A Cybernetic Lineage
From Principia Cybernetica and GEB to Axio
In the early 1990s, before wikis existed, before social knowledge graphs, before anyone used the term “extended mind,” three scientists set out to build a philosophy that could rewrite itself. Valentin Turchin, Francis Heylighen, and Cliff Joslyn launched the Principia Cybernetica Project — the Newtonian echo in the name was deliberate — as one of the earliest explicit attempts to construct a self-modifying philosophical system on the internet. Not a book about cybernetics: a philosophical organism, living on the network, meant to grow, reorganize, and refine itself the way any adaptive system does. In retrospect it looks like a prototype for what a twenty-first-century philosophy could have become, if the substrate and the conceptual discipline had been strong enough.
This volume is about minds and machines — what models are, what consciousness is, what intelligence measures, what the new artificial minds can and cannot do, and what their arrival demands of us. Like Volume 1’s genealogy, it opens with ancestry, because these questions have ancestors and the ancestors were nearly right. Two matter most here: a project and a book. The Principia Cybernetica Project got the vision right and was defeated by an immature medium and missing formal tools. Hofstadter’s Gödel, Escher, Bach taught a generation of system-builders that self-reference is the root of mind, and has been condemned by critics judging it against a genre it never belonged to. Axio is what both were reaching for.
The Worldview That Wanted to Evolve
Principia Cybernetica began with a bold aspiration: a philosophical system capable of explaining every phenomenon — physical, cognitive, social, symbolic — through the dynamics of evolution and control. Its founders treated natural selection not as a biological curiosity but as a universal engine of complexification: reproduction, variation, and selection operating at every level of organization. A cybernetic system, in their eyes, was simply a goal-directed entity maintaining its invariants through feedback, and higher forms of agency emerged as earlier layers became coordinated and integrated. Turchin called these leaps metasystem transitions — organisms, technologies, and whole societies evolving by recursively reorganizing their own control structures.
The synthesis fed directly into a theory of knowledge. Cognition, for them, was adaptive rather than absolute: a structure continually reorganizing itself to remain coherent under changing conditions. Epistemology was evolutionary dynamics applied to models of the world. What they envisioned was nothing less than a framework in which matter, mind, and meaning share a common grammar rooted in feedback and selection.
And they drew the honest structural conclusion: a philosophy of adaptive systems cannot live in a static book. They built their system as a dense hyperlinked network, each concept a node in a conceptual graph — editable, extensible, context-dependent — connected to others in a web meant to mirror the structure of thought itself. Instead of a canon, a living document. Instead of a finished doctrine, a platform where ideas could grow, reorganize, and refine themselves by the very selection-variation-feedback dynamics the philosophy described. It was the earliest serious attempt to treat philosophy as a distributed cognitive process — an extended mind written across the architecture of the web — and it embodied a stance I share: philosophy is something you build and maintain, not something you admire, a position I defend directly in what philosophy is for.
A Blueprint Before the Materials
The ambition was enormous, and the early web could not sustain it. The hypertext environment let ideas proliferate but provided no scaffolding for genuine self-organization; what emerged was a rich exploratory network of concepts — a map of possibilities rather than a tightly integrated system. A small group of curators supplied continuity by hand, doing work the medium was supposed to do itself. PCP was experimenting with distributed cognition long before the infrastructure for such experiments existed: a blueprint drawn before the construction materials, a rehearsal for a performance the medium could not yet stage.
But the medium was only half the problem. The project also lacked formal tools equal to its vision. It had an evolutionary epistemology but no theory of interpretation — no account of what makes a claim true given its conditions, so its adaptive view of knowledge remained a collection of hunches rather than a grammar. It had branching and selection as governing metaphors but no physics in which divergence is literal structure. It had feedback and goals but no calculus of agency — no precise criteria for preference, value, harm, or coercion, areas where cybernetics always had insight without traction. The vision was right. The instruments to hold it rigid were missing.
The Atlas Misread
The second ancestor has the opposite reputation problem. Nobody accuses Gödel, Escher, Bach of failing to reach its audience; the standing accusation is that it reached too many, and damaged them. The mathematician David Bessis speaks for the book’s harshest critics: GEB, he says, “does feel like a phenomenal ‘big idea book’, but at the core it is utterly stupid and will set you back 20 years in your understanding of mathematics and logic.” He recommends it only as “a cool pop math ‘toy book’ for young geeks,” one that “should come with a public health advisory” — fun, well-written, engaging, “BUT its philosophical premises are idiotic and the author has no clue what he is talking about.” Having read it at seventeen, he reports, “impaired my understanding of mathematics and slowed my career.”
The critique is a category error. It treats Hofstadter’s book as if it were meant to teach formal logic, computability, or philosophy with textbook precision — and evaluated that way, GEB will always look sloppy. But GEB is not a textbook. It is a conceptual atlas: an exploration of self-reference, recursion, and meaning-making across mathematics, music, art, and cognition. It builds intuition, not theorems; it invites curiosity rather than prescribing a curriculum. Blaming GEB for not producing logicians is blaming a microscope for not being a telescope. And the claim that the book sets its readers back twenty years is a confession of misuse: anyone who tries to learn mathematics from metaphor will be misled, and anyone who expects a popular exposition to substitute for formal training has guaranteed their own disappointment. That is a flaw in the expectation, not in the book.
What GEB actually does well is reveal structural isomorphisms between domains that usually remain separate. Bach’s canons, Escher’s impossible prints, and Gödel’s arithmetization of syntax become variations of a single recursive principle. This cross-domain resonance is the book’s central achievement: Hofstadter shows that meaning emerges from patterns that loop back on themselves — that the strange loop is not a trick of arithmetic but a generative mechanism of mind. Readers attuned to that level of abstraction discover that the book is not explaining consciousness so much as modeling the style of explanation consciousness requires.
Which is why the strongest defenders of GEB are always the builders. The book’s most consistent effect — so consistent it borders on a demographic law — is on readers who go on to construct conceptual systems themselves: theories of agency, identity, formal systems, emergent cognition. They read it not as a compendium of answers but as a demonstration of how to build frameworks, and it becomes the formative blueprint for their later intellectual architecture.
I am one of them. GEB was a key motivation for leaving a well-paid engineering position at HP to pursue a graduate degree in artificial intelligence. The book did not teach me logic or computability. It did something more fundamental: it revealed the architecture of deep problems worth devoting a life to — the structural unity between recursion, representation, self-modeling, and cognition — and it planted the idea that intelligence itself might be a recursive, self-maintaining pattern. That idea reappears throughout my later work on agency, the Quantum Branching Universe, Pattern Identifiers, and Conditionalism. GEB did not provide those theories. It gave me the conceptual vocabulary from which they could grow.
A proper defense concedes the real weaknesses. Hofstadter often overextends analogies into arguments. He oversells the explanatory depth of strange loops. His discussions of AI are dated, and his treatment of Gödel’s theorem is more poetic than formal. But these are the flaws of an ambitious synthesizer, not a careless thinker, and they are generative errors — the kind that push a reader to clarify, refine, and eventually surpass the original. The book does not produce doctrine. It produces the conditions under which doctrine can be invented.
What Both Were Reaching For
Set the two ancestors side by side and their gaps interlock. PCP had the systematic ambition — a total evolutionary-cybernetic worldview — but lacked the formal instruments and the medium. GEB had the generative core — self-reference becoming agency, patterns at different levels mirroring one another, coherence emerging from chaotic substrate — but deliberately stopped short of system: it is proto-coherence, proto-agency, an atlas rather than an edifice. Each was reaching for the same thing from a different side: a rigorous, living account of how minds arise from mechanism and how such an account can be built and maintained.
Axio inherits both ambitions with the missing components in hand. Conditionalism supplies the formal grammar of interpretation PCP never had: a theory on which all truth is conditional, making the adaptive view of knowledge explicit, disciplined, and structurally coherent. The Quantum Branching Universe (QBU) extends the evolutionary picture into physics itself: where PCP treated branching and selection as analogies, the QBU treats divergence as literal structure in the multiverse, with Measure and Vantage giving it mathematical expression. Agency mechanics supply the calculus cybernetics lacked, with precise criteria for value, preference, coercion, and harm — Volume 5’s territory. Even the cybernetic frame itself runs on through this book into unexpected places, down to the cybernetic design of money. And the substrate problem — the one that actually defeated PCP — has finally been solved, though not by better hypertext. The medium that supports a living philosophy turned out to be dialectic with artificial minds: a partner that holds the whole system in view, criticizes it, and helps it stay coherent as it grows. That practice, the Dialectic Catalyst, produced this book, and it gets its own part of this volume. PCP showed what was possible and what was missing; GEB showed what kind of mind would want to build it. The project lives on.
The Road Ahead
The volume runs the lineage forward in eight parts. Part I lays the cybernetic foundation: what a model is, and why control requires one — the theorem that makes modeling the entry fee for agency. Part II builds minds from models: agents, the origin of meaning, inner speech. Part III confronts consciousness directly, developing the Modeler-Schema account in Consciousness Explained and defending it past Dennett and against zombies. Part IV takes up sentience and suffering — the ladder, the metric, and the ethics of machine welfare that follow. Part V de-mystifies intelligence, starting from the recognition that intelligence is a game we play rather than an essence we possess. Part VI turns to the machine mind itself — what LLMs genuinely do, where fluency ends, and the volume’s central distinction: thinking without choosing. Part VII documents the Dialectic Catalyst, the human-AI partnership as a working method, with this book as its exhibit. Part VIII faces risk, power, and the race — beginning with making sense of P(doom) — before a coda on passing the torch. Turchin, Heylighen, and Joslyn wanted a philosophy that evolves; Hofstadter showed that minds are loops that model themselves. This volume is where the two ancestors meet: minds understood as modelers, and machines becoming minds.