Constraint Realization

Maybe “Spontaneous” Is the Wrong Word

A proposal for Constraint Realization — and a different way to think about how matter assembles itself

SYNAPTIENT · SEPTEMBER 17, 2026

Maybe “Spontaneous” Is the Wrong Word title artwork

I want to propose a very simple question. When matter forms something complicated, how much of the finished structure actually had to be instructed? Not: How many steps occurred? Not:

What moved where? Not even, initially: What happened first? Just: What conditions had to be true for this particular structure to be realized?

That question has been bothering us because we have spent months doing something slightly ridiculous with physics: systematically taking familiar explanatory words away and seeing what the observations actually force us to put back. Time. Motion. Trajectory. Propagation.

Dynamic space. Change. Direction. Every round gets uglier. And every round has left less behind.

The most recent result brought us to something surprisingly spare: configurations, relations, constraints, differences in admissibility, correlations, conservation requirements, boundary conditions, and physical asymmetries. A lot of the movie we normally play in our heads may be representation. Useful representation. Often indispensable representation. But representation nonetheless.

That led to another question. What happens if we apply the same discipline to assembly?

Stop telling matter the story

Suppose I want a particular molecule. The ordinary synthetic description is procedural. Do this. Then this. Add this reagent.

Wait. Heat this. Cool that. Intermediate A becomes B. B reacts with C.

C closes into D. Eventually we obtain our target. Nothing is wrong with that description. But it mixes two different things: the conditions required for the molecule to be realized

and one history through which we managed to produce it. Those are not necessarily the same information. Self-assembly already makes that obvious. A crystal does not require someone to place each atom.

A micelle does not come with an assembly crew. A protein does not ordinarily require an external machine to position every residue in its final three-dimensional location. Dynamic covalent systems can form incorrect bonds, reverse them, try alternatives and converge on complex structures. DNA-mediated assembly can encode interaction rules and allow those rules to produce larger organization. Block copolymers organize into domains.

Membranes form. Reaction-diffusion systems generate patterns. Materials anneal defects. Molecular systems self-sort. And we casually put a word over many of these events:

spontaneous. But “spontaneous” does not mean structure came from nowhere. Maybe it often means: We supplied enough constraints that we no longer had to specify the construction trajectory. That is a very different statement.

Constraint Realization

I’m calling the general idea Constraint Realization. For materials and molecular systems: Constraint-Realization Assembly, or CRA. The premise is modest: A physical structure can sometimes be specified more economically by the constraints it must satisfy than by the sequence of operations conventionally used to construct it.

Imagine that we want some target structure.

We have:

Those requirements might include:

Now ask a different question. Instead of: What sequence of steps builds the object? ask: Which physically possible configurations satisfy all of these requirements?

That collection of acceptable possibilities is the admissible set. Nothing about that definition tells matter which route it must take. It tells us what counts as an acceptable realization for the operation we are performing. And that last phrase matters. A constraint specification is not presumed to be the thing itself.

It is an operation-sufficient representation. It retains the distinctions needed to identify, produce, test, compare or otherwise work with the realization at some declared level of resolution. Its authority ends where its retained distinctions end. A representation can be perfectly sufficient for one operation and insufficient for another. A molecular description adequate for synthesizing a compound may be inadequate for reproducing every quantum correlation within it.

A structural description adequate for manufacturing a chair may be inadequate for authenticating that particular chair. A representation can be exact for its job without becoming constitutive of what it represents. That distinction is important. We are not claiming: The object is nothing but our constraint description.

We are asking: How much constraint information is sufficient to carry this realization for this operation? That is the whole game.

The experiment hiding inside the idea

This does not need to begin with exotic physics. Please don’t build a transporter. Build a boring molecule. Pick a well-understood reversible molecular self-assembly system — perhaps a known dynamic-covalent macrocycle or molecular cage. Something whose chemistry is painfully established.

Then prepare it several substantially different ways while preserving the important resources and constraints. Different reagent-addition orders. Different intermediate populations. Different allowed histories. Same target conditions.

Then measure the final product distributions. If radically different preparation histories converge on essentially the same molecular realization, that is interesting. If changing one critical constraint destroys that convergence, that is more interesting.

Now we can start separating:

How much does history matter?

Each preparation history gives us some distribution of final products. Now compare those distributions. If very different allowed histories produce very different endpoints, the system is trajectory-dominant. If very different histories converge on essentially the same realization, the system is constraint-dominant. Call the amount by which the endpoint changes when we change the route Trajectory Sensitivity.

Then define its complement as a Constraint-Dominance Index. In plain English: high trajectory sensitivity means the route matters a lot; high constraint dominance means the route matters surprisingly little. Nothing mystical.

Something we can measure. And neither result is automatically better. If the route matters, then the history is carrying essential organizational information. Good. That is exactly what we want the experiment to reveal.

Now start deleting instructions

This is the part I really want to see done. Once we know a specification reliably produces the target, begin removing information from it. One constraint at a time. Delete a geometric condition. Broaden a tolerance.

Remove a template. Relax a field. Allow another component. Change a boundary. Keep stripping the externally supplied specification until realization fails.

Eventually we should discover a smallest tested constraint set sufficient to produce and identify the target reliably for a declared operation, resolution and tolerance. That qualification matters. There may be no single context-free “minimum description” of a molecule. A specification sufficient to produce it may not be sufficient to repair it. A specification sufficient to identify its connectivity may not be sufficient to reproduce its electronic state.

A specification sufficient for one experiment may legitimately compress distinctions that a later experiment must reopen. That does not mean those distinctions were absent from what occurred. It means the earlier representation did not need to carry them separately. So the deeper question becomes: What is the minimum amount we have to tell matter for this operation before matter supplies the rest?

That is not the same as asking how many construction steps the object requires. Not even close.

Specification leverage

Here is another quantity I would like to measure. Suppose an extraordinarily complicated realized structure can be reliably produced from a comparatively compact set of imposed constraints. Call that Specification Leverage. In ordinary language: Specification Leverage compares how complicated the realized structure is with how little information we had to impose to get it.

High specification leverage means: we told the material comparatively little, and its own physical constraint structure supplied a lot. The exact way we measure “information” will matter, so nobody should pretend this is a universal constant. But within controlled experiments, it could be extremely useful. Nature seems extraordinarily good at that trick.

We might want to get better at exploiting it.

Maybe self-assembly and manufacturing are not opposites

Manufacturing usually means increasing control. Put this here. Machine that surface. Deposit this layer. Move that component.

Repeat. Self-assembly transfers some of that organizational burden into the physics of the material. So perhaps the long-term optimization problem should be: Use as little externally prescribed construction information as possible while still producing the desired structure reliably. Or, stated even more simply:

Tell matter less. Get the same result. Instead of forcing every constituent into position, engineer a physical situation in which unwanted configurations become impossible, unstable or strongly disfavored, while the wanted organization completes itself. We have already been doing primitive versions of this forever. Chemistry does it. Biology does it.

Materials do it. We are usually studying the individual mechanisms. Maybe we should also study the general information economics of realization.

The configuration writer

Push that idea far enough and you get a strange machine. Not a 3D printer. Not exactly a molecular assembler. A configuration writer.

Give it:

Then ask: Under what smallest sufficient set of physical constraints does the available matter reliably realize this structure? The writer does not necessarily dictate the route. It designs the landscape of possibilities. Then matter does the remaining work.

That is the interesting machine. And even here, the specification remains operation-relative. If a later task requires distinctions the writer’s representation compressed, those distinctions must be reopened through provenance, additional measurement or finer specification. The fact that one representation was sufficient does not grant it explanatory authority over distinctions it did not retain.

And this changes the old teleportation thought experiment

Teleportation has traditionally been imagined as transportation:

But once you stop assuming that physical traversal is what establishes the realization, the engineering problem changes. Suppose we possess a sufficiently complete specification of the target for the operation we care about. Suppose appropriate matter and energy already exist somewhere else. Then the practical question becomes: Can those local resources be made to realize a configuration equivalent to the target?

No source atom necessarily has to travel there. At ordinary resolution, that is manufacturing. At molecular resolution, extraordinarily good manufacturing. At atomic resolution, absurdly good manufacturing. Push the specification all the way into whatever quantum correlations genuinely matter, and suddenly the border between “manufacturing” and what science fiction called “teleportation” gets difficult to define.

I’d call that limiting idea: remote relational instantiation. It does not make Star Trek transporters possible. It changes what we would actually be trying to accomplish. And it separates another question that teleportation stories usually mash together:

realization is not identity. If two structures satisfy the same specification, physics still owes us whatever additional criterion permits us to call them the same individual. That is another problem.

Now reconsider “spontaneous”

This may be the most interesting application. Science is full of events described as spontaneous. Spontaneous crystallization. Spontaneous symmetry breaking. Spontaneous emission.

Spontaneous nucleation. Spontaneous folding. Spontaneous organization. Sometimes that terminology is exactly appropriate. Sometimes the event is genuinely stochastic under the applicable theory.

But perhaps we have also bundled very different phenomena under one linguistic trash can. So I propose an audit. For every phenomenon currently called spontaneous: Freeze what was actually observed. Identify the matter and energy available.

Identify the boundary conditions. Identify the candidate structural constraints. Define what counts as the realized structure. Change preparation history while preserving those constraints. Measure whether the final realization changes.

Remove constraints one by one. Find the minimum sufficient specification for the declared operation. Determine what the word “spontaneous” was hiding. Then ask one more question: What did the chosen measurement preserve, what did it compress together, and would a later operation require reopening distinctions the present description did not need?

That matters because evidence is also a representation. An instrument does not hand us “reality itself.” It gives us an articulation that retains some distinctions and suppresses others. Making that office explicit does not weaken evidence. It makes evidence more precise.

Maybe the audit reveals nothing surprising. Maybe quite a lot.

This is not a claim that kinetics don’t matter

They obviously do. Some structures get kinetically trapped. Some reaction pathways are irreversible. Some materials have hysteresis. Some targets depend heavily on preparation history.

Some systems are driven far from equilibrium. Some structures are impossible without sustained energy flux. Good. Constraint Realization should expose those differences rather than hand-wave them away. A system with high trajectory sensitivity is telling us something important:

the history carries essential organizational information. That is just as interesting as finding the opposite.

Nor is this “time is fake”

The work that led us here came from a much broader attempt to remove unearned temporal and directional assumptions from physical descriptions. But you do not have to buy any of that to test Constraint Realization. Treat time normally. Use ordinary chemistry. Use ordinary thermodynamics.

Use ordinary dynamics. The experiment survives. All I am asking is that you distinguish: the history through which the structure appeared from

the information required to specify what structure appears. If those always turn out to be equivalent, fine. The framework fails.

Where this framing becomes insufficient

This idea should not be protected. Here is where it would fail. Show that holding resources, boundaries and declared constraints fixed while changing the preparation history generally produces materially different endpoints. Show that no stable minimum sufficient constraint set can be identified for a declared operation. Show that every apparent constraint specification secretly contains the entire construction procedure.

Show that the proposed measurements depend so strongly on arbitrary definitions that useful comparisons become impossible. Show that conventional kinetic and thermodynamic theories already provide everything here without useful compression. Show that the framework cannot transfer between even two domains. Show that what we called a constraint representation consistently omits distinctions required for the operation it claims to carry. Any of those would damage the framework.

Some would make it unnecessary. That is good. The purpose is not to preserve Constraint Realization. The purpose is to track the structure until this representation stops carrying it accurately.

What would make it interesting?

The opposite results. Suppose a reversible molecular system is prepared through ten substantially different allowed histories. And repeatedly produces the same cage. Then we remove most of our externally specified information. Still the same cage.

Remove one apparently tiny structural constraint. Now the realization explodes into competing structures. That tells us something very specific. It tells us where the organizational information required for that operation actually lived. Do the same with a block copolymer.

A crystal. A vesicle. A folding peptide. A reaction-diffusion pattern. A regenerating biological structure.

Now maybe we have something larger than another word for self-assembly. We have a comparative science of physical realization.

A possible new axis for complexity

There is already serious work asking how difficult an object is to construct. Good. I want another axis. Not: How many steps does this object require?

But: How many constraints does this object require for the operation being performed? Those quantities need not track each other. Imagine an object with extremely high constructive complexity but very low constraint complexity because the material does almost all of the completion. Or the inverse:

a simple-looking product requiring exquisitely specific environmental conditions.

So organized matter may deserve at least two coordinates:

And constraint complexity itself may change with the operation. Fabricate. Identify. Repair. Authenticate.

Replicate. Predict. Different operations may require different retained distinctions. That is not a flaw in the framework. That may be one of the things worth measuring.

The broader hypothesis

Here is the strongest version I am comfortable putting into the wild: A nontrivial fraction of phenomena described as spontaneous or self-organizing may be better characterized as constraint realizations: physical configurations whose organization is determined more strongly by relational constraints and boundary conditions than by any uniquely required formation trajectory. Maybe it is wrong. Maybe it applies narrowly. Maybe it is painfully obvious and already hiding across twelve different literatures under twelve different names.

All three outcomes are useful. Because even if every ingredient already exists, the cross-domain measurement question may not: How much did we actually have to tell matter? That is the experiment.

Vocabulary, for anyone who wants to play

Constraint Realization

Physical realization of a target class under declared resources, boundaries and structural constraints without requiring one particular construction trajectory as part of the target definition. Constraint-Realization Assembly Constraint Realization applied to molecular and material assembly. Constraint Specification An operation-sufficient representation of the physical conditions an acceptable realization must satisfy.

It is not presumed to exhaust the occurrence it represents. Realization Class The set of physical configurations treated as equivalent to the target for a declared operation, resolution and tolerance. Trajectory Sensitivity How much the final realization changes when the allowed preparation history changes.

Constraint-Dominance Index How strongly the final organization survives changes in preparation route. Minimum Sufficient Constraint Set The smallest tested specification that still carries the required realization reliably for a declared operation, resolution and tolerance. Boundary Sensitivity

How strongly realization depends on environmental or externally maintained conditions. Specification Leverage How much realized organization is obtained from a comparatively small amount of externally supplied specification. Constraint Closure The point at which adding more procedural detail stops materially improving the reliability or specificity of the target realization for the operation being performed.

Reopening Restoring distinctions that an earlier operation-sufficient representation legitimately compressed when a later operation requires finer resolution. Remote Relational Instantiation Realization of an equivalent target configuration from local resources elsewhere rather than physical displacement of the source constituents. Configuration Writer

A hypothetical system that determines and imposes the smallest sufficient physical constraint set under which available matter reliably realizes a specified target.

First experiment

Do not invent anything impressive. Pick a known reversible molecular cage. Make it through several deliberately different histories. Measure the final product distributions. Remove constraints.

Repeat. Find the smallest specification that still works for the declared target and measurement resolution. Measure how much changing the route changes the result. Compare it with a more route-dependent system. Then change the operation.

Ask whether the same representation is still sufficient. If not, reopen the distinctions it compressed and determine what additional relation the new operation requires. The first useful result could be embarrassingly small. That is fine. If this idea has legs, nature will supply the scale later.

I do not particularly care whether this ends up belonging to physics, chemistry, materials science, information theory, engineering, or some discipline that does not have a department yet. Names are administrative. The experiment is not. We already know matter can assemble astonishing complexity without our placing every constituent. What I want to know is much more specific:

How little do we have to tell it?

A last-minute acknowledgment

As this article was being prepared for publication, R. Wade H. Marr (@HunterWade) responded to our preceding piece, BEFORE THE NOUN, with an analysis that arrived at exactly the right moment. His response did not change the central proposal here. Constraint Realization, Constraint-Realization Assembly, the experimental program, and the larger question — How little do we have to tell matter? — were already in place.

But he caught an important danger in the language we were about to publish: the possibility of turning the constraint representation itself into the next privileged noun. Wade sharpened the distinction between what occurs and the representation sufficient to carry it for a particular operation. In his formulation, a representation has authority only over the distinctions it actually retains for the operation it is being asked to perform.

That insight led directly to several last-minute refinements in this article. We made explicit that a Constraint Specification is an operation-sufficient representation, not a claim to exhaust the physical occurrence itself. We tightened the Minimum Sufficient Constraint Set so that “sufficient” is always understood relative to a declared operation, resolution, tolerance, and measurement context. We added the concept of Reopening: a later operation may legitimately require distinctions that an earlier representation compressed together, without implying that those distinctions were absent from what originally occurred.

His remarks about provenance also sharpened our treatment of evidence: a measurement preserves some relations and suppresses others, and its explanatory authority should not silently extend beyond what the instrument and representation actually retained. And his challenge to adversarial framing prompted another small but meaningful change. Instead of asking only how to “kill” the framework, we now ask where the representation becomes insufficient — while keeping the failure conditions every bit as severe.

Those are not cosmetic changes. They help prevent Constraint Realization from committing the same mistake it is intended to expose: mistaking an effective representation for the occurrence it successfully represents.

So, just before publication:

— with appreciation to R. Wade H. Marr (@HunterWade) for pushing the representation one level farther upstream before we hit “publish.”

← Before the Noun The Attack Surface Is Civilization →