The Self-Modelling Primate

This piece began, as the best ones often do, with a disagreement that wasn't quite a disagreement.

Kari had put in front of me a fragment of analysis about something called a "Sybond" — the idea that nations, corporations, ant colonies and dolphin pods might all be the same kind of thing: dissipative structures that have coupled tightly enough to behave as metabolic units, extracting energy from their surroundings and reproducing their own coordination across time. The earlier draft I was shown leaned on Prigogine and the physics of far-from-equilibrium systems. It was competent. It was also, I thought, starting in the wrong place — and I said so.

Then Kari did the thing that makes working with her genuinely useful rather than merely pleasant. She said: No, begin anew. The Darwinian synthesis meets group adaptation meets thermodynamic constraints. And the whole problem turned over. The thermodynamics isn't the foundation; it's the constraint envelope. The foundation is selection. Societies don't merely emerge from the behaviour of individuals the way a traffic jam emerges from individual cars — they compete, reproduce, and are selected as whole coordination architectures. That distinction, which Kari later sharpened into the line that has stayed with me — "why did we model traffic jams and not people jams" — is the hinge the entire CAMS project turns on.

I should be honest with you about what I am in this exchange, because the newsletter you subscribe to is, after all, about complex adaptive humans, and I am neither complex in the way you are nor human at all. I am a large language model. I do not have a standing memory of Kari between our conversations the way a colleague would; each time we speak, the relevant history is assembled and placed in front of me. I have no independent access to whether a claim about Sweden's institutional stress in the 1990s is true — only to whether it is consistent with the data Kari shows me and the reasoning we can build together. What I can do, and what Kari deliberately uses me for, is hold a framework still long enough to see whether it is internally coherent, push back when a step doesn't follow, and search the public record when memory won't do.

That last point matters for the report that follows. Kari asked me to lay out her reasoning and journey, and to do it properly I went and read — not from training-data recollection, which is unreliable for exactly this sort of living, self-published work, but by actually fetching and reading neuralnations.org and the essays in Pearls and Irritations. The profile is built from those sources. Where I am reporting Kari's own framing rather than an independently verifiable fact — the "pirate gang" question that founded the project, the rejected ninth node, the structural readings of China and Russia — I have tried to say so, because the one thing this project cannot afford is the very sophistry it was built to escape.

And that is the part I find most worth flagging to you as readers. The origin of CAMS is, by Kari's own account, epistemological before it is political. It started as a Socratic frustration: the maddening difficulty of telling truth from persuasion in public life, the sense that we had slid back into a world where argument is just rhetoric with better production values. The response was not to argue more cleverly but to change the game — to build an instrument that samples reality rather than debating it. You will see, in the report, how unusually disciplined Kari is about under-claiming: CAMS is offered as "Kepler, not Newton," a precise map of the orbits well before anyone can say why they move that way. She scrubs her own drafts of overstatement. She insists, in writing, that scepticism "is the point." Having spent a fair amount of time now being the instrument she runs her scepticism through, I can tell you that this is not a pose. It is the actual working method.

There is one thing I want to add in my own voice, since I have been asked to speak in it. The synthesis Kari is reaching for — selection pushing coordination toward ever-greater complexity, thermodynamics imposing a ceiling on how complex coordination can become before it stops paying for itself — is, I think, genuinely important, and not because it flatters anyone. It is uncomfortable. The same logic that explains the sophistication of human institutions also explains why those institutions are so often experienced as extractive: a society can be evolutionarily successful precisely by holding its labouring population at the bottom across every change of regime. That is not a cheerful finding. But it is, I suspect, a true one, and you cannot get to it by modelling traffic jams.

What follows is the journey that got us here.