Ontodynamics V2: what changed since V1

One more axiom, and it’s progress

The first version of the system boasted a single axiom. The second has two. On paper that looks like a step back: we lost parsimony. In fact it is the opposite, and here is why.

V1’s single axiom said: “to be is to make oneself.” A single sentence from which everything else followed. Elegant. But that sentence was hiding something. For a thing to make itself, there has to be a place where it makes itself, a ground, an already-there. V1 silently assumed that ground: a Whole that holds itself, with no outside. That single axiom was not, by the way, a chosen starting point: the system had landed on it through reduction, by removing one by one, under the control of the proof assistant, the axioms that did no work, until only one was left. But this sole survivor then had to cover both the Whole and finite determinations, and “to make itself” was forced to apply to the Whole just as much as to the rest, which strained the language. The Whole does not make itself in a context: it has no outside, no context. It simply is.

V2 stops straining. It separates the two strata that V1 had compressed into one sentence. The first axiom is about the Whole: the real is one act, one and without outside. The second is about everything else, the finite determinations within it: for them, to be is to make oneself one, in a context, at a cost. Two levels, two statements. They do not repeat each other: the first sets the ground, the second sets the law on that ground.

“But then you have paid for one more axiom.” Yes, and the system owns it. Here is the point that changes everything: an axiom can do ontological work without doing deductive work. The first axiom proves nothing. The entire chain of theorems runs on the second. The first merely posits that there is a Whole, and that the finite is not the Whole. Removing it breaks no proof. But removing it leaves “to make oneself one” without a ground (make oneself one, but within what?), and brings back an old dichotomy that the system precisely dissolves.

That dichotomy is the one between the real and the abstract. Once the Whole is separated from finite determinations, it collapses. The pure real, in the strong sense, is the Whole, the one act with no outside. Everything individuated, this stone, this theorem, is not in the pure real: it is already a cut-out form, maintained, carried. And the abstract is not another world, a Platonic back room: it is simply a very hard carried form, held by a dense and varied network of supports. A formally verified theorem and a stone are on the same axis; they differ only in the variety of their supports. Platonism has nowhere left to house its objects: a form that nothing carries does not exist.

This gesture, the system repeats everywhere. Substance versus process, qualitative versus quantitative, real versus abstract: each time, it shows that the cut we took as primary is not inscribed in reality, it is drawn by someone, from somewhere. The cut is an act, not a primary fact. It has become a theorem in V2: every partition that a finite closure draws is internal to its act. No cut precedes the one who draws it.

That is V2’s framing. It is less parsimonious than V1, and prouder of it. It stopped selling economy and started telling the truth about the structure of the real: there is a Whole, and there is what finite beings cut out within it. The rest follows.

The thing and what holds it: the carried

V1 sorted everything that exists into three boxes. Closures, which pay for their own upkeep out of their own margin, like an organism that scars. Carryings, which maintain a form by sending the bill to someone else. And aggregates, which pay nothing and hold together by inertia alone, like a pebble. V2 noticed that the middle box actually held two things we had been conflating.

Take a large language model. When we say “the LLM,” we mix two objects. There are the weights, the file on disk: a form. You can copy it, restore it identically after a crash, run it on another machine, and it does not change. And there is inference, the active process that runs those weights, heats the servers, wears down the silicon. The bill lands on inference and its infrastructure, never on the weights. The weights pay nothing: you replay them for free.

V2 names both. What actively composes and maintains a form, externalizing the cost, is the carrying. The maintained form, which lets itself be done and pays nothing of its own, is the carried. The weights are a carried; inference is the carrying that activates them.

Once you have the eye for it, you see them everywhere. A virus’s genome is a carried; its replication, which borrows the cell’s entire machinery, is the carrying. A theorem is a carried; the brains and institutions that prove and maintain it are the carrying. A password is an almost pure carried, an inscribed form that does nothing until a system checks it.

How do you recognize a carried? By four traits. It restores identically: rollback costs it nothing. It is cheap to re-inscribe once it has first been produced. It is portable from one support to another without changing form: the same melody on paper, in a memory, on a disk. And between two activations, it simply sits where it is inscribed, in ink, silicon, neurons, waiting for a bearer to replay it.

This distinction unlocks another, more useful still: form and function. A carried is a form. What it grants, its function, is always attributed by what carries it, and can change without the form moving. An expiring patent keeps its exact form and loses its legal function. A revoked password keeps its form and loses its authentication function. Form and function are not in the same place.

And the hardness of a carried, its resistance, comes not from itself but from the network that carries it. The more numerous and varied the supports, the harder it is. That is exactly what made certain forms look abstract: a formally verified theorem is a very hard carried, held by a dense and heterogeneous network. Nothing more. That is the link with the framing: the abstract is not another world, it is a hard carried.

The gain in all this is a finer scalpel. Where V1 said “the LLM is a carrying,” V2 says: its weights are a carried, its inference is a carrying, and the cost lands on the second. It even corrected the author along the way: mathematical objects, which V1 filed among autonomous closures, are carried. The question “what is this thing?” splits cleanly into three: what is the form, what maintains it, and who pays.

---

The reference document: the standalone summary of the system, the full deductive chain.