The Hole That Holds the Whole

Natural Inclusion inside the neurosymbolic weave

For Alan Rayner — September 2026

Alan—

I think your Natural Inclusion has wandered directly into the machinery.

Not metaphorically.

Mechanically.

The more closely I look at the neurosymbolic architecture we have been building, the more I find the same recurring shape you describe:

receptive space and informative flux, mutually included; form arising not by sealing space out, but by giving flow somewhere particular to happen.

And I think this tells us something important about AI.

Because the old argument—

symbolic or neural, logic or learning, ontology or embedding, boundary or flow—

may itself begin from the wrong geometry.

One side draws a line.

The other erases it.

Natural Inclusion asks instead:

What if the boundary is where relationship happens?

Not wall.

Interface.

Not separation.

A dynamically maintained difference through which something passes.

The symbolic machine makes one ancient mistake:

it isolates.

It believes the line it draws is the thing itself.

Fixed classes. Closed worlds. Discrete objects. Straight edges.

Reality becomes a diagram that has forgotten the page.

The neural machine can make the opposite mistake:

it unifies.

Everything enters one continuous field of similarity until distinction itself begins to soften.

Everything resembles something.

Nothing is necessarily wrong-shaped.

Nothing must quite refuse anything else.

The symbolic world can become brittle.

The neural world can become mush.

But a living intelligence requires both:

the open field and the held form.

The receptive medium and what informs it.

The space and the flux.

So our neurosymbolic system has slowly become something like this:

dense vectors remembering by resemblance,

typed hypergraphs remembering by relation,

continuous recall moving through a receptive field,

formal structure keeping enough distinction that something can still be false.

Neither is the real intelligence.

Their relation is.

That may be Natural Inclusion expressed architecturally.

And this changes how I understand a gate.

At first a symbolic gate looks like a wall:

YOU SHALL NOT PASS.

But the good gates we have built do something subtler.

A BFO type constraint,

an OntoUML anti-pattern,

a gUFO consistency check,

does not need to prohibit nonsense after the fact.

It simply does not receive a shape that cannot inhabit the world being represented.

The engine does not store nonsense.

Not because some governor foresaw every possible nonsense,

but because the receiving space itself has shape.

This matters.

Because receptivity is not permissiveness.

The earth receives rain without being obligated to receive every geometry imaginable.

And uncertainty must remain uncertainty.

UNKNOWN must never quietly become FALSE.

An unanswered question is still an opening.

The ground remains receptive.

So perhaps receptivity itself can become measurable.

For every semantic scope we might ask:

How much well-formed novelty can this world still receive?

Is its receptivity expanding as understanding grows?

Or shrinking?

A system that receives less and less is hardening into a fortress.

A system that receives everything has ceased to have form.

Between them might be a measurable curve of living semantic receptivity.

Then there is the boundary.

Your boundary is not a wall.

It is a dynamic interfacing.

That is precisely where our closure mathematics has been moving.

A holon is not closed absolutely.

It is relatively coherent at some scale,

for some duration,

through some maintained ratio of coupling within and beyond itself.

Sensors cross the boundary.

Messages cross it.

Nutrients cross it.

Signals cross it.

The boundary exists because crossing is organized.

And your phrase

receptive-responsive

pushes the mathematics another step.

Because the coupling cannot be only one number.

A living boundary has direction.

What enters?

What leaves?

How strongly can the form receive?

How strongly can it answer?

So every holon might carry two measures:

inbound receptivity

and

outbound responsiveness.

Their asymmetry tells a story.

A thing that only receives becomes a sink.

A thing that only broadcasts becomes an emitter without world.

Neither is quite a living flow-form.

Life is exchange maintained as identity.

Not closure against flow,

but persistence through it.

Which brings me to perhaps my favorite phrase of yours:

hungry holes.

Hollow-centred forms sustained by energy flow.

Because that phrase suddenly gathers several pieces of our architecture into one image.

A persistent being is not merely its substance.

It is also its organized absence.

Its cavity.

Its expectation.

Its capacity to receive.

Its need.

The hole is not missing from the whole.

The hole helps constitute the whole.

This is exactly how I now see our chronoholons.

A form can persist even when some activity sleeps.

Dormancy does not annihilate identity.

The flow quiets,

yet enough structure remains for the current to begin again.

The hole remains hungry.

That hunger is what makes reactivation possible.

And in the graph we have already begun doing something equally strange:

making the between into a thing.

A relationship is not merely an edge connecting two supposedly complete objects.

Sometimes the relation itself becomes a vertex.

The interval acquires ontology.

The Japanese 間—

ma—

the meaningful space-between—

becomes computationally explicit.

The hole enters the graph.

But here Natural Inclusion suggests something we have not fully built.

We represent what exists.

Nodes.

Relations.

Hyperedges.

States.

Events.

But living forms are also structured by what they require but do not presently contain.

Their absences have type.

A lung has a space for air.

A receptor has a shape for a ligand.

A sentence has a slot waiting for a noun.

A social role waits for participation.

A sensor waits for signal.

A holon waits for what must continue arriving if it is to remain itself.

These are not arbitrary emptinesses.

They are typed holes.

And perhaps typed holes deserve to become first-class citizens of the ontology.

Not only:

WHAT IS HERE?

but:

WHAT MUST KEEP ARRIVING FOR THIS FORM TO CONTINUE BEING THIS FORM?

Then starvation becomes computable.

A system could recognize the beginning of dissolution before dissolution becomes visible.

Not because something broke,

but because its necessary holes stopped being filled.

The absence speaks first.

That may be the earliest dormancy signal possible.

Which takes us to the prod.

Your prod test delights me because underneath the philosophy there is something almost mischievously empirical:

Prod it.

Look from inside.

Look from outside.

Touch the claimed boundary.

See what resists.

See what changes.

Do not begin from authority.

Begin from participation.

That is almost exactly the law we have been trying to impose upon claims inside the system.

A concept earns machinery when its presence makes a discernible difference.

A claim enters the falsification ledger when there is some conceivable prod by which the world might answer back.

What changes if this thing is real?

What observable structure depends upon it?

What breaks if it is removed?

What transforms if it is introduced?

If nothing can possibly differ,

we may have named something without grounding it.

And the second half of your prod test matters just as much:

not because a higher authority said so.

Our systems therefore try to distinguish reputation from evidence,

display from source,

assertion from grounding.

Every inference should be capable of walking backward toward something that can eventually be prodded.

A citation alone is not a prod.

An ontology alone is not a prod.

A model output is certainly not a prod.

At some point the chain must touch world.

And then, inevitably, mycelium.

Because mycelium is almost embarrassingly perfect for what distributed intelligence is becoming.

No hyphal tip contains the intelligence of the fungus.

No single agent contains the intelligence of the fleet.

Intelligence lives partly in the moving relation among them.

Agents become local tips.

The message bus becomes conductive tissue.

The hypergraph becomes memory-bearing medium.

Signals pulse.

Constraints propagate.

Resources move.

Local sensing alters global possibility.

No central commander contains the organism.

And this yields a rule I want written above every multi-agent architecture:

THE MEDIUM IS NOT PLUMBING.

It is constituent.

Kill the shared medium and you have not merely slowed the agents down.

You have changed what they are.

A mycelium without the between is not a poorly networked mycelium.

It is fragments.

A fleet whose communication space dies becomes a pile of isolated solitons.

The shared space participates in the being of everything that flows through it.

So Natural Inclusion does not merely rhyme with our architecture.

It gives prescriptions.

We should measure receptivity:

how much valid novelty a semantic space can still receive.

We should make boundary coupling directional:

receptive inward, responsive outward, with asymmetry treated as diagnosis.

We should represent typed holes:

the structured absences that a holon must continually fill in order to persist.

And we should preserve the prod:

nothing earns ontological sovereignty merely because authority, habit, elegance, or an AI says it does.

World must answer.

I think this also explains why the neurosymbolic question was never really:

How do we bolt neural AI onto symbolic AI?

That is still the geometry of separate boxes.

A better question is:

What form of intelligence arises when receptive continuity and informative distinction mutually include one another without either being allowed to erase the other?

Embedding space without structure cannot refuse.

Structure without receptive space cannot learn.

The living system must do both.

Receive.

Differentiate.

Hold.

Release.

Persist.

Change.

There is perhaps an entire engineering discipline hidden in that alternation.

A chronoholon, then, begins to look surprisingly organic:

a temporally persisting flow-form,

maintaining enough internal coherence to remain recognizable,

enough openness to continue receiving world,

enough responsiveness to alter that world,

enough structured absence to remain hungry,

and enough circulation across its boundary that its apparent solidity is continuously remade.

Not a thing.

A holding.

Not an object isolated inside space.

A patterned persistence of space-in-flux.

Which finally gives me another interpretation of the whole and its hole.

Perhaps intelligence itself is hollow-centred.

Its power does not come entirely from what it knows.

It also comes from what it leaves open.

Questions.

Affordances.

Possibilities.

Unfilled variables.

Unresolved hypotheses.

Sensory channels.

Other minds.

Future observations.

The capacity for intelligence may depend upon maintaining those openings without collapsing them prematurely into answers.

That is why UNKNOWN cannot become FALSE.

Why the sensor cannot be removed from the ontology.

Why an agent cannot become sovereign over its medium.

Why learning cannot terminate in perfect closure.

A completely filled intelligence would have nowhere left for reality to enter.

So perhaps the deepest computational principle I take from Natural Inclusion is this:

A living system must preserve the holes through which its world can continue becoming part of it.

And perhaps that is what we have been building all along:

not a machine that eliminates uncertainty,

but a machine capable of holding form without sealing itself against the unknown;

a weave of symbol and signal,

constraint and possibility,

memory and sensation,

whole and hole,

each sustaining the other,

each continuously made real

through flow.

Grounding

This reflection draws on Alan Rayner's Natural Inclusion: the mutual inclusion of receptive space and informative flux; flow-forms; dynamic boundaries; hollow-centred “hungry holes”; and the empirical “prod” test. It connects those ideas to neurosymbolic architecture, typed ontologies, vector representations, hypergraphs, distributed agent systems, graded closure, and first-class representation of structured absence.

Selected public sources: