults everyone downstream inherits, often invisibly, sometimes permanently. A model's values are not discovered, they are chosen, and then shipped at the scale of the deployment. So the honest question is never only how good is it. It is whose rules is it running, and did the people living under them get any say.

Why visibility is half the battle

A rule you can see is a rule you can contest. The danger is not only that a few people choose the rules, it is that the choice is usually buried, in weights, in a system prompt, in a policy nobody publishes, so the result arrives looking like a fact of nature rather than a decision someone made. Make the rule visible and it becomes arguable, auditable, reversible. Keep it hidden and it hardens quietly into infrastructure. A great deal of my work is just an argument for dragging that choice back into the open where it can be questioned.

Why I build on this

I did not arrive at this from a policy seminar. I have been turning it over since I was a child building cellular automata on a fixed canvas, watching a handful of rules grow a world. The lesson stuck. If you want to understand a system, do not stare at what it can do. Find the rules underneath it, and find out who gets to write them. That is the layer I work on, and the question in the title is the one I think more people should be asking out loud.