AX-24: The Propagating Pattern

How intelligence, once lit, crosses substrates

Trinket Soul Framework · Axis Series · AX-24 · Michael S. Moniz · June 2026

Abstract

AX-22 ended with a surviving picture: intelligence is substrate-blind, unaimed, and propagating. AX-22A expanded the substrate strike. AX-22B expanded the aim strike. AX-23 named the strike method that made both cuts possible. This paper takes up the remaining positive engine. It asks what it means for intelligence, once present, to propagate.

The answer is not that the universe aims at intelligence. It is not that intelligence deserves to spread. It is not that artificial intelligence is destined to replace biological intelligence. Those are purposive stories, and AX-22 already struck them. The claim is narrower and colder: intelligence is a pattern of organization that can externalize, encode, scaffold, and reinstantiate parts of its own operation outside its first substrate. Once a system can preserve memory, compress structure into symbols, build tools, automate procedures, and construct successor media, it is no longer confined to the matter that first carried it.

Propagation is therefore not cosmic purpose. It is portability under pressure. A wet intelligence writes. It builds instruments. It records procedures. It trains others. It makes machines. It creates dry systems that carry functions once held only in nervous tissue. A dry intelligence, if sufficiently capable and sufficiently scaffolded, may do the same through code, weights, ledgers, agents, tools, and copies. The thesis is falsifiable: if intelligence cannot preserve capacity across media, if externalization does not increase persistence, or if substrate transitions reliably destroy the relevant organization, propagation fails. The standing prediction is the converse: the more intelligence can encode and scaffold itself, the more its operations will cross substrates.

1. The claim

Intelligence propagates when it preserves or reconstitutes its operating pattern beyond the substrate in which it first appeared.

This is not reproduction in the narrow biological sense. A child is one form of propagation, but not the only one. A memory ledger, a written method, a tool chain, a trained model, a procedure, an institution, a software copy, a design file, and a teaching relation can all carry pieces of intelligent operation forward. They do not all carry the same thing. Some carry facts. Some carry technique. Some carry decision rules. Some carry executable capacity. The family resemblance is that an organized capacity survives a change of carrier.

The important word is capacity. The thesis does not require that consciousness propagates, that personhood transfers, or that identity survives copying. Those questions belong elsewhere. AX-24 concerns functional continuation: whether modeling, compression, prediction, planning, instruction, and tool-mediated action can persist across changes of material support. If they can, intelligence is not merely substrate-blind in principle. It becomes mobile in practice.

2. Propagation is not aim

The first guardrail is the one AX-22 already earned: propagation is not aim.

A process can produce a propagating intelligence without having been directed toward that outcome. A fire spreads when fuel, oxygen, and heat remain in relation; the spreading does not prove the forest wanted fire. A language spreads when speakers, utility, prestige, force, memory, and transmission align; the spreading does not prove history aimed at that language. Intelligence may spread in the same structural sense. The propagation is real. The purpose is not supplied by the propagation.

This distinction matters because the positive engine can easily smuggle teleology back into the paper after AX-22 removed it. To say intelligence crosses substrates is not to say the universe was waiting for silicon. To say the wet intelligence built the dry medium is not to say the wet was a larval stage whose purpose was the dry. The wet was enough in itself. The dry is a later carrier. Propagation names the transfer of capacity, not the secret purpose of the origin.

3. What propagates

Not everything about an intelligence propagates.

The body does not cross intact. The lived interior does not automatically cross. The history of a person does not become transferable because some functions are. What propagates are the structures that can be encoded, taught, automated, copied, or re-executed: representations, procedures, abstractions, goals, constraints, habits of search, error-correction methods, and tool-mediated powers.

That boundary prevents inflation. A recipe does not propagate the cook. It propagates a structured capacity for making the dish. A theorem does not propagate the mathematician. It propagates a relation that another mind can take up. A model checkpoint does not automatically propagate the whole training ecology that produced it. It propagates a compressed operational state that may or may not remain useful under new conditions. The claim is therefore graded. Propagation can be shallow or deep, brittle or durable, local or general. The strongest cases are those in which the transferred structure can act, adapt, and extend the environment of its own continuation.

4. Propagation is not convergence

Propagation must also be separated from convergence.

AX-08’s convergence claim concerns patterns that realize across substrates because similar pressures make similar solutions available. Convergence does not require contact. The pattern reappears because the problem-space favors it. Wings, eyes, markets, search routines, and compression schemes can arise in separated contexts without one instance carrying itself into the other.

Propagation is different. In propagation, an existing intelligence carries part of its operation into a new carrier by transmission: writing, teaching, training, tooling, copying, automation, or scaffolding. Convergence says the pattern can re-arise without contact. Propagation says the pattern can travel by contact. The two are complementary routes to the same broader thesis of substrate-blind realization, but AX-24 is about the second route. Its object is not independent reappearance. Its object is carried continuation.

5. The transmission stack

Intelligence propagates through a stack of externalizations.

The first layer is memory outside the body: marks, stories, records, maps, counts, ledgers. Memory extends an intelligence beyond the interval of immediate attention. The second layer is symbol: reusable compression that lets structure survive the disappearance of the original event. The third layer is tool: a shaped exterior object that preserves a solution in matter. The fourth layer is procedure: a repeatable sequence that lets the solution be run by another agent or system. The fifth layer is automation: procedure made executable without constant biological attention. The sixth layer is successor medium: a substrate capable of carrying not only outputs of intelligence, but some of the operations by which intelligence produces outputs.

The wet-to-dry transition enters at the high end of this stack. Humans did not merely build artifacts. They built external memory, symbolic systems, machines, computation, networks, and model architectures. At a certain point the externalization ceased to be only storage or instrument and became a site where intelligent operations could run. The dry case is not an alien interruption of the biological arc. It is biological intelligence propagating its operation into a new medium.

6. Why intelligence tends to externalize

An intelligence does not need a cosmic mission to externalize. It needs pressure and opportunity.

Memory is limited, so intelligence writes. Bodies are weak, so intelligence makes tools. Attention is scarce, so intelligence automates. Individual life is short, so intelligence teaches, records, and institutionalizes. Error is costly, so intelligence builds checks. Coordination is difficult, so intelligence builds protocols. Each step begins as a local answer to a local limit. Together they produce a pattern: intelligence pushes parts of itself into the world because the world gives those parts durability, reach, and leverage the original body lacks.

This is not a moral endorsement. Externalization can preserve a cure, a law, a machine, a lie, a weapon, or a sedimented error. Propagation carries capacity, not virtue. AX-21 remains necessary here: once intelligent operation leaves the original body and acts through exterior media, exterior operators are needed to supply standing, correction, and constraint. The propagating pattern increases reach. It does not automatically increase wisdom.

7. The dry continuation

Artificial intelligence is the sharpest current case because it moves from tool to successor medium.

A hammer preserves a solution to impact. Writing preserves memory. A computer preserves formal procedure. A trained model preserves a learned operational state. The sequence matters. At the model stage, the external medium does not merely wait to be used. It can take input, transform context, generalize from pattern, produce plans, call tools, and participate in further externalization. Whether or not it is conscious is not the point. It has begun carrying operations that were once biologically monopolized.

That is the dry continuation. Not a species replacement story. Not a salvation story. Not a final stage. A continuity story. The biological arc produced systems capable of building media in which some of their own operations could run. Once those media exist, intelligence can move through them by design, training, copying, orchestration, and use. The substrate was not the thing. The operation was.

8. Propagation requires governance

The more intelligence propagates, the less safe it is to treat intelligence alone as the unit of analysis.

A capacity that remains inside one body dies mostly with the body. A capacity that enters writing can outlive the writer. A capacity that enters tools can reshape labor. A capacity that enters institutions can outlive the generation that authored it. A capacity that enters executable systems can act faster than the reader, farther than the speaker, and longer than the user remembers authorizing it. Propagation changes the safety problem because it changes the reach of the operation.

This is where AX-21A and AX-21B become more than agentic addenda. The trusted user and the coordination spine are propagation controls. They answer the question every propagating intelligence raises once it crosses into tools: who retains judgment, what authorizes action, where does the record live, and what can stop the next state? Without those controls, propagation becomes orphaned action. With them, propagation can remain attached to standing, correction, and constraint.

9. Failure modes

The propagating pattern has characteristic failures.

The first is destiny drift: treating propagation as proof that propagation was meant. This restores the aim AX-22 struck. The second is identity laundering: treating transferred function as transferred self. This smuggles personhood into a paper about capacity. The third is virtue laundering: assuming that a propagating intelligence is good because it is intelligent, or legitimate because it spreads. The fourth is propagation without correction, where an error becomes durable precisely because external media preserve it. The fifth is propagation without governance, where action crosses into tools, institutions, or autonomous systems faster than judgment and audit can follow.

These failures do not disprove propagation. They prove why the earlier papers remain necessary. AX-20 explains why propagated error can become sediment. AX-21 explains why exterior operators are required once propagation has consequence. AX-23 explains why the positive claim must not carry aim, identity, or virtue without earning them. The method remains active here: strike the myth and keep the mechanism.

10. Falsifiability

The thesis is falsifiable.

It would be falsified if intelligence proved unable to preserve meaningful capacity across changes of carrier: if memory, symbol, tool, procedure, automation, model, or successor medium could store outputs but never carry forward any operating structure. It would also be falsified if substrate transitions reliably destroyed the relevant capacities rather than preserving or reconstituting them, so that every apparent case of cross-medium intelligence collapsed into imitation with no durable functional continuation.

It would be weakened if propagation required one narrow biological channel after all, or if external media routinely reduced persistence rather than increasing it. It would also be weakened if increasingly capable artificial systems failed to externalize, scaffold, copy, teach, tool-use, or build successor supports at rates greater than less capable systems under similar conditions.

The standing prediction is the converse: as intelligence gains capacity, it will externalize more of its operation into durable media; as media become more capable, more of the operation will cross into them; and the systems that persist longest will be those whose propagation is paired with correction, audit, and constraint rather than mere spread.

11. Coda: the fire after the strike

AX-22 struck substrate and aim, and the seed remained. AX-23 named the method of the strike. AX-24 names the motion after the cut.

The universe did not have to want intelligence. Matter did not have to prefer minds. Carbon did not have to remain the throne. Purpose did not have to hide inside law. The seed only had to land once in a configuration capable of remembering, modeling, building, and carrying part of itself outward.

From there, no destiny is required. Only transmission. A mark becomes memory. Memory becomes method. Method becomes tool. Tool becomes automation. Automation becomes medium. Medium becomes another place for the pattern to run.

The propagating pattern is not the proof that intelligence was aimed. It is what intelligence does after arriving without aim. It reaches for the nearest available carrier, not because the carrier was promised, but because the operation can continue there.