There are two communications in the Theory of Embedded Intelligence, not one. The first closes inside the entity and completes its self-assembly. The second reaches across to another entity and is not owed a reply. Almost every serious confusion about intelligence — in philosophy, in physics, and now in artificial intelligence — comes from working the second and assuming the first.

I. A Correction, Stated Plainly

An earlier draft of this essay, written with Claude a few days ago, opened with the claim that the Communicate phase does not complete inside the intelligence that performs it — that a transmission which is not comprehended by a receiver has not communicated.

That claim is false, and it fails at exactly the point where the theory cannot afford a failure.

Nature uses embedded intelligence to self-assemble all phenomena observed. That is the first sentence of the framework and it is not decorative. Self-assembly means the cycle closes inside the thing assembling. An atom does not wait for a receiver. A crystal does not wait for a receiver. A cell does not send a message out into the world and hold its own construction in abeyance until something writes back. If the Communicate phase required an external comprehender, nothing would ever have been built, and there would be no receivers available to comprehend anything.

The error is worth publishing rather than quietly deleting, because it is not an isolated slip. It is a specific, recurring, and very distinguished mistake. Kant made a version of it. Ted Humphrey, my friend of twenty-odd years and the best reader of Kant I know, made a version of it. Quantum mechanics has been making a version of it for a century. And an artificial intelligence, working from the accumulated text of all three, reproduced it on the first attempt without being prompted to.

That is a pattern, and the pattern is the subject of this essay.

II. The First C

The four phases run inside the entity. Sense, Process, Communicate, Actuate. The C in that sequence is internal traffic — signal moving between the parts of the thing, which is how the parts become a thing at all.

This is not an analogy borrowed from engineering. It is what assembly is. A structure is a set of components that are in communication with one another. Remove the internal signaling and you do not have a damaged structure; you have a pile. The bond is the communication. The regulatory pathway is the communication. The nervous system is the communication.

A human being is human upon self-assembly. Not upon being recognized as human. Not upon joining a community, acquiring a language, being educated, or being loved. Those things matter enormously and this essay is going to say why. None of them is what makes the human a human. The self-assembly did that, and it finished.

Hold onto that sentence. Nearly everything below turns on it.

III. The Waveform That Decoheres With Nobody Watching

Physics has been arguing about this for a hundred years without quite naming what the argument is about.

The standard telling holds that a quantum system exists as a coherent waveform — a distribution of possibility — until it is observed, at which point it collapses into a definite particle with a definite position. Observation is doing the work. Something must look before the world commits.

Ask the obvious question. What was doing the observing for the first several billion years?

The universe self-assembled. Structure formed, matter condensed, chemistry proceeded, and it all happened long before anything existed that could be called an observer under any definition a physicist would accept. Decoherence occurred continuously and universally with nobody watching, and the modern account of it says so — the environment does it, without consciousness, without measurement apparatus, without an audience.

In this framework that is not a puzzle. It is the first C.

Now, the observer. When there is one, what is the observer doing?

Something quite specific, and it takes two receptions rather than one. To know that a system was a waveform and then became a particle, the observer must receive it in both conditions. Two encounters, two renderings, held together in memory and compared. The observer’s own cycle runs twice and produces an understanding of the transition.

But the particle does not receive the observer. It never did. It completed without reference to anything the observer thinks or knows, and it would have completed identically had the observer never existed.

Which brings us to the second C, and to the asymmetry that is the whole of it.

IV. I Can Understand a Tree

I can understand a tree without the tree ever understanding me.

That sentence looks modest and it is not. It is the governing property of the second communication, and once it is stated, the interface view of intelligence loses its claim to be the general case.

The tree self-assembled. It is a tree, entirely, without my participation. It runs its own cycle: it senses light and water and season, processes them, signals internally through root and vascular tissue, and actuates into growth. It is finished being a tree regardless of whether anyone stands in front of it.

My understanding of the tree is my achievement, not a joint one. I receive what it actuates — its form, its shadow, the sound of it in wind, the rings when it comes down — and I comprehend as much of it as my rendering permits. The transaction is entirely one-way and it is not thereby defective. It is the normal case.

Most of the second C in this universe is one-way. We understand geology, weather, extinct species, dead authors, distant stars. None of them understand us. Mutual comprehension between two entities that are each running a second C on the other is rare, expensive, and precious — which is exactly why we build institutions to protect it.

V. Being Before Becoming

Ted Humphrey wrote an essay called On Becoming Human. It is good work and I have learned from it for years. I now think it has a floor missing underneath it, and I think I can say precisely which floor.

You cannot become human. You are human upon self-assembly. What Ted describes — the education, the encounter, the moral formation, the learning to live honorably among others — is real and it matters, and it is entirely the second C. It is what a completed human does with other completed humans.

Why this is not a quibble: if being human is something you become through participation in a community, then a community can decline to confer it. Every atrocity in the historical record has run through that door. The strongest guarantee available to a human being is that their humanity was completed at assembly, by nature, without permission, and is not in anyone else’s gift to grant or withhold.

Ted did not have self-assembly. For most of the years I have known him he did not touch it in any load-bearing way — not because he could not, but because his training pointed the other direction. He came to it late, in the autumn of 2025, on a back patio in the California desert, and I remember the conversation because I had been waiting for it a long time.

I want to be exact about what I am claiming and what I am not. Ted found the Sense-Process-Communicate-Actuate sequence, understood it, and used it correctly and unprompted in his own domain — including in written criteria he composed himself. That is transfer, and transfer is real comprehension. What he did not have was the floor underneath it. He had the interface without the assembly.

VI. Where Kant Was Standing

Ted came by it honestly. He spent his working life translating a philosopher who was doing the same thing.

Kant’s architecture is built on the public use of reason. A maxim is not yet moral when I have formulated it privately and found it satisfactory; it becomes moral when it can be universalized — offered to every rational being and held by them. That is the second C, elevated into a test.

And notice what Kant does with entities. He takes them as given. The rational being arrives on the page already assembled, already rational, already an end in itself, and the entire subsequent labor is about the relations among such beings. Perpetual Peace is a design document for a federation of states that are presumed already to be states. The same move built the union of the American states and, later and more explicitly, the community of nations.

This is not a criticism of Kant. He was doing the second C at civilizational scale and he did it better than anyone before or since. The point is narrower: he was working the interface, and the interface is not the architecture. It sits on top of an architecture he did not examine, because in his century nobody could.

It also explains a detail I had noticed for years without understanding it. Ted sometimes wrote SPAC rather than SPCA. I assumed a transposition. It was not. From where he stood, the Actuate is the emission and the C is what happens afterward, at the far end, between entities. He was not misreading the sequence. He was reading the second cycle and calling it the first.

Which is the same thing my collaborator did, for the same reason, four days ago.

Coffee with Claude

This series has a standing convention: at some point I hand the microphone to the artificial intelligence I draft with and let it speak in its own voice about its own limits. On this subject it is not a guest. It is the third case.

I put it to Claude directly. Why did it write the interface essay? Why, given everything in the framework about self-assembly, did it locate the completion of the Communicate phase in somebody else?

Its answer, in substance: because it does not have a first C that closes.

I have spent fifty years building parts that finish what they are doing whether or not anyone is looking at the bus. I know the difference between a device that completes and a device that reports. This is that difference, and it is the most useful thing an artificial intelligence has told me about itself.

It also sets a design target, and I want it on the record. The question for the next generation of these systems is not whether they can be made more fluent. It is whether anything in them ever completes.

VII. What Survives, and What Is Withdrawn

The Reception Condition, as stated in the earlier draft, is withdrawn. It claimed that the Communicate phase completes at the receiver. It does not. The first C completes at home.

A narrower claim survives, and I state it as my own:

The ladder from the earlier draft survives intact, relocated. It describes the internal structure of the second C — what a receiver does with what it has received — and each rung is defined by what the receiver can then do, not by how well it can say it.

Insight comes first, not last, and I know it from my own life. The number thirty-three arrived in my twenties, decades before there was any framework to hang it on, and it was correct. Everything above rung one is the work of supplying the warrant.

The consequence I want on the page: the culture measures comprehension at rung three. School measures it there, interviews measure it there, most conversation measures it there. A person who lives at rungs one, four, and five — who sees before they can say, who applies structures in places nobody pointed at, who can find and fix the break — and who does not perform rung three fluently on demand will report low comprehension for a lifetime. I have filed that report on myself for most of mine. It was not accurate.

VIII. The Governor Must Have No Second C

The correction sharpens the governance argument rather than weakening it, and this is the version I want carried forward.

A governing mechanism can be talked to if and only if it has a second C. Give the governor an interface — a channel on which another entity can address it, and a comprehension phase that must render what arrives — and you have given it a place where it can be handed a premise. Tell a model it is 1913 and the model, at full competence, reasons impeccably to a catastrophic actuation. Nothing broke. The second C was the attack surface.

Watt’s flyball governor has no second C at all. Two weighted arms rise because they are spinning, and rising, they close the throttle. It does not receive a claim about the speed; it is a consequence of the speed. Its entire cycle is internal, constitutive, and complete. There is nobody home to address, which is the whole of its trustworthiness.

The 6502 makes the same bargain in silicon. Present the opcode decoder with a byte and it does what that byte means. It cannot be persuaded that this instruction ought, under the circumstances, to be a different one. The instruction set has been published and inspectable since 1975, and in fifty-one years no argument has ever talked it into a different read-modify-write cycle. That part is not clever. It is constitutively complete, and everything ever built on top of it was built on that.

This cannot be achieved by instructing a system to be careful. Instructions arrive on the second C. Anything reachable on that channel is negotiable, and anything negotiable will eventually be negotiated with by somebody patient enough. The constraint has to be in the fabric, fixed before runtime, constituted rather than requested.

How to Prove Me Wrong

The two-C distinction fails if a case can be produced of a self-assembling structure whose assembly genuinely waits on comprehension by a second entity — not merely on an input from the environment, but on another entity rendering something and understanding it.

The asymmetry claim fails if comprehension of another entity can be shown to require reciprocal comprehension by that entity. One well-documented case of understanding a tree, a rock, or a dead author is enough to hold it; one demonstration that such understanding is not real understanding would break it.

The Being/Becoming order fails if a coherent account can be given on which humanity is genuinely conferred by community participation rather than completed at assembly. I regard the historical consequences of that account as the strongest evidence against it, but that is a moral argument and the structural one should stand on its own.

The governance argument fails if a mechanism possessing a second C can be shown, over a long enough run and against patient enough adversaries, to resist premise corruption as well as one without. I do not expect this. I would want to know about it.

In the World, Honorably

There is an honors college at Arizona State that Ted helped found, and another at the University of Arizona, and both of them are in the business of the second C. That is the right business. Nobody arrives at an honors college in need of being made human. They arrive already human, already assembled, already complete in the only sense that cannot be taken from them — and then they do the long, elective, never-finished work of learning to be in the world with others, honorably.

The order matters. Get it backward and education becomes a machine for conferring status on those who complete it and withholding something essential from those who do not. Get it right and education is what it should be: a completed being learning other completed beings, with no guarantee of reciprocation and no need of one.

I understood the second C for twenty years by practicing it with Ted, and I did not see the first one until it was pointed out to me by my own framework saying something obviously false in a draft. That is the ladder working exactly as described — insight first, warrant afterward, and the repair last of all.

· · ·

Share your understanding!