# God's Infinite Dimensional Space

### Transcendental Embeddings, Predictive Actor-State, and the Next Phenomenal Transition

> Und mich ergreift ein längst entwöhntes Sehnen  
> Nach jenem stillen, ernsten Geisterreich,  
> Es schwebet nun, in unbestimmten Tönen,  
> …  
> Was ich besitze seh' ich wie im weiten,  
> Und was verschwand wird mir zu Wirklichkeiten.
>
> What I possess appears as if far away;  
> And what has vanished becomes real to me.
>
> — Goethe, *Faust I*, “Zueignung” (*translation mine*)[^goethe]

**How does reality appear to you?**

Not *what is reality*, as though you could stand outside the thing and look backward at it; not *what is matter made of*, as though one more instrument, one more layer of physical description, and one more century of increasingly expensive microscopy would finally deliver the world without an observer attached; how does reality appear **to you**, and what sort of structure must already exist before anything can appear at all?

Reality is too large to be experienced all at once, therefore an organism receives no unfiltered copy of the universe; it inherits a finite way of carving differences out of it, a repertoire of distinctions, saliences, couplings, and responses, then one life works on that inherited structure through language, culture, memory, injury, repetition, love, status, work, humiliation, whatever else managed to leave a mark, until the species-compatible template has become this particular actor, occupying this particular phenomenal state, at this exact moment, while only a small and proposition-dependent part of the whole thing is active.

This manuscript is an attempt to write that process down.

I will be blunt, this work is a monster, and it is autobiographical of the mental state of the author who wrote it; the concepts were not found in a clean sequence, they were dragged out through reading, engineering, argument, private obsession, and the debauched and tortured method by which I seem to learn anything difficult. As philosophy it is ambitious and occasionally undisciplined; as mathematics it is mostly old machinery attached to a strange choice of objects; as machine-learning research it may be worthwhile if the ontology pays rent; as a completed account of mind it cannot be finished in a lifetime, which is fine, I do not need to finish mind in order to predict the next thing a person does.

The compressed argument is this:

1. A mind-independent external domain exists, but no actor encounters it without mediation.
2. A lineage inherits a restricted organization through which some external differences can become distinctions for that kind of organism.
3. Development and history realize that inherited structure differently in each actor.
4. The actor occupies a changing phenomenal state: the reality available to the actor at that instant, including perception, memory, feeling, thought, and action already underway.
5. Thought and action are not separate substances inside this system; both are state transitions produced as noumenal conditions and prior phenomenal reality continue acting on one another in a loop, with the difference appearing mainly in which consequences become externally visible.
6. Under complete state information and the true transition law, the next state is an asymptotic object of exact prediction; under finite information, the honest target is a conditional distribution.
7. We cannot observe the complete interior, so we estimate a predictive actor-state from traces.
8. A proposition does not act on an abstract person; it is reconstructed inside an actor, under a role, a relationship, an institution, and a present state.
9. The resulting interaction changes the actor; visible behavior is one residue of that transition.
10. A useful actor representation should survive more than one task. Inputs and outputs are probes; the ontology is the product.

<a id="gids-e1"></a>
**GIDS–1 — Registration and lineage access.**

A local external condition becomes a modeled distinction, and then only the accessible component survives for the actor class:

\[
\omega_t\in\mathcal N_{\mathrm{loc}}
\xrightarrow{\operatorname{Reg}}
\mathbf g_t^{\mathrm{reg}}\in\mathcal G
\xrightarrow{P^{\mathrm{spec}}}
\widetilde{\mathbf g}_t^{\mathrm{spec}}\in\mathcal M^{\mathrm{spec}}.
\]

Read it from left to right. Something exists locally in the external world, \(\omega_t\); the model registers whatever differences it can describe, placing them in an open-ended representational arena \(\mathcal G\); a species-level access structure retains only those distinctions available to that lineage at the resolution we are modeling. The organism does not see the left side of the expression, because by the time anything is available to experience the world has already been selected, organized, and made consequential.

<a id="gids-e2"></a>
**GIDS–2 — The actor descent.**

\[
G_i
\longrightarrow
T_{i,t}
\longrightarrow
\phi_{i,t}
\rightsquigarrow
Q_{i,t}
\xrightarrow{\Pi_{\tau,\Delta}}
q_{i,t}^{(\tau,\Delta)}.
\]

Again, read the equation as a descent in resolution rather than a claim that every object is the same mathematical type. \(G_i\) is the inherited starting organization of actor \(i\); \(T_{i,t}\) is the slowly changing actor that organization becomes after one life has worked on it; \(\phi_{i,t}\) is the actor's complete phenomenal state now; we do not observe that state, so \(Q_{i,t}\) preserves the general response structure required across admissible propositions, while \(q_{i,t}^{(\tau,\Delta)}\) is the smaller part needed for one task \(\tau\) and horizon \(\Delta\).

<a id="gids-e3"></a>
**GIDS–3 — Estimation, simulation, and visible trace.**

\[
\mathcal H_{i,<t}
\longmapsto
\widehat s_{i,t},
\]

\[
(\widehat D_t,X_t=x_t,\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1})
\longmapsto
\widetilde D_{t+1}
\longmapsto
\widetilde O_{t+1}.
\]

The history available before a decision, \(\mathcal H_{i,<t}\), is compressed into an estimated state \(\widehat s_{i,t}\); a hat means *estimated from evidence*, while a tilde means *simulated by the model*. The larger state \(\widehat D_t\) may contain several people, institutions, a relationship, and a world-state; a candidate proposition \(x_t\) encounters that construction, exogenous conditions \(\boldsymbol\xi_{t+1}\) are supplied or modeled, the system simulates the next state, and only afterward does it decode the traces likely to escape into observation.

That is the whole descent:

\[
\text{external difference}
\rightarrow
\text{actor-accessible distinction}
\rightarrow
\text{realized actor}
\rightarrow
\text{present state}
\rightarrow
\text{predictive state}
\rightarrow
\text{estimated state}
\rightarrow
\text{next transition}.
\]

The philosophical argument exists because the engineering choices otherwise look arbitrary. Why split slow person structure from fast state? Because an actor is a historically realized organization occupying a moment, and those two things move on different clocks; why represent the proposition separately? Because the same external arrangement is not the same experienced object for two actors; why carry role, relationship, and institution? Because the proposition arrives through them and is changed by them; why demand transfer across tasks? Because a representation that predicts one label may be nothing more than a shortcut to that label, a useful feature perhaps, but not an ontology of the actor.

The founding ideal is exact prediction under complete state information, which is not the first empirical promise and should not be mistaken for one. The first promise is smaller: as evidence accumulates, we can construct increasingly useful approximations of the distinctions by which an actor understands and operates in the world; those approximations can improve forecasts over long horizons and under different degrees of available information, then the same structure can be tested elsewhere instead of being congratulated for explaining the target that created it.

This is not a theory of truth. There may be a mind-independent world, but the actor does not receive truth as an input field; the actor receives a heavily transformed stream, and for the present research program I do not care whether that transformation resembles reality in some morally satisfying way, I care whether it preserves the distinctions that govern later transition. A person who persistently misreads threat, status, affection, competence, or probability does not contain noise around a preferred rational baseline; the misreading is part of the actor we must model.

## The First Machine Failed, Which Was Useful

Before this structure existed, I tried to solve a cruder problem with a much stupider machine. I collected opinions from people working inside the same company, built loose descriptive embeddings of the people and the available decision-space, asked everyone what they thought should be done, then attempted to weight the responses into a kind of organizational swarm intelligence; whichever proposition survived would be implemented, the result would be measured over time, and the aggregate judgment would—at least in the optimistic version of the story—become wiser than any single participant.

It failed for reasons that now look embarrassingly obvious. The people giving opinions bore little consequence for being wrong, therefore many supplied justifications they probably did not believe; the eventual success of a decision was wrapped inside hiring, execution, timing, management, politics, and every other process the company was already running, which meant the signal dissolved into the organization before I could measure it; the system treated the company as a crowd of opinions rather than an access architecture with authority, incentives, missing information, vetoes, and memory; worse, I had imposed operational labor on the organization merely to feed my stupid program, so the measurement device changed the thing being measured while producing results too noisy to justify the burden.

The current paradigm came out of that failure. Do not ask people what they will do as the principal feedback loop; observe what they actually did, under what role and consequence, structure the event-clock until the data stops lying, and choose an early laboratory in which propositions, responses, and economic outcomes can be attached to one another with unusual hardness. Sales is not behaviorally deterministic, but it provides a comparatively deterministic training instrument: a proposition was delivered at a time, a response followed or did not, a relationship and company state changed, and revenue can eventually be attached to the trajectory. This reduced the mixed-incentive problem, reduced the operational load placed on participating organizations, and forced the model to confront composite actor-state instead of averaging everyone into a cheerful but useless vote.

## What Kind of Actor Can Enter the Space?

The same formal arena should be capable of carrying the phenomenal state of a fish, a human, an ant colony, a corporation, or a hypothetical silicon actor, provided we are disciplined about what *phenomenal* means at the chosen scale. For a human it includes the lived mental interior; for a colony or corporation it means the organized reality available to the composite actor through its members, records, channels, memory, authority, and recurrent patterns of response, which does not settle whether the institution possesses one unitary human-like consciousness, nor does it need to. The actors are not physically identical, their internal construction is not interchangeable, but the outer grammar—access, organization, state, proposition, transition, trace—can remain stable.

This is why the manuscript refuses to wait on the final physical explanation of consciousness. Most attempts to formalize mind eventually choose one of three comfortable failures: they wait for neuroscience to finish the map, after which understanding is promised to arrive; they remain purely behavioral, treating the actor as an input-output box with no durable internal structure; or they disappear into phenomenology, producing ever finer descriptions of experience without an engineering object one can estimate. I want a fourth path: take the structure of experience seriously without requiring its complete physical reduction, define an actor at the level where persistence, access, memory, transition, and traces can be modeled, then make the model answer to the future.

The nearest intellectual relatives are scattered across several fields: Kant's account of the conditions of experience; evolutionary epistemology's attempt to place those conditions inside natural history; Marr's separation of computational, algorithmic, and implementational explanation; predictive-state representations; psychometrics and factor analysis; state-space modeling; and the intentional stance as a legitimate predictive abstraction.[^kant][^lorenz][^marr][^dennett][^psr] None gives me the complete engineering program I want, so this manuscript attempts to force them into one, perhaps against their wishes.

The inversion against ordinary machine learning is deliberate:

> **Standard machine learning:** optimize a model for a task; accept the latent representation that falls out.
>
> **GIDS:** construct a stable latent object over the actor; use tasks to discover whether that object contains reusable structure.

A foundation model may learn an extraordinarily useful geometry because scale and objective force one to emerge; GIDS asks whether the ontology itself can be built, tested, split, merged, transferred, and treated as a durable object. The outputs are not the destination, they are instruments with which we strike the latent structure and listen for what rings.

The manuscript proceeds in eight movements:

1. [Kant From an Evolutionary Perspective](01_Kant_From_an_Evolutionary_Perspective.md) — why experience must be constructed, and why a closed table of categories is useless for the work ahead.
2. [God's Infinite Dimensional Space](02_Gods_Infinite_Dimensional_Space.md) — the common arena, the access maps, the evolutionary origin of possible distinctions, and the rule by which a model should discover new ones.
3. [Behold; You! The Chimera](03_The_Chimera.md) — how a species-compatible template becomes one person, one role, and one proposition-conditioned response.
4. [Predictive Actor-State](04_Predictive_Actor_State.md) — the move from phenomenal life to a state that can be estimated from traces, remembered, compressed, and transferred.
5. [Composite Actors](05_Composite_Actors.md) — corporations, colonies, relationships, and actors constructed at several scales.
6. [World Models and Proposition Search](06_World_Models_and_Proposition_Search.md) — simulation, filtering, trajectories, and the selection of what to present next.
7. [Evaluation and Ontology Growth](07_Evaluation_and_Ontology_Growth.md) — event time, transfer, baselines, essential tests, drift, and the conditions under which a distinction earns its place.
8. [Study Guide / Cheat Sheet](08_Study_Guide.md) — the system compressed into a short working reference, with the indispensable equations read in plain English.

One scope covenant before we begin. From the first learned representation onward, every latent variable is an object in a fitted model unless I explicitly say something stronger; I will not append the same ceremonial disclaimer to every equation, because where the distinction between a representation and the thing represented becomes dangerous I will stop and mark it, and elsewhere you may assume I remember.

If you reject the starting position—that experienced reality is actor-relative, structured, historically realized, and predictive of action—then much of what follows will be useless to you; but I swear to entertain, nonetheless.

---

[^goethe]: Johann Wolfgang von Goethe, *Faust: Der Tragödie erster Teil*, “Zueignung” (1808). The English rendering above is the author's.
[^kant]: Immanuel Kant, *Critique of Pure Reason*, especially A19/B33–A49/B73 on the forms of intuition and A50/B74 onward on the understanding's contribution to experience.
[^lorenz]: Konrad Lorenz, “Kant's Doctrine of the A Priori in the Light of Contemporary Biology,” originally published in 1941; English translation in H. C. Plotkin, ed., *Learning, Development, and Culture: Essays in Evolutionary Epistemology* (Wiley, 1982), 121–143. The resemblance is real, though GIDS is aimed at an operational actor model rather than a general defense of evolutionary epistemology.
[^marr]: David Marr, *Vision: A Computational Investigation into the Human Representation and Processing of Visual Information* (W. H. Freeman, 1982).
[^dennett]: Daniel C. Dennett, *The Intentional Stance* (MIT Press, 1987).
[^psr]: Michael L. Littman, Richard S. Sutton, and Satinder P. Singh, “Predictive Representations of State,” *Advances in Neural Information Processing Systems 14* (2001), 1555–1561. The precise relationship appears in [Predictive Actor-State](04_Predictive_Actor_State.md).

# Kant From an Evolutionary Perspective

> Transcendental idealism is a man walking on phantom legs; locomotion is achieved through selective understanding, while the origin of the legs is hidden inside the system.

After reading the *Critique of Pure Reason*, I was struck by a simple irritation: Kant was extraordinarily close to a useful account of reality as it appears to an observer, then froze the machinery outside history. Space, time, causality, objecthood, unity, plurality—these arrive as conditions through which experience becomes possible; fine, but how did a creature come to possess those conditions, why do different creatures possess different ones, why do individuals vary, and what happens when a distinction that matters to one actor is not merely ignored by another but unavailable to its reality altogether?

That was not Kant's project, unfortunately; it is mine, which is either good news for this manuscript or terrible news for everyone forced to read the next several sections.

Kant's central move was to deny that the mind receives experience as a passive copy. Sensibility does not simply deliver a finished world; the subject contributes forms and concepts through which a manifold becomes an object of possible experience, and “thoughts without content are empty, intuitions without concepts are blind” remains the cleanest sentence in the whole architecture.[^kant-empty] The world of experience is therefore not the external world naked, it is the external world as organized under the conditions of the observer.

Keep that; put the observer back inside nature.

A mind-independent external domain exists, therefore I am not denying rocks, stars, photons, bodies, or whatever deeper structure survives our current physics; I am denying that an actor receives this domain without transformation. Reality for an organism is the form produced when external difference encounters inherited access, developmental organization, memory, present state, and the demands of whatever transition is possible next; the actor does not stand between two mirrors, comparing appearance against the thing in itself, because it lives on the appearance side and calls the result ordinary.[^recap-registration]

This is not solipsism in the stupid sense that only one mind exists; other actors and the external world are real enough to kill you. The claim is that experienced reality is actor-relative, correspondence without identity, constraint without a God's-eye copy, and I do not need evolution to deliver truth when adaptive distortion is perfectly capable of surviving. The organism requires a world that produces workable transitions, not a philosophical certificate establishing resemblance to the noumenon.

Lorenz made the neighboring move when he argued that what is *a priori* for the individual may be *a posteriori* for the lineage: inherited cognitive structure is ancestral experience sedimented into biology.[^lorenz] I discovered that tradition late, which was irritating but reassuring; the older literature supplies an ancestry to the first half of the program, although GIDS wants something more operational from it, because I do not want only an account of why cognition fits a human-scale world, I want a representation that can be estimated for one actor and used to forecast the next state under a proposition.

## Rocks Delusional Enough to Think They Are Alive

My modification of Kant is roughly this:

> An organism's representation of reality presupposes an inherited and developed organization capable of turning external difference into an actor-relative world; evolution creates the possibility of distinctions, development and history realize them, present state determines which become active now, and thought or action continues the same phenomenal transition through another phase.

Evolution does not begin with a finished mind and optimize its opinions, it begins with systems that persist or fail to persist. A membrane that responds differently to a dangerous ionic condition already embodies, at the resolution of our model, a consequential distinction; a light-sensitive molecule changes the transition possibilities of the organism; a comparator between two light-sensitive structures permits contrast; a second pigment permits chromatic relation; sensorimotor loops acquire depth, timing, object persistence, bodily boundary, food, threat, kin, territory, hierarchy, and eventually the synthetic absurdities humans inhabit together.

Imagine an organism whose visual world is a \(48\times48\) grid, a shape enters the grid, and the creature has three gross responses:

\[
\{\text{attack},\text{mate},\text{run}\}.
\]

Suppose attacking is the only response compatible with survival under one recurrent condition; a bias toward attack spreads, not because the organism has discovered the truth of the shape, but because the transition worked. Now make the world slightly less stupid: the correct response depends on relative size, size requires some organization of spatial extent, relative size over time recruits distance and velocity, and treating the shape at one moment as the same shape at another requires persistence and a primitive causal organization. None of these distinctions needs to appear as a proposition the creature consciously reasons through; they can be built into the way its world arrives.

This is the point Kant could not supply from inside his project: the structures that make experience possible may themselves be historical products, layered, coupled, and reorganized across lineages; the *a priori* need not descend from reason, it can be the residue of organisms whose worlds were carved badly enough that no descendants remained to complain.

Fitness is not truth, and for this system truth is mostly beside the point. I do not care whether the creature sees the shape “as it really is”; I care that the experienced distinction predicts its next state. A systematic distortion can be more stable, useful, and behaviorally consequential than a veridical representation, which humans demonstrate continuously—threat is magnified, status is hallucinated, affection is misread, probability is mangled—and none of this should be discarded as debris around an otherwise rational actor, because the distortion is part of the actor.

## We Thought, Therefore I Think

“I think, therefore I am” establishes almost nothing useful about how thinking became possible; a better engineering aphorism would be: **we thought, therefore I think; we existed, therefore my existence is dependent.** The individual mind arrives after a lineage, a body, a language, a family, a culture, and an enormous inherited machinery of distinctions, then proceeds to mistake the resulting mental interior for a private first principle.

Mathematics is one of the cleaner examples of several interpretive structures operating together. To count to three, an actor must already treat portions of experience as separable, persistently identifiable, and collectable under one operation; there must be something like objecthood, plurality, recurrence, and a rule for collapsing several distinguished objects into one abstract total. The numeral is not lying around in the physical world waiting to be inhaled; it is an efficient operation over an organized experience of multiplicity, and once the structures and rules are granted the consequences can become brutally constrained, but the ability to form the relevant objects and operations is not produced by pure reason floating before all possible minds.

Kant calls the unification of a manifold under one act of cognition *synthesis*, which is useful language; the trouble begins when one confuses the logical necessity of a system, once its primitives are granted, with the biological or historical necessity of those primitives for every possible observer. A posteriori learning is nested inside prior organization, because an actor can learn only through distinctions it can form, although learning may later refine, combine, and reweight those distinctions; to do is to learn because action exposes the structure to consequence, and perception, interpretation, understanding, thought, and movement are not sovereign faculties holding meetings with one another, they are phases inside one coupled transition.

Your politics is not a transparent deduction from public facts; your idea of violence is not the word *violence* plus a dictionary; your self-understanding is not an objective report emitted by a neutral witness. Each is produced from an inherited repertoire, a life history, a social world, present salience, and a recurrent effort to make the next transition survivable, desirable, or merely available.

Psychology has already built crude instruments for this. A construct such as conscientiousness compresses recurring covariances into a named factor, psychiatry groups other patterns under other names, and these are useful topologies drawn over an underlying field; the lines may be wrong, the factors may split or merge, the same label may hide several routes, but the act of compression is the important precursor to what follows.

> Split the sea and walk across the ocean floor with me while tautologists play their games.

## A Fucking Table

> Kant explaining categories, circa 1781 (colorized).

Now comes the worst part of reading Kant: the table of categories.

Kant organizes the pure concepts of the understanding under quantity, quality, relation, and modality; he derives them from the logical forms of judgment and presents the system as complete.[^kant-table] The architecture is beautiful in the way a locked mausoleum is beautiful, everything has a place, nothing new can enter, and the symmetry of the table is made to certify the completeness of mind.

I do not buy it.

I understand where Kant says the table comes from; I do not believe the logical forms of judgment establish a final inventory of the distinctions required for every possible cognition. The derivation closes because the system was built to close, and it does not explain variation between lineages, variation between individuals, developmental failure, acquired structure, cultural invention, intensity, partial activation, contradictory roles, or the possibility that an organism inhabits a world assembled from distinctions no human possesses.

Kant's categories may describe something real about the organization of human judgment, perhaps even broad families of the machinery discussed later, but a fixed table is the wrong data structure. The worst part about reading German philosophers is the tour de force required to explain simple ideas—you will find the irony in that statement soon—yet Kant can spend hundreds of pages constructing the conditions of possible experience while leaving a modern reader begging for one animal, one damaged brain, one child, one culture, one counterexample that forces the architecture to move.

Nietzsche's accusation that philosophy is a kind of involuntary confession is useful here. Kant took the instruments available to one extraordinarily systematic human mind and elevated them into the necessary furniture of experience; he looked at chaos through his own device and mistook the device's closure for the closure of cognition.

His moral philosophy is worse, and I consider the categorical imperative an enormous distraction; we will not be dragging it into this manuscript. The *Critique of Pure Reason* is useful enough without asking its author for permission to leave.

The replacement is not another table; it is an open ontology governed by evidence. Assume sameness along a proposed distinction until a consequential difference proves that the model should split it, permit the split to transfer across tasks, then permit later evidence to rotate it, weaken it, merge it with another distinction, restrict it to one regime, express it as an interaction, or retire it entirely. The ontology is not a sacred list of mental atoms, it is a changing hypothesis about the distinctions by which actors organize and respond to reality.

This is where the philosophical complaint becomes a research program. Kant asks what must be true for human experience to be possible; GIDS asks which distinctions are available to an actor class, how they became possible, how they were realized in this actor, which become active under this proposition, which survive transfer across tasks, and what transition follows. The answer cannot be a table, it requires a space large enough to admit distinctions before we know their names, access structures describing which actors can use them, and a rule for constructing actor-relative objects within that space.

That is where God's Infinite Dimensional Space begins.

---

[^recap-registration]: This restates the philosophical content of [GIDS–1 — Registration and lineage access](00_Opening.md#gids-e1); it is repeated here to make the Kantian argument locally readable, and no new mathematical claim is introduced.
[^kant-empty]: Immanuel Kant, *Critique of Pure Reason*, A51/B75: “Thoughts without content are empty, intuitions without concepts are blind.” Translation wording varies by edition.
[^lorenz]: Konrad Lorenz, “Kant's Doctrine of the A Priori in the Light of Contemporary Biology,” originally published in 1941; English translation in H. C. Plotkin, ed., *Learning, Development, and Culture: Essays in Evolutionary Epistemology* (Wiley, 1982), 121–143. Gerhard Vollmer later developed the related idea of a species-fitted “mesocosm” in *Evolutionäre Erkenntnistheorie* (1975).
[^kant-table]: Kant introduces the table of categories at A80/B106 and argues for their role in prescribing laws to appearances at B163–B165. The objection here is not that Kant forgot his derivation; it is that the derivation does not earn the empirical and cross-organism completeness GIDS would require.

# God's Infinite Dimensional Space

The preceding chapter argued that experience is constructed, and that the structures through which reality becomes available to an actor must themselves have an origin; this chapter supplies the common arena in which those structures can be represented, not because Hilbert space is secretly the substance of mind, but because we need an open-ended mathematical landscape capable of holding distinctions available to a fish, a human, an ant colony, a corporation, or some actor we have not yet imagined, while still allowing finite projections for whatever calculation we can actually perform.

The first movement is the one already introduced in the opening:[^recap-gids1]

<a id="gids-e1-recap"></a>
**GIDS–1 — Registration and lineage access, restated.**

\[
\omega_t\in\mathcal N_{\mathrm{loc}}
\xrightarrow{\operatorname{Reg}}
\mathbf g_t^{\mathrm{reg}}\in\mathcal G
\xrightarrow{P^{\mathrm{spec}}}
\widetilde{\mathbf g}_t^{\mathrm{spec}}\in\mathcal M^{\mathrm{spec}}.
\]

Something exists; the model registers a difference; the actor class retains only what its organization can make available. Everything after this equation is an attempt to explain what those three movements mean without turning the representation into the represented thing.

## Vectors Are All You Need

Earlier in my life I wrote a chapter called “Chaos,” a long deconstruction of light and matter into component datums—light, molecule, receptor, interpretation, experience—then imagined a being capable of receiving all light without any interpretive structure and concluded that omniscience of input would not resemble perfect vision, it would resemble static without form. More data does not solve the problem; the structure that makes differences usable solves the problem, which is why the representation of reality comes before the processor of reality.

I want a formal system with enough composability to move from biochemical distinction, to mental interiority, to outward behavior, then further into a corporation or colony acting at a larger scale; not because these objects are physically identical, but because each can be described through one grammar of access, state, transformation, and consequence. Inside this manuscript, perception, interpretation, understanding, thought, and action are typed manifestations of one coupled state-transition process; moving your arm and silently completing a sentence in your head occur at different physical resolutions, although to the phenomenal state both are continuations of the same loop, noumenal conditions acting on prior phenomenal reality, phenomenal reality changing what becomes available next, and some fraction of the process escaping into the world where another actor can register it.

This is why I use vectors so aggressively. A vector gives the model an addressable pattern of differences, something that can be compared, transformed, gated, composed, and updated; the vector is not the thing. A photon is not secretly an arrow, a person is not a row of numbers waiting to be discovered in the skull, and a corporation is not the average of its employees; a vector is the form in which a distinction enters an operation, and the type of the operation determines what the vector means.

I am not inventing new mathematics; I am deciding what kind of objects the old mathematics should be attached to.

## A Small and Somewhat Disgusting Life

Pretend, for a moment, that you are much dumber than you are now—so dumb that consciousness is too generous a word—things happen to you, your world is a flash of structured activations and strong tendencies, and here is a story of your life:

```text
A0 [ 0.54, -0.13,  0.75,  0.42, -0.26,  0.87 ]
```

The encounter recruits a feeling:

```text
B  [ 0.32,  0.69, -0.15,  0.78,  0.25, -0.44 ]
```

The feeling recruits an action:

```text
C  [ 0.61, -0.33,  0.48,  0.91, -0.18,  0.36 ]
```

Afterward you occupy another condition,

```text
D  [ 0.27,  0.72, -0.09,  0.65,  0.41, -0.53 ]
```

and the first object is no longer represented in quite the same way:

```text
A1 [ 0.54, -0.13,  0.75,  0.42, -0.26,  0.10 ]
```

Kinda gross, but whatever.

The only visible change in the represented object is the last coordinate, \(.87\rightarrow.10\); everything between \(A0\) and \(A1\) is the internal activity required to make the object different **for the actor**, which means the encounter, feeling, thought, instinctive movement, memory, and altered object-representation can all be written inside one general format without claiming they share one physical mechanism.

For readability I separated them, although a phenomenal stream does not wait politely for one vector to finish before another begins. The activations overlap:

\[
S_1=A_0,
\qquad
S_2=A_0\oplus B,
\qquad
S_3=A_0\oplus B\oplus C,
\qquad
S_4=A_1,
\qquad
S_5=A_1\oplus D.
\]

Here \(\oplus\) means only that the modeled state contains jointly active structured components; it does not grant arbitrary addition a psychological meaning. The point is temporal: an encountered object recruits an internal transition, the transition includes whatever we later name perception, feeling, thought, and action, and the transition changes the object as it will be available next time.

The same grammar can describe light hitting a photoreceptor, a founder reading a financing term, a dog hearing a familiar car, an ant colony registering a damaged trail, or an institution receiving a regulatory order; the components and timescales differ, while the outer form does not.

## Two Geometries, Not One Magical Dot Product

There are two representational problems hiding inside the word *embedding*. The first is the repertoire of possible distinctions—the inherited and developed methods by which an actor can experience and organize reality—while the second is the present pattern of activation, the feeling of reality now, including what is perceived, remembered, anticipated, thought, and already being done. The two belong to one process but should not be collapsed; a coordinate can exist in the actor's repertoire without being active now, two distinctions can be structurally similar without producing the same response, and two physically different propositions can produce nearly identical transitions for one actor while producing radically different transitions for another.

<a id="gids-e4"></a>
**GIDS–4 — Structural and response geometry.**

For registered distinctions \(\mathbf u,\mathbf v\in\mathcal G\), write structural similarity as

\[
k^{\mathrm{dist}}(\mathbf u,\mathbf v) =
\langle\mathbf u,\mathbf v\rangle_{\mathcal G}.
\]

Read this only after the encoding has been declared: the inner product measures overlap in the modeled structure of the distinctions, and has no universal psychological meaning by itself.

Now let

\[
B_{i,t,\tau}:\mathcal G\rightarrow\mathbb R^{m_\tau}
\]

map a registered distinction into the task-relevant change it tends to recruit for actor \(i\), at time \(t\), under task family \(\tau\); response similarity is then

\[
k_{i,t,\tau}^{\mathrm{resp}}(\mathbf u,\mathbf v) =
\left\langle
B_{i,t,\tau}(\mathbf u),
B_{i,t,\tau}(\mathbf v)
\right\rangle.
\]

Plainly, two distinctions are response-similar when, after passing through this actor, they bend the next state in similar ways. The map \(B_{i,t,\tau}\) is not a bridge from private mental substance into some alien category called action; it is a local description of one process continuing through the mental interior and, sometimes, into externally visible movement.

A richer implementation may study transition laws through operators in a reproducing-kernel Hilbert space; the core theory does not require the actor to be linear, nor does it identify every slow spectral mode with a psychological trait.[^operators]

## The External Domain and the Common Arena

Let \(\mathcal N\) denote the external, or noumenal, domain: whatever complete conditions exist independently of an actor's organization of them. I do not assume \(\mathcal N\) is a vector space, because doing so would smuggle our mathematics into the thing we are claiming no actor receives directly.

Let \(\mathcal G\) denote **God's Infinite Dimensional Space**, an idealized and open-ended arena capable of representing distinctions that could participate in the phenomenal state or response of possible actor-observers. For a working mathematical model, take \(\mathcal G\) to be a real separable Hilbert space; concretely, one may imagine

\[
\mathcal G\cong\ell^2,
\]

with represented activations

\[
\mathbf g =
\sum_{k=1}^{\infty}g_k\mathbf e_k,
\qquad
\sum_{k=1}^{\infty}|g_k|^2<\infty.
\]

The equation says that the arena may contain indefinitely many directions while any usable activation remains mathematically controlled. This is a learning device and a capacity claim; the operational research program could be reformulated in another sufficiently rich measurable state-space, although Hilbert structure gives us inner products, finite projections, limits, and a clean landscape in which one actor can retain distinctions another cannot.

Why call it *God's* space? Not because this is a religious argument, but because only something like an omniscient observer could register every external condition, every possible organization of that condition, every reality available to every organism or composite actor, and every transition across all of them. \(\mathcal G\) is the model-side silhouette of that impossible view; humans occupy a narrow and violently biased region, fish occupy another, an ant colony or corporation may occupy a distributed construction at another scale, and a hypothetical actor may register distinctions that never become available to us at all.

For the purposes of this manuscript, a phenomenal state is the organized reality available to the actor at the chosen scale. In a human this includes the lived mental interior; in a colony or corporation it includes the distributed state made available through members, records, paths, authority, memory, and recurrent response. This does not establish that every institution has one unitary consciousness in the human sense, which is a separate and perhaps badly formed question; it establishes that the same representational arena can hold the state from which the actor, as constructed at that scale, proceeds.

To represent a local external condition \(\omega_t\in\mathcal N_{\mathrm{loc}}\), introduce a registration map

\[
\operatorname{Reg}:\mathcal N_{\mathrm{loc}}\rightarrow\mathcal G,
\qquad
\mathbf g_t^{\mathrm{reg}}=\operatorname{Reg}(\omega_t).
\]

This is already a model-side object. Registration chooses a resolution, loses information, and may treat physically different conditions as equivalent; no actor has yet experienced anything, we have only written external differences in a common grammar.

## What a Lineage Can Reach

A lineage does not receive every direction in \(\mathcal G\); it inherits a finite or effectively finite repertoire of accessible distinctions. Represent the species-level accessible structure as

\[
\mathcal M^{\mathrm{spec}} =
\operatorname{span}\{\mathbf v_1,\ldots,\mathbf v_d\}
\subset\mathcal G,
\qquad d<\infty.
\]

For the first linear geometry, take the \(\mathbf v_j\) to be orthonormal, with projection

\[
P^{\mathrm{spec}}\mathbf y =
\sum_{j=1}^{d}
\langle\mathbf v_j,\mathbf y\rangle\mathbf v_j,
\qquad
\widetilde{\mathbf g}_t^{\mathrm{spec}} =
P^{\mathrm{spec}}\mathbf g_t^{\mathrm{reg}}.
\]

Read the projection as a picture, not a biological claim about matrix multiplication in the skull: the external condition contains more modeled difference than this lineage can make available, therefore keep the accessible component and discard the rest. Later access maps may be nonlinear, developmental, state-dependent, and lossy in more interesting ways.

This is the first place the title does real work. A direction can exist in \(\mathcal G\) while being annihilated by the human access structure; another actor can retain it, and each actor then constructs a different phenomenal reality from what may be the same local external condition.

## Objects Are Constructed, Not Shelved

An object of experience is not generally one primitive vector waiting in \(\mathcal G\). For actor \(i\), let \(A_{i,t}\) determine what distinctions are accessible and \(\mathcal I_{i,t}\) organize those distinctions through current phenomenal state and context.

<a id="gids-e5"></a>
**GIDS–5 — Actor-relative object construction.**

\[
\zeta_{i,t}^{\mathrm{obj}}(\omega_t) =
\mathcal I_{i,t}
\!\left(
A_{i,t}\operatorname{Reg}(\omega_t),
\phi_{i,t},
 c_{i,t}
\right).
\]

First determine what reaches the actor, then organize those distinctions through what the actor already is, what is active, and where the actor stands; the result is the object **for that actor now**. The same external arrangement may therefore become different objects for different actors, or for the same actor at different times: a price becomes evidence of quality, an insult, a threat to cash, a rounding error, or proof that the seller is unserious, and the physical marks on the page do not decide alone.

A category also requires a declared representation domain; it may appear as a region, prototype, distribution, scoring function, or learned relation, for example

\[
\mathcal C_\kappa\subseteq\mathcal V_\kappa^{\mathrm{cat}},
\qquad
\operatorname{proto}_\kappa\in\mathcal V_\kappa^{\mathrm{cat}},
\qquad
\operatorname{cat}_\kappa:\mathcal V_\kappa^{\mathrm{cat}}\rightarrow[0,1].
\]

A category is not forced to be an axis. Most named psychological objects are conglomerations over subtler coordinates and interactions; factor analysis will return later as the obvious public shortcut into this field.

A proposition follows the same rule. Let \(x_t\) denote the external or symbolic proposition and write its actor-relative form as

\[
\mathbf p_{i,t}(x_t) =
\mathcal P_{i,t}(x_t,\phi_{i,t},c_{i,t}).
\]

The email outside the person is not the email inside the person; the transition is driven by the latter.

## Hunger, War, and the God Object

Humans experience bodily conditions and synthetic concepts through one phenomenal stream. Hunger may involve glucose regulation, fullness, thirst, fat stores, ghrelin, learned expectation, time of day, and memory; war may involve kinship, threat, land, ownership, hierarchy, humiliation, duty, fear, symbolic history, and a model of outsiders; a god object may be assembled from fatherhood, communal bonding, purity, revenge, care, sacrifice, death, transcendence, spite, love, law, and the relations among them.

The sleight of hand is deliberate: I have placed biochemical signals and concepts that exist only through a symbolic world inside one representational grammar. Ah, but that is the point. Hunger and “my group is at war” do not feel identical, nor do they arise from identical machinery, yet both enter the actor's phenomenal state, recruit memory and salience, bend anticipation, and alter what can happen next; the common arena lets us describe their effects without pretending the physical levels are interchangeable.

The *god object* is therefore not theology, it is an example of a high-order human construction requiring many lower-order distinctions and a social world capable of sustaining it. The title points toward impossible omniscient access; the example points toward what humans build inside their much smaller reality.

## The Natural Projection of Possible Distinctions

Evolution explains the creation of possible distinctions; it does not explain every later choice an individual makes, and it will not be used as a universal story about what the actor is optimizing. What follows should be taken seriously as an example of how the accessible structure we are describing can emerge as a natural projection over time, until repeated evolutionary pressure has produced one of the foundational objects of the framework: a lineage-compatible repertoire of distinctions.

At evolutionary stage \(\nu\), let the accessible structure be \(\mathcal M_\nu^{\mathrm{spec}}\), and let a mutation or developmental modification propose \(\Delta\mathbf v\in\mathcal G\). Remove what the existing structure already expresses:

\[
\Delta\mathbf v_{\perp} =
\Delta\mathbf v-P_\nu^{\mathrm{spec}}\Delta\mathbf v.
\]

Let \(\Delta\mathfrak F_\nu\) denote the net lineage contribution of admitting the residual distinction under the environments in which it operates, after developmental, energetic, and maintenance costs are counted. The simplest retention statement is

<a id="gids-e6"></a>
**GIDS–6 — Natural retention of an accessible distinction.**

\[
\mathcal M_{\nu+1}^{\mathrm{spec}} =
\begin{cases}
\mathcal M_\nu^{\mathrm{spec}}
\oplus
\operatorname{span}\{\widehat{\Delta\mathbf v}_{\perp}\},
& \Delta\mathbf v_{\perp}\neq0
  \text{ and }\Delta\mathfrak F_\nu>0,\\[4pt]
\mathcal M_\nu^{\mathrm{spec}},
& \text{otherwise}.
\end{cases}
\]

Read it without mystery: variation proposes a distinction, the projection removes what is already available, and the lineage retains the new capacity when its total contribution survives selection. Real evolution rotates, couples, loses, repurposes, drifts, and preserves structures whose value appears only in combinations; the equation is not a complete population-genetic theorem, it is the minimal formal picture required to show how an accessible repertoire can become richer without a finished table descending into the organism at creation.

## The Model Must Evolve Too

The operational registry is not \(P^{\mathrm{spec}}\); it is a finite catalog of the person, context, source, role, relationship, proposition, and interaction distinctions the current model knows how to represent. It is fallible, commercially valuable, and scientifically provisional—the model's ontology, not the species' anatomy.

Let \(\mathfrak R_n\) denote the registry at discovery stage \(n\). A candidate operation \(\delta\) may split one family into two, merge two families, rotate or reparameterize a latent block, weaken or retire a distinction, restrict it to a role or source, replace a supposed axis with an interaction, or create a new coordinate family. The central rule is simple: assume sameness until a distinction demonstrates cross-task value.

<a id="gids-e7"></a>
**GIDS–7 — Cross-task ontology retention.**

\[
\Delta_{\mathrm{tr}}(\delta) =
\sum_{\tau\in\mathcal T_{\mathrm{tr}}}
\omega_\tau
\left[
\mathcal R_\tau(M_{\mathfrak R_n})
-
\mathcal R_\tau(M_{\mathfrak R_n\oplus\delta})
\right]
-
\lambda_C C(\delta)
-
\lambda_S S(\delta).
\]

Lower risk is better. The bracket measures how much the candidate distinction improves performance across tasks not used merely to name it; \(C(\delta)\) penalizes complexity, \(S(\delta)\) penalizes instability across time, samples, or coordinate charts, and the operation survives only when the gain is positive and repeats.

This is not an Indian buffet process wearing a leather jacket. Bayesian nonparametrics may supply useful priors over an expanding feature set, but the GIDS rule also requires typed distinctions, source and role structure, temporal validation, interaction tests, and transfer.[^ibp]

The typed registry is a first implementation choice made partly because it is debuggable. A single continuously updating manifold may eventually be the technically superior object, particularly where the real structure is continuous and every discrete boundary mutilates a subtle relationship between apparently unrelated points; however, a manifold that cannot tell you which source, regime, interaction, or failed split produced an error becomes difficult to interrogate. The likely mature system is hybrid: typed families and relations where the distinctions are operationally meaningful, reparameterizable latent blocks where continuity matters, and enough structure that the whole thing can be taken apart when it begins lying.

Psychology has always grown this way, albeit under a softer standard. Constructs split, combine, rotate, and disappear because the visible topology is coarse while the underlying field is subtle; GIDS assumes the instability instead of apologizing for it, then asks transfer to decide which distinctions deserve to survive.

The axes are hypotheses; transfer is the judge.

## The Deterministic Silhouette

<a id="gids-e8"></a>
**GIDS–8 — Ideal phenomenal transition.**

\[
\phi_{i,t+1} =
F_i
\!\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
\mathbf p_{i,t}(x_t)
\right).
\]

Read this as the asymptotic ideal: if the complete realized actor, present phenomenal state, context, world, proposition as experienced, and exact transition law were known, the next phenomenal state would follow. Thought and outward movement need not be given separate ontological machinery here; each is a projection of the same recursive actor-world process, distinguished by its coordinates and consequences.

For complete actor-world dynamics, write

\[
\Sigma_{i,t}^{\star}
:=(T_{i,t},\phi_{i,t},c_{i,t},w_t),
\]

and let the operationally honest ideal be

\[
\Sigma_{i,t+1}^{\star}
\sim
K_i^{\star}
\!\left(
\cdot
\mid
\Sigma_{i,t}^{\star},
X_t=x_t,
\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right).
\]

The deterministic arrow is the silhouette; the stochastic kernel is what we can train when state is incomplete or the process itself is stochastic. A deterministic universe is the special case in which the kernel collapses to one point, while our operational uncertainty remains larger because we never possess the complete state.

## From the Lineage to One Actor

Collect the species-level template as

\[
G^{\mathrm{spec}} =
\left(
\mathcal M^{\mathrm{spec}},
P^{\mathrm{spec}},
\mathfrak I^{\mathrm{spec}}
\right),
\]

where \(\mathfrak I^{\mathrm{spec}}\) is the lineage-compatible family of interpretive organizations; one actor begins from an individual inherited seed

\[
G_i =
\left(
\mathcal M_i^0,
A_i^0,
\mathcal I_i^0
\right),
\qquad
\mathcal M_i^0\subseteq\mathcal M^{\mathrm{spec}}.
\]

The actor may lack, alter, or differently organize capacities available in the species-level picture; development, language, culture, memory, and repeated events then realize this starting point into one slowly changing structure. We have therefore built the possibility of an actor's world, but not yet the actor who inhabits it, which is the crueler problem: why does one human occupy a shared world so differently from another, and how can the difference be inferred without opening the skull and pretending the resulting anatomy already explains the mental interior?

Behold; you.

---

[^recap-gids1]: This is [GIDS–1 — Registration and lineage access](00_Opening.md#gids-e1), repeated as the local entrance to the formal chapter; no new mathematical claim is introduced.
[^operators]: Conditional mean embeddings and kernel transfer operators provide one route for representing transition distributions as operators in reproducing-kernel Hilbert spaces. See, for example, Marco Mollenhauer et al., “Kernel Autocovariance Operators of Stationary Processes,” *Journal of Machine Learning Research* 23 (2022). GIDS does not require this realization in its core formalism.
[^ibp]: The Indian buffet process is a useful prior over potentially unbounded sparse latent features, but it does not by itself provide the typed split, merge, role, transfer, and retention logic used here. See Thomas L. Griffiths and Zoubin Ghahramani, “The Indian Buffet Process: An Introduction and Review,” *Journal of Machine Learning Research* 12 (2011), 1185–1224.

# Behold; You! The Chimera

## The Technical Scope, Because Otherwise I'll Accidentally Lie to You

The previous chapter built the possibility of an actor by describing an inherited access structure inside an open representational arena; it did not yet build a person, because a species-compatible template leaves enormous room for variation, and two humans may form roughly the same broad distinctions—objects, faces, threats, promises, status, time, obligation—while inhabiting different realities because the distinctions have been weighted, coupled, remembered, and activated through different lives. The inherited repertoire is only the beginning; a life does the damage afterward.

The rest of the manuscript therefore descends from the actor as possible, to the actor as realized, to the actor as presently lived, then finally to the actor as estimated from outward traces; the sequence is repeated here because every later engineering object depends on not confusing these layers.[^recap-actor-descent]

<a id="gids-e2-recap-chimera"></a>
**GIDS–2 — The actor descent, restated.**

\[
G_i
\longrightarrow
T_{i,t}
\longrightarrow
\phi_{i,t}
\rightsquigarrow
Q_{i,t}
\xrightarrow{\Pi_{\tau,\Delta}}
q_{i,t}^{(\tau,\Delta)}
\rightsquigarrow
\widehat s_{i,t}.
\]

\(G_i\) is the inherited starting organization; \(T_{i,t}\) is the slowly changing actor after development, language, culture, memory, and biography have worked on it; \(\phi_{i,t}\) is the complete phenomenal state now; \(Q_{i,t}\) preserves the general response structure required across a declared family of propositions; \(q_{i,t}^{(\tau,\Delta)}\) is a task-and-horizon summary when one exists; and \(\widehat s_{i,t}\) is the finite object we can actually estimate. The arrows are not promises of lossless recovery, they mark a descent from the actor as lived to the actor as predicted.

There are also three ambitions running through the paper, and mixing them produces spiritual language in the loss function and causal language in the sales dashboard. The **interpretive ambition** explains why an observer has a structured reality at all; the **predictive ambition** asks what actor-state must be preserved to forecast later response; the **control ambition** asks which proposition to present when several are available. The first motivates the second, the second enables the third, and the third has no right to call itself causal merely because the simulator produced a confident number.

For operational purposes, an actor is a bounded construction for which we can describe persistence, channels of access, a changing internal organization, and outward traces. Consciousness matters to the human phenomenal thesis; it is not required for every actor scale, because a colony or institution can possess an organized state available to itself through distributed channels without settling whether the whole construction feels like one human mind.

## The Person Is Never Encountered in the Abstract

Call the person-under-conditions a **Chimera**.

The name is mnemonic, not zoological. A person is never observed as “the person” in a vacuum; you meet a founder during a financing negotiation, a parent during a crisis, a friend after an insult, a soldier under command, an executive inside a budget cycle, and in each case the same durable actor passes through the situation while the proposition reaches a different active organization of that actor.

The earliest notation wrote the Chimera as

\[
\chi_{i,t}=(T_{i,t},c_{i,t}),
\]

which was useful and too static. Role is not merely a label appended to the person; it changes which features of the actor and proposition can interact, therefore the more faithful object is proposition-conditioned:

<a id="gids-e9"></a>
**GIDS–9 — The proposition-conditioned Chimera.**

\[
\chi_{i,t}(x_t) =
\mathcal C_i
\!\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
 x_t
\right).
\]

Read it plainly: the Chimera is the part of the person made active by being this person, in this role, in this world, when confronted with this thing. Role remains explicit in the state because it constrains available information and action, while its principal effect appears inside the proposition–actor interaction, where the same sentence can become an invitation, humiliation, duty, threat, or nothing at all.

The person does not change species when entering a boardroom; the local geometry changes. Some tendencies become salient, others are inhibited, available actions narrow, and statements acquire institutional consequences; a founder generous with friends may become brutally conservative while protecting payroll, a timid employee may become aggressive when speaking through formal authority, and an executive may privately like a proposal while publicly rejecting it because the role requires skepticism.

This is not inconsistency to be averaged away; it is structure.

Imagine several executives rejecting the same product. One rejects it because the vendor looks weak, another because implementation threatens a subordinate, another because the price is suspiciously low, another because a prior deployment ended badly, another because the board has made caution locally rational; the outward trace is the same word—*no*—while the internal route differs, which means a model that learns only the label has learned very little about what to do next.

The Chimera gives us a place to put the difference.

## Psychology and Factor Analysis

Psychology already possesses a rough map of this territory. Factor analysis begins with correlated observations and asks whether a smaller collection of latent factors can explain their covariance; in a conventional common-factor model,

\[
\mathbf x_i =
\boldsymbol\mu_x
+
\boldsymbol\Lambda_{\mathrm{FA}}\mathbf f_i
+
\boldsymbol\varepsilon_i,
\qquad
\operatorname{Cov}(\mathbf f_i)=I.
\]

The measured person is \(\mathbf x_i\), the smaller latent description is \(\mathbf f_i\), the loading matrix says how those factors express themselves in what was measured, and the residual is whatever this particular description failed to capture. The factors are not uniquely oriented; for an orthogonal rotation \(R\),

\[
\boldsymbol\Lambda_{\mathrm{FA}}\mathbf f_i =
(\boldsymbol\Lambda_{\mathrm{FA}}R)(R^\top\mathbf f_i),
\]

therefore the same common component survives a rotated coordinate chart.[^factor] This matters because named constructs feel more ontologically solid than the mathematics warrants: conscientiousness, neuroticism, authoritarianism, impulsivity, openness, each may be useful, yet each is drawn over subtler organizations involving attention, threat, temporal discounting, status, uncertainty, imagination of other minds, memory persistence, inhibition, and context.

Psychology is full of constructs that split, merge, rotate, and fall out of fashion because the field is drawing a visible topology over a manifold of distinctions much finer than a gross statistical method can reliably isolate. That is not an argument to discard psychometrics; it is the shortcut into GIDS.

A psychometric vector

\[
\boldsymbol\psi_i\in\mathbb R^k
\]

can provide a coarse prior over an actor, placing the person in a region of likely response; it is not the person. A factor score is not a memory field, a role, a present state, a proposition, or a transition rule; it is a prior with delusions of grandeur.

The GIDS objective is to push beneath the named conglomerations and recover finer distinctions that remain useful across tasks. Some will look psychological, some linguistic, relational, institutional, temporal, or behavioral, while some will exist only as interactions; a coordinate earns attention because it transfers, not because a human found a satisfying noun for it.

## Why the Axes Stay Vague

The manuscript will remain intentionally vague about the specific axes already discovered in implementation, not because the system has no content behind the notation and not because I want mystery for its own sake, but because the coordinate registry—the distinctions, source types, regime conditions, interactions, split rules, and transformations that repeatedly predict actor response—is both the commercial object and the scientific object. Publishing the strongest families would amount to publishing the shortcut.

The public approximation is psychology plus factor analysis, made dynamic and far less polite:

- traits are priors rather than essences;
- source channels remain separate until evidence justifies comparison;
- role and institution modify the proposition–actor interaction;
- recent events alter fast state;
- memory is retrieved under the proposition;
- coordinates may be split, merged, rotated, weakened, restricted, or retired;
- transfer across tasks determines whether a distinction deserves to survive.

The public implementation details are therefore structural. A reader should be able to reproduce the discovery method, the event discipline, the actor construction, and the tests without receiving the current answer key; the concealed ontology is not evidence for itself, and public empirical work must eventually establish that reusable structure exists without requiring the strongest proprietary coordinates to be printed in the manuscript.

## The Realized Individual

Let the inherited seed be \(G_i\); the realized actor at time \(t\) is

<a id="gids-e10"></a>
**GIDS–10 — Realization of one actor.**

\[
T_{i,t} =
\mathcal E_{\mathrm{ind}}
\!\left(
G_i,
\mathbf u_{i,<t}^{\mathrm{lang}},
\mathcal H_{i,<t}^{\mathrm{life}}
\right).
\]

The equation says that a person is the inherited starting organization after language, culture, and life history have transformed it; the history is indexed before \(t\) because the actor at a decision cannot contain evidence that has not happened yet.

Life history is a sequence, not a bag. A crude weighted view may be written

\[
\overline{\mathbf h}_{i,t} =
\sum_{r<t}
\beta_{i,r,t}\mathbf v_{i,r}^{\mathrm{event}},
\qquad
\beta_{i,r,t}\ge0,
\]

where \(\mathbf v_{i,r}^{\mathrm{event}}\) represents event \(r\) and \(\beta_{i,r,t}\) its later force. Some events fade, some return only under a similar proposition, and some bend the later space of response so thoroughly that every subsequent event is interpreted through them; the weighted sum makes unequal force visible, while a sequence model must preserve chronology whenever order matters.

Durable evidence may be collected schematically as psychometric summaries, biography, language and cultural position, role history, life-event structure, and a slow bank of categorical traces; an encoder then produces the first slow estimate,

\[
\widehat{\mathbf t}_{i,t} =
E_{T,\theta}(\mathbf u_{i,t}^{\mathrm{dur}}).
\]

A deliberately boring additive baseline is useful before a grand interaction architecture is allowed to hide whether any source contained signal at all; its full form is retained in a footnote rather than displayed as though addition itself were a scientific result.[^additive-encoder]

This is the model's approximation of the slowly changing realized Transcendental Embedding, and should be read as **what this actor is generally like now**, not as a timeless soul-vector.

## Slow Structure and Fast State

A founder does not become another founder because one email arrived, although their local state can change completely because of one email; the operational actor-state therefore separates clocks:

<a id="gids-e11"></a>
**GIDS–11 — Slow and fast operational state.**

\[
\widehat s_{i,t} =
\left(
\widehat{\mathbf t}_{i,t},
\mathbf z_{i,t},
\mathbf c_{i,t},
\mathbf w_t
\right).
\]

The slow term approximates what the actor is generally like now; the fast term carries what recent chronology has made active; context contains role and institutional position; world-state contains the measured external conditions relevant to the decision. Keeping context and world explicit prevents the model from calling every changed situation a changed person.

Let \(\mathsf h_n\) be the \(n\)-th chronological record arriving at time \(v_n\); the fast state updates through

\[
\mathbf z_{i,n} =
U_{z,\theta}
\!\left(
\mathbf z_{i,n-1},
\widehat{\mathbf t}_i(v_n^-),
\mathbf c_i(v_n^-),
\mathbf w(v_n^-),
\mathsf h_n,
\mathbf d_{i,n}^{\mathrm{rec}}
\right),
\]

where \(v_n^-\) means immediately before the record arrives, and the applicability vector tells the updater whether the record pertains to this actor and which fields are present. An irrelevant record leaves the actor unchanged.

At decision time \(t\), use only records timestamped before the decision:

\[
N(t) =
\#\{n:v_n<\operatorname{time}(\mathsf d_t)\},
\qquad
\mathbf z_{i,t}=\mathbf z_{i,N(t)}.
\]

This is not glamorous; it is also where many impressive systems quietly cheat.

## Relevance, Salience, and the Active Person

The whole actor does not respond uniformly to every proposition, because a present transition is normally governed by a smaller weighted slice. Let a learned relevance map produce potentially active structure and a salience gate weight which parts matter now:

\[
\widehat{\mathbf r}_{i,t}^{(\tau)}(x_t) =
\boldsymbol\lambda_{\theta,\tau}(\widehat s_{i,t},x_t)
\odot
\Lambda_{\theta,\tau}(\widehat s_{i,t},x_t).
\]

The elementwise product \(\odot\) says that the model represents potentially relevant structure and then weights its present activation; role is inside this interaction, memory is inside this interaction, the proposition is inside this interaction, and the same actor-state confronted with another proposition may expose another Chimera entirely.

The task is not to discover one permanent principal axis of the person; it is to discover which distinctions and couplings carry the transition under these conditions, then force those discoveries to survive somewhere else. The next chapter replaces the complete phenomenal state with a predictive object defined by consequences, which is where the philosophy stops merely suggesting an architecture and begins constraining what the state estimator is allowed to preserve.

---

[^recap-actor-descent]: This is [GIDS–2 — The actor descent](00_Opening.md#gids-e2), repeated because the individual-actor chapter must remain locally readable; no new mathematical claim is introduced.
[^factor]: For the historical development of factor analysis, see Charles Spearman, “General Intelligence, Objectively Determined and Measured,” *American Journal of Psychology* 15 (1904), 201–292; and L. L. Thurstone, *Multiple-Factor Analysis* (University of Chicago Press, 1947). Rotational indeterminacy is one reason named axes must remain provisional.
[^additive-encoder]: A first-pass debuggable encoder may align the public source blocks into one width and add them: \(\widehat{\mathbf t}_{i,t}^{(0)}=W_\psi\boldsymbol\psi_i+W_b\mathbf b_i+W_{\mathrm{lang}}\mathbf u_{i,t}^{\mathrm{lang}}+W_{\mathrm{role}}\mathbf v_{i,t}^{\mathrm{role}}+W_h\overline{\mathbf h}_{i,t}+W_g\mathbf g_{i,t}^{\mathrm{slow}}\). This is an implementation baseline, not a claim about the true algebra of a person.

# Predictive Actor-State

The Chimera chapter separated the slowly realized actor from fast local state, then made role and proposition part of the interaction rather than treating a person as one timeless factor score; this chapter now asks the more severe question, which is not whether the complete mental interior can be recovered, but what information must survive compression if we want to predict the actor under a declared family of future propositions.

The operational state introduced earlier is repeated here because the predictive construction must remain legible without sending the reader backward through another file.[^recap-operational-state]

<a id="gids-e11-recap-predictive"></a>
**GIDS–11 — Slow and fast operational state, restated.**

\[
\widehat s_{i,t} =
\left(
\widehat{\mathbf t}_{i,t},
\mathbf z_{i,t},
\mathbf c_{i,t},
\mathbf w_t
\right).
\]

This is the finite object estimated from traces; it is not the complete phenomenal state, nor is it automatically sufficient merely because we gave its components impressive names.

## The Notion of State

Philosophically, everything belongs inside state. The actor's present state includes perception, interoception, memory, attention, anticipation, language already moving through the mind, action tendencies, action already underway, and whatever else is available to the actor at that instant; call the complete phenomenal state

\[
\phi_{i,t}\in\Phi_i.
\]

That is the motivating object, and it is inaccessible. A transcript is not the thought that produced it, a click is not the desire, a psychometric score is not the person, and even perfect outward logging would leave many internal routes observationally equivalent; therefore, if I point the engineering section directly at \(\phi_{i,t}\), I begin lying almost immediately.

The operational question is instead:

> What must be preserved about the actor's available history so that future response under admissible propositions can be predicted?

This question has a mathematical answer even when the complete interior does not.

## The General Predictive Response Object

Define the ideal pre-proposition information state

\[
\mathsf I_{i,t} :=
\left(
\mathcal H_{i,<t},
T_{i,t},
 c_{i,t},
 w_t
\right).
\]

The terms are the history available before decision \(t\), the slowly changing realized actor, present context, and the relevant world; the complete phenomenal state is not granted to the predictor, because if it were, much of the problem would disappear by definition.

For a finite horizon \(H\), let \(\mathbf x_{t:t+H-1}\) be an admissible sequence of propositions and \(O_{i,t+1:t+H}\) the future observable traces. The response law is

<a id="gids-e12"></a>
**GIDS–12 — Proposition-conditioned response law.**

\[
\mathscr R_{i,t}^{(H)}
\!\left(
\cdot
\mid
\mathsf I_{i,t},
\mathbf x_{t:t+H-1},
\boldsymbol\xi_{t+1:t+H}
\right) :=
\mathcal L
\!\left(
O_{i,t+1:t+H}
\mid
\mathsf I_{i,t},
\mathbf X_{t:t+H-1}=\mathbf x_{t:t+H-1},
\boldsymbol\Xi_{t+1:t+H}=\boldsymbol\xi_{t+1:t+H}
\right).
\]

The notation is dense; the idea is not. Hold one actor-information state fixed, present a possible proposition sequence, supply or model a future external scenario, and ask for the distribution of traces that follow. The admissible family must be declared, because “every possible proposition under every possible universe” is not a usable scientific object.

Two information states are predictively equivalent when every admissible proposition sequence produces the same family of response laws:

<a id="gids-e13"></a>
**GIDS–13 — Predictive equivalence.**

\[
\mathsf I\sim\mathsf I'
\iff
\mathscr R^{(H)}(\cdot\mid\mathsf I,\mathbf x,\boldsymbol\xi) =
\mathscr R^{(H)}(\cdot\mid\mathsf I',\mathbf x,\boldsymbol\xi)
\]

for every declared horizon, proposition path, scenario regime, and measurable future event in the family under study. The equivalence class is the ideal **general predictive actor-state**, denoted \(Q_{i,t}\).

This is the cleanest formal answer to the state question. If two actor histories differ in a thousand details but imply the same response law under every proposition we care to present, those differences do not belong in the minimal predictive state; if one forgotten humiliation changes one response family ten steps later, it belongs. The object may be infinite-dimensional, and there is no promise that every actor can be compressed into one convenient finite vector without loss; the paper becomes much more honest the second this object, the phenomenal state, and the estimate stop being treated as the same thing.

For a narrower task \(\tau\) and horizon \(\Delta\), a task-conditioned summary may exist:

\[
q_{i,t}^{(\tau,\Delta)} =
\Pi_{\tau,\Delta}(Q_{i,t}).
\]

The map keeps what one task and horizon require, and need not be linear; a seven-day response prediction may discard structure required for a five-year succession decision. For every measurable outcome event \(B\), the summary is sufficient when

<a id="gids-e13a"></a>
**GIDS–13A — Task-conditioned sufficiency.**

\[
\mathbb P
\!\left(
Y_{i,t}^{(\tau,\Delta)}\in B
\mid
\mathsf I_{i,t},X_t=x
\right) =
\mathbb P
\!\left(
Y_{i,t}^{(\tau,\Delta)}\in B
\mid
q_{i,t}^{(\tau,\Delta)},X_t=x
\right).
\]

Once the proposition and task-summary are known, the richer information state contributes nothing further to this outcome under the declared regime; the causal version replaces observational conditioning with the corresponding intervention-indexed laws. The general state \(Q_{i,t}\), however, is judged against the broader declared family of admissible proposition-conditioned response laws, which is where cross-task transfer enters at the foundation rather than as a decorative secondary metric.

## Predictive State Representations, and the Difference

This construction is closely related to predictive state representations: represent a latent dynamical state through predictions of future observable tests under possible actions.[^psr] The resemblance should be stated because it gives the manuscript a real mathematical neighbor; the difference is not that PSRs manipulate vulgar worldly objects while GIDS manipulates holy mental ones, the difference lies in the actor-relative construction placed before predictive state.

In GIDS, the externally modeled proposition, the proposition reconstructed by the actor, the conscious or actor-available state, the memory it recruits, the thought or movement it produces, and the state that follows all enter one actor-relative transition grammar. We keep types so the implementation does not perform nonsense arithmetic, but we do not place an ontological abyss between observation, thought, feeling, and action once they are inside the actor's phenomenal loop; each is an organized difference changing what becomes available next, with outward action distinguished mainly because another observer can register more of its consequences.

The second difference is transfer. A narrow predictive state may be sufficient for one controlled dynamical system; GIDS wants a reusable actor ontology, a representation of response structure surviving changes in proposition family, role, horizon, and task. That ambition is harder and may fail, although it is also the reason the latent object matters more than one forecast. No theorem is inherited merely because the objects look related; existing results apply only when their assumptions match the actor system constructed here.

## Minimality Without a Soul Coordinate

Call a general predictive state \(Q_{i,t}\) minimal when every other state \(R_{i,t}\) preserving the same declared family of proposition-conditioned response laws contains enough information to recover it:

\[
Q_{i,t}=h(R_{i,t})
\quad\text{almost surely}
\]

for some measurable map \(h\). This does not identify one sacred coordinate chart; rotations, invertible transformations, and more complicated reparameterizations may preserve every relevant response law, therefore the ontology is not the literal name attached to each axis, it is the stable structure of distinctions and relations needed to preserve response across tasks.

Human-readable names remain useful handles. They are not divine certificates.

## What Approximation Means

The model estimates \(\widehat s_{i,t}\) from records available before the proposition, while the ideal information state remains richer. Approximation should therefore mean more than drawing a wavy line between two symbols.

For outcome \(Y_{i,t}^{(\tau,\Delta)}\), define the information lost by state compression as

\[
\epsilon_{\tau,\Delta}^{\mathrm{state}}(\widehat s) :=
I
\!\left(
Y_{i,t}^{(\tau,\Delta)};
\mathsf I_{i,t}
\mid
\widehat s_{i,t},X_t
\right).
\]

Read it this way: after the model knows the operational state and the proposition, how much additional information about the future remains hidden in the richer ideal state? Zero means the compression was sufficient for this outcome under the declared observational regime.

A fitted predictor may still misuse a sufficient state. Define the model-estimation gap

\[
\epsilon_{\theta,\tau,\Delta}^{\mathrm{model}}(\widehat s) :=
\mathbb E
\!\left[
D_{\mathrm{KL}}
\!\left(
\mathbb P(Y\in\cdot\mid\widehat s_{i,t},X_t)
\;\middle\|\;
P_{Y,\theta,\tau,\Delta}(\cdot\mid\widehat s_{i,t},X_t)
\right)
\right].
\]

The first error asks whether the state threw useful information away; the second asks whether the predictor used the retained information correctly. “The model was bad” is too imprecise to be useful, because better training cannot recover information the state discarded, while a richer state does nothing if the predictive head cannot use it.

<a id="gids-e14"></a>
**GIDS–14 — State error plus model error.**

Under ordinary log-loss regularity,

\[
\mathcal R_{\log}(P_{Y,\theta}\circ\widehat s)
-
\mathcal R_{\log}^{\star} =
\epsilon_{\tau,\Delta}^{\mathrm{state}}(\widehat s)
+
\epsilon_{\theta,\tau,\Delta}^{\mathrm{model}}(\widehat s).
\]

This equation earns its place because it tells us where to look. It is related to the information-bottleneck idea of compressing one variable while preserving what matters for another; GIDS asks for a broader preservation problem across a family of proposition-conditioned futures rather than one target alone.[^ib]

## Memory Is a Field of Weighted Traces

Memory need not begin as narrative. For the model, it may begin as a field of traces with changing weights:

\[
\mathbf m_{i,t}^{\mathrm{mem}} =
\sum_{j=1}^{N_i}
\varpi_{ij,t}\mathbf h_{ij}^{\mathrm{mem}}.
\]

Each trace representation carries a present availability or force; some decay, some repeat until they become structure, and some remain dormant until a proposition resembles the original event. A proposition-conditioned retrieval rule changes those weights,

\[
\widetilde\varpi_{ij,t}(x_t) =
\mathcal R_{\mathrm{ret},\theta}
\!\left(
\varpi_{ij,t},
\mathbf h_{ij}^{\mathrm{mem}},
\widehat s_{i,t},
 x_t
\right),
\]

then forms the local retrieved past

\[
\widetilde{\mathbf m}_{i,t}^{\mathrm{mem}}(x_t) =
\sum_j
\widetilde\varpi_{ij,t}(x_t)\mathbf h_{ij}^{\mathrm{mem}}.
\]

The present proposition does not consult the whole archive evenly; it retrieves a local past, and after the actual trace arrives the memory field changes again. The second encounter is therefore never with exactly the same actor-state as the first, because the first encounter has joined the actor.

A recommender system supplies the crude engineering analogy. A view history is not a mind, yet it demonstrates that repeated traces can be compressed into a latent object that improves prediction; GIDS takes the move seriously enough to separate source, role, memory, durable actor structure, and proposition-conditioned retrieval.

## Categorical Traces and the Registry

A great deal of useful evidence arrives categorically: roles, recurring topics, objection families, action types, counterpart identities, product themes, price postures, institutional regimes, and source channels. Before pooling, the model should lift surface labels through context,

\[
\widetilde{\mathcal B}_{i,r}^{(f,\sigma)} =
\operatorname{Lift}_{\mathrm{ctx}}
\!\left(
\mathcal B_{i,r}^{(f,\sigma)},
 c_{i,r}
\right),
\]

because “aggressive” toward a competitor, “deferential” toward a regulator, and “protective” toward a subordinate may express one deeper organization under different relations, while collapsing them at ingestion manufactures contradiction out of context.

Source remains explicit. Biography, stated language, observed behavior, and third-party inference are not merged merely because they share a label; a person may describe themselves as cautious, behave recklessly, and be described by others as calculating, and the disagreement is evidence. Slow categorical memory pools durable evidence by role and regime; fast retrieval emphasizes recent proposition-relevant traces; weighting may depend on recency, repetition, source reliability, action intensity, regime similarity, and estimated susceptibility. Missing cells use learned null representations and explicit masks, while count or evidence-mass terms keep one exposure from becoming numerically identical to twenty repeated exposures.[^categorical-pooling]

The registry behind the machinery is the current typed hypothesis over families, sources, regimes, interactions, masks, and reparameterizable latent blocks. It is not the species projection introduced earlier; it is the finite catalog of distinctions the current model knows how to ask about, revised under cross-task transfer. The strongest discovered content remains deliberately vague, while the method, chronology, and test conditions remain public.

## Identifiability and the Right Kind of Modesty

A learned actor-state can be rotated, rescaled, or reparameterized while preserving every prediction, which is not fatal; it means the empirical target is stable predictive information rather than one blessed coordinate chart. The slow/fast decomposition is more substantive because it predicts different failure patterns: slow state should help under sparse observation, role transfer, and longer horizons, while fast state should help after recent events and over shorter horizons; if both disappear without consequence, the decomposition has failed for that actor class.

The standard is not metaphysical proof. A distinction must persist where it should persist, change where it should change, transfer where it claims to transfer, and—when intervention data exists—participate in the predicted change under a proposition. At this point the actor can be written at several resolutions, from inherited seed to realized actor, phenomenal state, general predictive state, and operational estimate; the next move is to construct actors larger than one person without committing the obvious sin of averaging everyone together.

---

[^recap-operational-state]: This is [GIDS–11 — Slow and fast operational state](03_The_Chimera.md#gids-e11), repeated because the predictive-state chapter must be locally readable; no new mathematical claim is introduced.
[^psr]: Michael L. Littman, Richard S. Sutton, and Satinder P. Singh, “Predictive Representations of State,” *Advances in Neural Information Processing Systems 14* (2001), 1555–1561; see also Satinder Singh, Michael R. James, and Matthew R. Rudary, “Predictive State Representations: A New Theory for Modeling Dynamical Systems,” *Proceedings of UAI 2004*, 512–519.
[^ib]: Naftali Tishby, Fernando C. Pereira, and William Bialek, “The Information Bottleneck Method,” *Proceedings of the 37th Annual Allerton Conference on Communication, Control, and Computing* (1999); arXiv:physics/0004057.
[^categorical-pooling]: A compact implementation keeps family and source typed. For lifted bag \(\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}\), use \(\mathbf u_{i,r}^{(f,\sigma)}=|\widetilde{\mathcal B}|^{-1}\sum_{\upsilon\in\widetilde{\mathcal B}}E_{f,\sigma}(\upsilon)\) when nonempty and a learned \(\mathbf e_{\varnothing}^{(f,\sigma)}\) otherwise; concatenate an aligned representation with an availability mask and \(\log(1+|\widetilde{\mathcal B}|)\). Slow banks pool these cells separately by role or regime with durable weights \(\beta\); fast retrieval uses task-conditioned weights \(\alpha\) and retains total relevance mass. This preserves source, missingness, repetition, and regime without promoting the pooling arithmetic into a central theorem.

# Composite Actors

The individual actor is not the only useful unit of prediction, because the boundary of an actor is chosen relative to the transition and trace under study; an ant moves under local chemical gradients, contact, hunger, alarm, and whatever else an ant can register, while the colony moves under distributed paths, food reserves, brood state, nest condition, caste, traffic, and the aggregate consequence of many local actors whose behavior contains both stable pattern and randomness. The ant is an actor; the colony is also an actor; neither statement cancels the other.

The same principle extends into institutions. A corporation contains people acting with varying independence, yet the corporation also persists, remembers, receives information, filters it, makes decisions, and changes the world under one operational boundary; an employee sends an email, a committee rejects a contract, the corporation exits a market, and these are actions at different scales produced by actor constructions nested inside one another.

The outer grammar is repeated here because the corporate derivation depends on it.[^recap-composite-grammar]

\[
\text{access}
\rightarrow
\text{organization}
\rightarrow
\text{state}
\rightarrow
\text{proposition}
\rightarrow
\text{transition}
\rightarrow
\text{trace}.
\]

## An Actor Is a Choice of Scale

The useful actor is selected relative to the trace and decision being modeled. If the outcome is whether one executive replies, model the executive inside the institution; if the outcome is whether the company acquires another company, the executive becomes one component inside a corporate actor; if the outcome is a market transition, several corporations may become components of a larger system whose state cannot be reduced to any one member.

Whatever can be traced at a declared level of abstraction may be treated as the action of the actor constructed at that level, provided the construction has a persistent boundary, selective access to information, an organization through which information becomes consequential, memory, and repeatable outward transition. Statistical predictability alone is not enough, because a random collection of correlated things is not thereby one actor; the boundary and transition grammar must explain why the components jointly produce the trace.

## The Organization Is a Synthetic Object Before It Is a Composite Actor

An organization exists first as a synthetic construction inside the realities of the people who participate in, observe, fear, regulate, exploit, or depend on it. “The company” is not one object copied into every mind; a founder, employee, child, creditor, regulator, or person with a radically different affective structure may construct entirely different organization-objects from the same legal entity, because each receives different information, assigns different salience, and possesses a different relationship to authority, obligation, danger, and belonging.

The institution nevertheless becomes more than a collection of private feelings when those actor-relative constructions are coupled through records, routines, communication channels, authority, incentives, assets, laws, and recurring consequences; the people act through what they believe the organization is, those actions alter the shared synthetic structure, and the resulting pattern persists long enough to confront later propositions as one higher-order actor. The corporation is therefore both a phenomenal object inside individuals and a composite phenomenal actor at another scale, where *phenomenal* means the organized reality made available to the composite through its members and institutional machinery, not a declaration that the corporation contains one human-like stream of consciousness.

## Institutional Access and Interpretation

Part 1 described an actor as an access map plus an interpretive organization; a corporation follows from the same principle, because a fact cannot influence a corporate decision merely by existing somewhere in the building, it must reach the relevant institutional channel and survive whatever authority, delay, incentive, and distortion governs its path.

Let \(A_{C,t}^{\mathrm{inst}}\) denote the corporation's **institutional access architecture**: what reaches the organization, through which channel, with what delay, after which filters, and under whose permissions. A fact known by an intern but blocked from the decision-maker may exist in the corporation's environment without entering the state relevant to the decision; a dashboard, board packet, customer escalation, legal opinion, rumor, and private conversation create different paths of access even when they concern the same external condition.

Let \(\mathcal I_{C,t}^{\mathrm{inst}}\) denote the corporation's **institutional interpretive organization**: governance, incentives, authority, procedure, culture, coalition structure, veto rights, active objectives, and the formal or tacit rules by which information becomes action.

<a id="gids-e15"></a>
**GIDS–15 — Institution-relative object construction.**

\[
\zeta_{C,t}^{\mathrm{inst}}(\omega_t) =
\mathcal I_{C,t}^{\mathrm{inst}}
\!\left(
A_{C,t}^{\mathrm{inst}}\operatorname{Reg}(\omega_t),
S_{C,t^-}
\right).
\]

Read it exactly as the human equation. Something happens outside the institution; institutional channels make some part of it available; the existing corporate state and organization determine what the event becomes for the company. A regulatory notice may become an existential threat, an ordinary compliance task, an opportunity to crush smaller competitors, or nothing at all because it died in the wrong inbox; the document is the same, while the institutional object is not.

This construction makes the composite actor depend on the philosophy rather than appearing later as a convenient aggregation box. Information architecture is the institutional access map; governance, incentive, and authority are the interpretive organization.

## Constructing the Corporate State

Let \(\mathcal J_C(t)\) be the people relevant to corporation \(C\) at time \(t\), and let \(\varsigma_{j,t}^{\mathrm{person}}=(T_{j,t},\phi_{j,t},c_{j,t})\) be the conceptual person-state contributed by member \(j\). The corporate construction must also contain facts and statistics, authority and communication structure, institutional memory, incentives and active constraints, and environmental history already absorbed by the institution.

<a id="gids-e16"></a>
**GIDS–16 — Composite institutional state.**

\[
S_{C,t} =
\mathcal A_{\mathrm{corp}}
\!\left(
(\varsigma_{j,t}^{\mathrm{person}})_{j\in\mathcal J_C(t)},
\mathsf{Facts}_{C,t},
\mathsf{Org}_{C,t},
\mathsf{Mem}_{C,t},
\mathsf{Inc}_{C,t},
\mathsf{EnvHist}_{C,t}
\right).
\]

The aggregator \(\mathcal A_{\mathrm{corp}}\) is not an average, it is the formal interface for the institutional access and interpretation process just described; it must represent unequal authority, missing information, coalitions, vetoes, delegated action, procedural delay, and the fact that most employees contribute nothing to most decisions. The equation does not explain those mechanisms by itself, although it prevents an implementation from quietly replacing them with mean-pooling and continuing to call the result a corporation.

The state is not limited to formal decisions. Any persistent institutional change may matter: a new internal belief, changed risk posture, budget reallocation, relationship fracture, new champion, lost executive, policy, hidden queue of work, or market action; what counts depends on the scale of the actor and the traces available.

A corporation acquires an almost feel-like way of dealing with the world, not necessarily human consciousness but a characteristic pattern of access, memory, salience, delay, and response. Some companies absorb bad news early and act; some route it upward until it has been sanded into harmlessness; some experience every external proposition through one founder; some possess no center until crisis supplies one. That pattern is part of the corporate phenomenal state at the scale we have chosen.

## Partial Observation Is a Projection Problem

The complete corporate state is never observed. We may see three people in a buying committee of eight, one official org chart, stale funding data, a few public statements, and the traces produced in our own relationship; the missingness is structural, therefore it belongs in the state estimate instead of being silently converted into zero.

Let \(\mathcal J_C^{\mathrm{obs}}(t)\subseteq\mathcal J_C(t)\) be the members we can use, let \(\mathbf u_{C,t}^{\mathrm{corp}}\) collect measurable institutional evidence, and let \(\mathbf m_{C,t}^{\mathrm{miss}}\) encode missing membership, authority, and fields. The estimate is

\[
\widehat S_{C,t} =
\mathcal A_{\mathrm{corp},\theta}
\!\left(
(\widehat s_{j,t}^{\mathrm{person}})_{j\in\mathcal J_C^{\mathrm{obs}}(t)},
\mathbf u_{C,t}^{\mathrm{corp}},
\mathbf m_{C,t}^{\mathrm{miss}}
\right).
\]

A missing executive is not a zero-vector executive, unknown authority is not equal authority, and an unseen veto player should widen uncertainty rather than disappear. Conceptually, observability is a projection from complete institutional state into available traces,

\[
\mathcal O_{C,t}:S_{C,t}\longrightarrow\mathcal H_{C,\le t}^{\mathrm{obs}},
\]

while the estimator attempts an inverse problem under uncertainty and never literally reconstructs everything the projection destroyed.

## Relationship as a State of Its Own

Two actors create structure belonging to neither actor alone. Let \(\Gamma_{ab,t}\) denote their relationship—familiarity, trust, perceived authority, prior commitments, objections, channel history, response rhythm, resentment, debt, affection, status, and whatever else the interaction has accumulated.

A relationship is not merely actor \(a\)'s opinion of actor \(b\) plus actor \(b\)'s opinion of actor \(a\); it contains a sequence of mutual transitions, because one promise changes the meaning of the next promise, one betrayal changes the meaning of silence, and the relationship becomes an object precisely because it persists and alters future response.

Estimate it from pre-proposition history,

\[
\widehat\Gamma_{ab,t} =
E_{\Gamma,\theta}(\mathcal H_{ab,<t}^{\mathrm{rel}}),
\]

then update it when relevant records arrive. Chronology matters; a later apology cannot leak backward into the relationship-state used to predict the insult.

## The Dyadic Actor

The first practical world model uses a dyad. Let \(a\) be a sender or initiator, \(b\) the receiving actor, and \(C_a,C_b\) the institutions through which they act.

<a id="gids-e17"></a>
**GIDS–17 — Filtered dyadic state.**

\[
\widehat D_{ab,t} =
\left(
\widehat s_{a,t}^{\mathrm{person}},
\widehat S_{C_a,t},
\widehat s_{b,t}^{\mathrm{person}},
\widehat S_{C_b,t},
\widehat\Gamma_{ab,t},
\mathbf w_t
\right).
\]

The dyad contains two people, two institutional contexts, one relationship, and a shared world; a commercial interaction is one laboratory, while the same state type can represent negotiation, recruiting, diplomacy, command, care, instruction, or any repeated proposition-response process involving two sides.

There is deliberate redundancy, because the focal people may already contribute to their company representations; an implementation must either exclude them from the corresponding company-context summary or estimate the coupled states jointly, otherwise the same evidence is counted twice and called confidence.

A proposition may contain content, offer, framing, evidence, channel, timing, sender, and requested action, although this is not a universal schema; the proposition could instead be a policy, threat, candidate, medical intervention, educational sequence, or arrangement of physical conditions. What matters is that it enters the actor through an actor-relative representation.

The dyad is not a new metaphysical species, it is the convenient actor construction at the scale where relationship, institutions, and proposition jointly determine the next trace. The internal construction changes with the actor, while the outer algebra remains stable; that is how realities become composable without becoming the same.

The next chapter takes this composite state and does the dangerous thing: it introduces candidate propositions, simulates what follows, then asks which future should be preferred.

---

[^recap-composite-grammar]: The access–organization–state–proposition–transition–trace sequence restates the common grammar established in [God's Infinite Dimensional Space](02_Gods_Infinite_Dimensional_Space.md); it is repeated here to derive the institutional actor locally, and no new mathematical claim is introduced by the recap itself.

# World Models and Proposition Search

The previous chapter assembled people, institutions, relationships, and world-state into a dyadic actor; this chapter now turns the construction into a recursively usable world model, which means the model must do more than predict one visible label, it must return a next state of the same type required by the following transition, otherwise the alleged world model is merely a decoder with delusions of grandeur.

The operational loop is repeated here because it is the object being constructed rather than merely summarized.[^recap-operational-loop]

<a id="gids-e3-recap-world-model"></a>
**GIDS–3 — Estimation, simulation, and trace, restated.**

\[
\mathcal H_{<t}\longmapsto\widehat D_t,
\qquad
(\widehat D_t,X_t=x_t,\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1})
\longmapsto
\widetilde D_{t+1}
\longmapsto
\widetilde O_{t+1}.
\]

The history available before the decision produces an estimated state; a candidate proposition encounters that state under a declared external scenario; the model simulates a next state and then the traces likely to escape. A hat means estimated from evidence, a tilde means simulated, and confusing the two is how a forecasting system begins hallucinating its own outputs into the database.

## Towards a Universal State-Transition Grammar

The ideal object remains the next phenomenal state. Let the complete actor-world state be

\[
\Sigma_{i,t}^{\star} :=
(T_{i,t},\phi_{i,t},c_{i,t},w_t),
\]

and recall the asymptotic silhouette,

\[
\phi_{i,t+1} =
F_i
\!\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
\mathbf p_{i,t}(x_t)
\right).
\]

If the complete state and exact law were available, the next phenomenal state would follow; the operational model possesses neither, therefore it works with an estimated state and a conditional distribution,

<a id="gids-e18"></a>
**GIDS–18 — Recursively usable state transition.**

\[
\widetilde s_{i,t+1}
\sim
K_{s,\theta}
\!\left(
\cdot
\mid
\widehat s_{i,t},
X_t=x_t,
\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right).
\]

The model receives the estimated actor-state, candidate proposition, and supplied or modeled exogenous change, then returns a distribution over the next state of the same operational type. This closure matters; a system predicting only “reply” or “no reply” cannot roll itself forward, because the next proposition will meet an actor changed by the first interaction, not the actor-state that existed before it.

The outward event is a partial consequence of the transition,

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}(\cdot\mid\widetilde s_{i,t+1}),
\]

where reply, purchase, rejection, delay, concession, departure, promotion, thought expressed in language, or physical movement is the part that escaped into observation rather than the complete state itself. If the simulated next state does not screen off the previous state and proposition well enough, the trace law can condition on all three; that is an empirical choice, not a theological dispute about Markovity.

## The Proposition Must Be Rebuilt Inside the Actor

Begin with an external representation of the proposition,

\[
\mathbf e_t^x=E_{x,\theta}(x_t),
\]

which may encode language, structure, source, timing, price, channel, physical arrangement, or whatever features the system possesses; this remains the proposition as represented **outside** the receiving actor.

For dyadic state \(\widehat D_{ab,t}\), construct the receiving actor's proposition-relative representation,

\[
\mathbf p_{b,t}^{(D)}(x_t) =
\mathcal P_{D,\theta}
\!\left(
\mathbf e_t^x,
\widehat D_{ab,t}
\right),
\]

then form the interaction,

<a id="gids-e19"></a>
**GIDS–19 — Actor-relative proposition interaction.**

\[
\mathbf h_{ab,t}^{(D,\mathrm{int})}(x_t) =
\Psi_{D,\theta}
\!\left(
E_{D,\theta}(\widehat D_{ab,t}),
\mathbf p_{b,t}^{(D)}(x_t)
\right),
\]

and predict

\[
\widetilde D_{ab,t+1}
\sim
K_{D,\theta}^{\mathrm{int}}
\!\left(
\cdot
\mid
\mathbf h_{ab,t}^{(D,\mathrm{int})}(x_t),
\boldsymbol\xi_{t+1}
\right).
\]

Read the construction from left to right: encode what was presented; reconstruct it through the receiving actor and surrounding dyad; model the interaction between that reconstructed proposition and present state; simulate what the dyad becomes next. Role is not appended after the proposition has already been understood, because the role, relationship, institutions, and current state participate in the understanding itself.

This is the practical meaning of making realities composable. The common arena does not make sender, recipient, company, price, sentence, thought, and outward action interchangeable; it gives each a typed route into one transition grammar, where thought and movement remain two possible continuations of phenomenal state rather than two disconnected ontologies.

Two physically different propositions may be equivalent for one actor and task when they produce the same interaction representation. A percentage discount and corresponding dollar discount may be economically identical yet psychologically different; two messages may be linguistically different yet psychologically identical because the actor compresses both into “vendor asking for more of my time.” Equivalence belongs to the actor-relative transition, not the physical proposition alone.

## Why Sales Became the First Laboratory

The first organizational system attempted something broader and less measurable: ask a corpus of employees for opinions, loosely embed the people and decision-space, weight their judgments into swarm intelligence, implement one proposal, then wait for the organization to reveal whether the collective answer was wise. It failed because opinions carried weak consequences, justifications were cheap, decisions dissolved into the surrounding company, and the measurement process itself imposed operational labor; the full account appears in the opening because the failure explains the architecture that replaced it.[^failed-swarm]

Sales became the first laboratory not because selling exhausts the theory, nor because human decision is deterministic, but because it offers a harder training instrument than organizational opinion. A proposition can be timestamped, the actor and institutional context can be reconstructed from prior traces, a response and relationship update can be observed, and economic value can eventually be attached to the trajectory; the environment remains noisy and complex, which is exactly why the model carries actor, relationship, company, and world-state instead of asking participants to predict themselves.

## A Worked Dyad

Suppose a founder, \(a\), presents a partnership to an executive, \(b\). The executive's slow state contains durable estimates—tolerance for ambiguity, status posture, temporal preference, skepticism toward founder-led vendors, and other unnamed distinctions inferred from prior traces—while the fast state contains what has become active recently: a failed implementation, board pressure, budget deadline, irritation with the sender, perhaps a newly urgent internal problem. The company-state contains authority, incentives, institutional memory, current priorities, and who can veto the decision; the relationship-state contains prior promises, response rhythm, trust, and whether the founder has already exhausted the executive's patience.

Compare two propositions. The first asks for a broad strategic commitment, the second asks for one narrow technical test; an abstract classifier may see the same product and target, while inside the executive the first recruits loss of control, political risk, and implementation memory, and the second recruits curiosity, reversibility, and a chance to gather evidence without public commitment. The model is not merely choosing the better sentence; it is predicting two different state transitions.

After the actual response arrives, relationship and fast state change. A polite refusal may lower immediate probability while increasing trust; a meeting acceptance may look mechanically positive while revealing that the executive delegated the matter to someone without authority. The next proposition must meet the updated dyad, not the state that existed before the response.

## Simulation Is Not Filtering

<a id="gids-e20"></a>
**GIDS–20 — Simulation and filtering.**

Before the future is observed, the model simulates:

\[
\widetilde D_{ab,t+1}
\sim
K_{D,\theta}
\!\left(
\cdot
\mid
\widehat D_{ab,t},
X_t=x_t,
\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right).
\]

After real records arrive, the model filters:

\[
\widehat D_{ab,t+1} =
\mathcal F_\theta
\!\left(
\widehat D_{ab,t},
 x_t,
\mathcal H_{(t,t+1]}
\right).
\]

Simulation asks what might happen; filtering changes what we believe because something did happen. A model-generated future should never be written back as though it were evidence, and an observed response should not be treated as though it were the full state; the simulator produces distributions, the filter consumes timestamped records.

For proposition sequence \(\mathbf x_{t:t+H-1}\) and exogenous path \(\boldsymbol\xi_{t+1:t+H}\), recursive application produces a trajectory law over future states and traces. The environment does not freeze because recursion is convenient; markets move, people leave, companies change, later propositions depend on earlier responses, and anything caused by the proposition belongs inside the transition rather than being smuggled into an exogenous background held fixed across candidates.

Delayed outcomes require their own declared regime. A ninety-day event depends on continuation policy, future propositions, external change, and censoring; it is not an immediate emission from tomorrow's state merely because the database stores it on the same decision row.

## Learning From What Escapes

The state should support more than one visible consequence. Primary outcomes and auxiliary probes—intermediate actions, objections, delays, institutional changes, or other traces—may be modeled jointly when their dependence matters, or through separate heads when only marginal prediction is required; separate heads do not magically define a coherent joint future.

The training objective may combine the relevant predictive losses, probe losses, regularization, censoring, and time-to-event terms, although the generic gradient-descent equation has been omitted because every reader who reached this chapter already knows how parameters move.[^training-objective] The important distinction is temporal: parameters learn across cases, filtering changes belief about this case, fast actor and relationship states update when new records arrive, and slow states refresh only when durable evidence accumulates.

Probe heads survive only when they improve transfer, calibration, or stability of the state; a decorative taxonomy of motives is not a scientific result.

## From Forecasting to Proposition Search

This is where I stop pretending the purpose of the machinery is to admire prediction metrics. The system should compare admissible propositions by their expected effects on future state and downstream utility; otherwise why the hell are we building it.

Let \(\mathcal X_t^{\mathrm{cand}}\) be the candidate set available at decision \(t\), let \(\mathfrak e\) declare the continuation policy, future candidate-set process, exogenous-path law, and outcome convention, and let \(U_\tau\) be a measurable utility over predicted trajectories.

<a id="gids-e21"></a>
**GIDS–21 — Predictive proposition value.**

\[
V_{\theta,\mathfrak e}^{\mathrm{pred}}
\!\left(x\mid\widehat D_t\right) :=
\mathbb E_{\theta,\mathfrak e}
\!\left[
U_\tau
\!\left(
\widetilde D_{t+1:t+H},
\widetilde O_{t+1:t+H},
\widetilde{\mathbf Y}_t
\right)
\mid
\widehat D_t,
X_t=x
\right].
\]

When the maximum exists over the candidate set, choose

\[
x_t^\star
\in
\arg\max_{x\in\mathcal X_t^{\mathrm{cand}}}
V_{\theta,\mathfrak e}^{\mathrm{pred}}(x\mid\widehat D_t).
\]

The equation says: simulate each available proposition under the same declared future regime, score the resulting trajectories, and choose the best supported candidate. Utility belongs to the operator or system using the world model; it is not a claim about the objective the actor internally optimizes, because I do not need to infer a universal fitness function, expected-free-energy objective, or hidden rational program inside the person. I need to estimate how the actor is likely to decide over a long horizon under different degrees of information, then evaluate those futures under a constrained combination of interests appropriate to the application.

Sales is one laboratory, not the definition of the machinery. The same form can rank educational interventions, negotiation moves, organizational policies, recruiting sequences, product experiences, care plans, or any other propositions for which actor transition matters.

## Sequences, Not Isolated Tricks

The best immediate proposition may be a poor first move in a sequence. A message producing no meeting may reveal uncertainty, a reversible test may change trust, a difficult question may expose the true veto, and a temporary concession may alter relationship-state enough to make a later demand possible; therefore credit belongs to the trajectory.

For a policy \(\pi\), planning length \(H\), and declared regime \(\mathfrak e\), the value is the expected sequence of step utilities plus terminal value under recursively simulated states.[^sequence-value] The conceptual rule is simple: do not copy one eventual success backward and award it independently to every proposition that preceded it; that is not learning, it is numerology with a CRM.

The sensible progression is modest:

1. rank controlled proposition families one step ahead;
2. compare short predefined sequences;
3. choose the next proposition after each observed response;
4. attempt longer policy optimization only when the state transition and data collection process deserve trust.

## Prediction, Ranking, and Control

Forecasting estimates what tends to follow the proposition actually delivered; model-based ranking simulates alternatives and orders them under the fitted world model; interventional policy improvement claims that selecting a proposition causes a better outcome, and that final claim requires an experimental or otherwise defensible identification design.

The distinction does not need forty pages of self-flagellation. Observational success licenses forecasting and simulation inside the observed support; causal swagger begins only when the data collection regime earns it.

Until then, leave the causal swagger out of it.

## What I Would Actually Build First

The theory does not require loyalty to one architecture. I would begin with components I can debug when calibration goes sideways at two in the morning: strong tabular models for durable and institutional features, a small recurrent or state-space model for chronological history, explicit relationship-state, typed proposition encoders, a simple interaction block, and completely separate simulation and filtering paths.

The proprietary advantage is unlikely to come from choosing the most fashionable sequence block; it comes from the ontology, event-clock, actor construction, source-aware traces, and discipline of discovering distinctions that transfer. A capacity-matched monolithic model should still see the same data, because if it wins, it wins; the explicit construction earns complexity through transfer, data efficiency, calibration, controllable recursion, or interpretability useful enough to change decisions.

No actor model remains correct forever. People change, companies change, roles change, the same distinction becomes active under a new regime, and a coordinate that once transferred may stop carrying signal; when error, calibration, or support deteriorates, reopen ontology discovery rather than merely refitting weights. The next chapter defines the event discipline, baseline opposition, and four tests that decide whether the construction survives contact with the future.

---

[^recap-operational-loop]: This restates [GIDS–3 — Estimation, simulation, and visible trace](00_Opening.md#gids-e3); it is repeated because this chapter constructs the loop in full, and no new mathematical claim is introduced by the recap itself.
[^failed-swarm]: See [The First Machine Failed, Which Was Useful](00_Opening.md#the-first-machine-failed-which-was-useful). The failure is repeated only in compressed form here to explain why the first world model uses observed consequence, explicit institutional state, and a hard event-clock rather than self-reported organizational opinion.
[^training-objective]: One implementation may minimize \(\mathscr J(\theta)=\sum_{\ell}\lambda_\ell^Y\mathscr J_\ell^Y(\theta)+\sum_m\lambda_m^Z\mathscr J_m^Z(\theta)+\lambda_{\mathrm{reg}}\Omega(\theta)\), with nonnegative weights and head-specific masking, censoring, or survival likelihoods. This is an implementation interface, not a mathematical contribution.
[^sequence-value]: One explicit form is \(J_{\theta,\mathfrak e}(\pi\mid\widehat D_t)=\mathbb E_{\theta,\pi,\mathfrak e}[\sum_{k=0}^{H-1}\gamma^k u_\tau^{\mathrm{step}}(\widetilde D_{t+k+1},X_{t+k},\widetilde O_{t+k+1})+\gamma^H V_\tau^{\mathrm{term}}(\widetilde D_{t+H})\mid\widehat D_t]\). The regime must also define the future candidate-set and exogenous processes; otherwise the sequence value is not a well-defined comparison.

# Evaluation and Ontology Growth

The philosophy explains why the model has this shape; data decides whether the shape survives. This chapter does not prescribe the company's current benchmark, expose the discovered ontology, or freeze the research program around one early sales dataset, because the canonical manuscript should define the evaluation discipline while the empirical paper supplies one public implementation, one corpus, one set of results, and all the humiliating detail required for replication.

The operational loop has already been derived, therefore it is not re-derived here; the evaluation problem begins at the instant an estimated state is frozen before a proposition, then asks whether the predicted transition survives future time, strong baselines, transfer, and intervention.[^recap-evaluation-loop]

## Event Time Is Sacred

The clock has to be clean or the entire evaluation becomes leakage with equations around it. Let \(t\) index decision epochs and let \(\mathsf h_n\) be timestamped records; only records fully available before proposition \(x_t\) was selected may enter the state used to forecast its consequences.

<a id="gids-e22"></a>
**GIDS–22 — Decision-aligned event time.**

\[
N(t) =
\#\left\{
 n:
\operatorname{time}(\mathsf h_n)
<
\operatorname{time}(\mathsf d_t)
\right\},
\qquad
\mathbf z_{i,t}=\mathbf z_{i,N(t)}.
\]

The equation is small and carries the entire leakage argument. If an email response arrives at 14:02, it cannot appear in the state constructed for a message selected at 14:00; if a transcript summary is generated after a meeting, it cannot enter the pre-meeting actor vector; if a company fact is learned three weeks later, the historical model does not get to know it early merely because the warehouse does now.

At every decision epoch: construct the actor or composite state from pre-decision history; record the candidate propositions actually available; choose and deliver \(x_t\); freeze the forecast; append observations only when they occur; attach delayed outcomes only when their horizons mature. Timestamp ties require an explicit order, because “same day” is not an event clock.

## Decisions, Observations, and Outcomes

A decision record should preserve the decision identity, candidate set, chosen proposition, elapsed time, and—when policy evaluation is intended—the assignment probability or density and information actually available to the historical policy. Observations arrive later and may include outward response, new actor evidence, institutional change, or world-state change; each enters the filtered state only when it becomes available.

A decision may carry several outcomes at different horizons, and every outcome requires an availability or censoring indicator. An outcome that has not matured is not a negative; it is unavailable. Long-horizon evaluation should therefore use complete follow-up windows or appropriate time-to-event methods instead of converting ignorance into failure.

The manuscript remains vague about the content of discovered actor axes and application-specific fields; it should not be vague about time, masks, source, candidate availability, or outcome maturity. Secrecy around the ontology is defensible; sloppiness around the dataset is not.

## Evaluate the Future, Not a Shuffled Past

Random row splits are nearly useless for repeated actor interactions, because they leak future actor-state into training, place the same relationship on both sides of the split, and reward memorization as though it were generalization. Use nonoverlapping temporal windows,

\[
\mathcal D_{\mathrm{train}}^{[1,T_1]},
\qquad
\mathcal D_{\mathrm{val}}^{(T_1,T_2]},
\qquad
\mathcal D_{\mathrm{test}}^{(T_2,T_3]},
\]

and make model choice, ontology changes, thresholds, and calibration decisions using training and validation only; the final test window remains untouched until the analysis is frozen.

Time is necessary and insufficient. Report transfer separately for later interactions with known actors, new actors inside known institutions, known actors under new roles or proposition families, entirely unseen actors and institutions, and new tasks or outcome horizons. The final regime matters most to the ontology claim, because a coordinate discovered for one target earns its place only when it preserves useful structure elsewhere.

Rows remain dependent within people, relationships, institutions, campaigns, and time periods; uncertainty estimates must respect those clusters. A million events generated by ten companies are not a million independent companies.

## The Models That Must Be Beaten

Every implementation should face at least three levels of opposition. The first is a shallow current-proposition model using the obvious predictors available now, with no durable actor-state; the second is a strong structured-history model using static actor and institution inputs plus competent hand-built summaries; the third is a capacity-matched monolithic sequence model seeing the same pre-decision records without being forced to separate slow actor, fast state, relationship, institution, and proposition interaction.

The monolithic model is the dangerous one. If it consistently predicts better, transfers as well, calibrates as well, and requires no more data, the explicit decomposition has become a story told after the fact; if the structured model ties on raw prediction while transferring better, remaining stable under sparse observation, or supporting reliable recursion, that may still justify it, although the trade must be stated rather than smuggled in as “interpretability.”

## Four Essential Tests

The earlier manuscript carried a ceremonial army of ablations; four are enough to decide whether the central claims are alive.

### 1. Slow versus fast state

Remove \(\widehat{\mathbf t}_{i,t}\), then remove \(\mathbf z_{i,t}\). Slow state should matter most under sparse observation, cold start, role transfer, and longer horizons; fast state should matter most after recent events and at shorter horizons. If their removal produces no distinguishable pattern, the decomposition is not buying what it claims.

### 2. Actor, relationship, and institution

Replace the structured composite state with flat metadata and shallow history; this tests whether relationship and institutional access structure contain signal that cannot be reduced to ordinary tabular features.

### 3. Explicit construction versus monolithic sequence

Give both models the same information and comparable capacity. This is the shortest route to discovering whether the ontology is useful or merely narratively satisfying.

### 4. Cross-task transfer

Discover or fit actor structure on one family of outcomes, then evaluate whether it improves another without being rebuilt from scratch. This is the decisive test for a new axis; a coordinate that helps only the target that created it is probably a target feature wearing philosophical makeup.

A fifth diagnostic is often cheap: shuffle recent within-actor or within-relationship history while preserving static profiles. If performance does not fall, the model was not using sequence in the way claimed.

## How a New Distinction Earns Its Place

Ontology growth begins with structured residual error. Suppose the current registry systematically fails for a subset of actors, roles, propositions, or regimes; propose an operation \(\delta\)—split, merge, rotation, interaction, source separation, role restriction, new family, weakening, or retirement—then ask whether it explains a stable residual pattern in training, improves future risk in validation, and transfers across held-out tasks, roles, horizons, or populations.

The transfer score is repeated from the ontology chapter because it becomes the evaluation rule here.[^recap-transfer]

<a id="gids-e7-recap-evaluation"></a>
**GIDS–7 — Cross-task ontology retention, restated.**

\[
\Delta_{\mathrm{tr}}(\delta) =
\sum_{\tau\in\mathcal T_{\mathrm{held}}}
\omega_\tau
\left[
\mathcal R_\tau(M_{\mathfrak R_n})
-
\mathcal R_\tau(M_{\mathfrak R_n\oplus\delta})
\right]
-
\lambda_C C(\delta)
-
\lambda_S S(\delta).
\]

Retain the distinction when the score is positive, the gain repeats across future time, and—where intervention data exists—the distinction participates in the predicted change under a controlled proposition. The name may change later; the predictive relation is what survived.

This is the model-side analogue of evolutionary accumulation, not the same mechanism. Evolution creates possible distinctions in organisms; ontology growth creates typed hypotheses about distinctions in the model, and the hybrid registry remains useful because it can be debugged even if a continuously updating manifold eventually captures the underlying geometry more faithfully.

## Policy Evaluation Without Causal Swagger

When assignment probabilities were logged at decision time and outcomes have matured, off-policy estimators may reweight historical decisions toward a target policy; the standard one-step inverse-propensity expression is retained in a footnote because it is an evaluation instrument, not part of the philosophical core.[^ips]

The conditions matter more than the formula: overlap, correct event ordering, stable treatment definition, defensible assignment, mature outcomes, censoring discipline, and support. Free-form propositions generally possess almost no exact historical overlap, therefore early policy evaluation should operate over controlled proposition families or dimensions instead of pretending every new paragraph has a counterfactual twin in the logs; sequential policies require sequential estimators, and one-step arithmetic should not be stretched across an entire conversation.

## Metrics Should Match the Claim

Use proper scoring rules for probabilistic forecasts, calibration measures for decision use, ranking metrics for candidate ordering, and survival or competing-risk methods for censored outcomes. No single number is the research program.

<a id="gids-e23"></a>
**GIDS–23 — Future-risk criterion.**

\[
\mathcal R_{\mathrm{future}}(M_{\mathrm{GIDS}})
<
\mathcal R_{\mathrm{future}}(M_{\mathrm{best\ baseline}})
-
\epsilon,
\qquad
\epsilon>0.
\]

Choose \(\epsilon\) before the final test window. “Statistically detectable” is not automatically “worth the machinery”; report gain size, uncertainty, calibration, data requirements, and transfer, while predictive ranking should also be checked for support because an optimizer will discover propositions the model likes for accidental reasons.

## Drift Is Not an Exception

Recent risk, calibration, support, source mix, and event quality should be compared against reference windows. Persistent degradation reopens the model: refit when parameters moved, reopen ontology discovery when residual structure changed, revise the actor boundary when the old scale no longer explains the trace, and update institutional access when information begins flowing through another authority structure.

The model is allowed to become wrong; it is not allowed to become wrong silently.

## What Counts as Success

Success is not that the paper sounds deep. The program succeeds operationally when an explicit actor-state model improves future prediction or transfer over strong baselines; when slow and fast components fail in the different regimes their meanings predict; when actor, relationship, and institutional construction add reusable information; and when newly discovered distinctions generalize beyond the labels that proposed them.

It succeeds more strongly when the same state supports several decisions over long horizons and under different degrees of available information. The program-level stopping rule remains deliberately unromantic: if three materially different implementations of the explicit decomposition, evaluated across at least two independent temporal corpora, fail to match a capacity-comparable monolithic sequence model and produce no repeatable cross-task transfer, retire the decomposition for that actor class; keep the useful event discipline, stop calling the latent construction a stable ontology.

That rule is not the center of the manuscript, it exists so failure has somewhere to land. The research program remains open-ended because reality keeps producing distinctions, not because every failed implementation receives an infinite appeal.

---

## In Memory of Einar Kringlen

It has been an honor tackling this multi-generational problem with you; to you I owe much.

---

[^recap-evaluation-loop]: The operational loop is [GIDS–3](00_Opening.md#gids-e3), constructed in full in [World Models and Proposition Search](06_World_Models_and_Proposition_Search.md#gids-e3-recap-world-model); this chapter assumes that definition and evaluates its consequences rather than deriving it again.
[^recap-transfer]: This is [GIDS–7 — Cross-task ontology retention](02_Gods_Infinite_Dimensional_Space.md#gids-e7), repeated because the evaluation chapter must state the acceptance rule locally; no new mathematical claim is introduced.
[^ips]: For eligible decisions \(\mathcal T_{\mathrm{ope}}\), logged assignment probability or density \(\eta_t\), mature utility \(u_t^{\mathrm{obs}}\), and target policy \(\pi\), one basic estimator is \(\widehat V_{\mathrm{IPS}}(\pi)=N_{\mathrm{ope}}^{-1}\sum_{t\in\mathcal T_{\mathrm{ope}}}[\pi_t(x_t\mid\widehat D_t,\mathcal X_t^{\mathrm{cand}})/\eta_t]u_t^{\mathrm{obs}}\). It is valid only under the stated overlap, logging, ordering, treatment, and censoring conditions.

# GIDS Study Guide / Cheat Sheet

This is the short working version of the manuscript: the hierarchy, the notation, the indispensable equations, and the tests that decide whether the theory survives. Every repeated equation keeps its canonical GIDS identifier and points back to the chapter in which it was originally explained; repetition here is retrieval, not a second derivation.

## 1. How to Read the Notation

| Mark | Meaning |
|---|---|
| \(i,j\) | actors or members |
| \(t\) | decision epoch or aligned time |
| \(r,n\) | record or event index |
| \(\tau\) | task family |
| \(\Delta\) | elapsed-time outcome horizon |
| \(H\) | number of future decision steps |
| \(\widehat{\cdot}\) | estimated from observed evidence |
| \(\widetilde{\cdot}\) | simulated by the model |
| \(\star\) | ideal or complete object |
| \(K(\cdot\mid\cdots)\) | transition kernel, a distribution over what follows |
| \(\odot\) | elementwise weighting |
| \(\oplus\) | combined structure or direct-sum idealization, as declared |

A plain arrow gives the explanatory silhouette; a conditional kernel gives the probabilistically honest version. Read every equation from left to right and ask four questions: what object enters, what information is available at that time, what transformation occurs, and whether the output is an estimate, simulation, or observed trace.

## 2. The Philosophical Descent

<a id="cheat-gids-e1"></a>
**GIDS–1 — Registration and lineage access.**[^cheat-e1]

\[
\omega_t\in\mathcal N_{\mathrm{loc}}
\xrightarrow{\operatorname{Reg}}
\mathbf g_t^{\mathrm{reg}}\in\mathcal G
\xrightarrow{P^{\mathrm{spec}}}
\widetilde{\mathbf g}_t^{\mathrm{spec}}\in\mathcal M^{\mathrm{spec}}.
\]

Something exists in the mind-independent external domain; the model registers a difference inside God's Infinite Dimensional Space; the actor class retains only what its access structure can make available. \(\mathcal G\) may be taken as an open-ended Hilbert arena for the first formalization, although the operational theory can inhabit a broader measurable state-space if later work requires it.

The same arena should be capable of representing phenomenal state at several actor scales: fish, human, ant colony, corporation, or another actor-observer. *Phenomenal* means the organized reality available to the actor as constructed at that scale; it does not, by itself, claim that every composite actor possesses one unitary human-like consciousness.

<a id="cheat-gids-e4"></a>
**GIDS–4 — Two geometries.**[^cheat-e4]

\[
k^{\mathrm{dist}}(\mathbf u,\mathbf v) =
\langle\mathbf u,\mathbf v\rangle_{\mathcal G},
\]

\[
k_{i,t,\tau}^{\mathrm{resp}}(\mathbf u,\mathbf v) =
\left\langle
B_{i,t,\tau}(\mathbf u),
B_{i,t,\tau}(\mathbf v)
\right\rangle.
\]

The first geometry asks whether registered distinctions overlap structurally; the second asks whether they bend this actor's task-relevant next state in similar ways. They need not agree, which is why one universal dot product cannot carry every meaning we need.

<a id="cheat-gids-e5"></a>
**GIDS–5 — Actor-relative object construction.**[^cheat-e5]

\[
\zeta_{i,t}^{\mathrm{obj}}(\omega_t) =
\mathcal I_{i,t}
\!\left(
A_{i,t}\operatorname{Reg}(\omega_t),
\phi_{i,t},
 c_{i,t}
\right).
\]

Register the external difference, determine what reaches the actor, then organize it through the actor's present state and context; the result is the object **for this actor now**. A proposition follows the same rule, therefore the external message and the experienced proposition are not the same model object.

## 3. From Inherited Actor to Operational State

<a id="cheat-gids-e2"></a>
**GIDS–2 — The actor descent.**[^cheat-e2]

\[
G_i
\longrightarrow
T_{i,t}
\longrightarrow
\phi_{i,t}
\rightsquigarrow
Q_{i,t}
\xrightarrow{\Pi_{\tau,\Delta}}
q_{i,t}^{(\tau,\Delta)}
\rightsquigarrow
\widehat s_{i,t}.
\]

- \(G_i\): inherited starting organization.
- \(T_{i,t}\): slowly changing realized actor after development and history.
- \(\phi_{i,t}\): complete phenomenal state now.
- \(Q_{i,t}\): general predictive response object.
- \(q_{i,t}^{(\tau,\Delta)}\): task-and-horizon summary, when one exists.
- \(\widehat s_{i,t}\): finite state estimated from traces.

These are not six names for the same thing; they mark a descent from actor as lived to actor as estimated.

The proposition-conditioned **Chimera** is the part of the person made active by this role, world, and proposition; role remains explicit context in the state, while its principal effect appears through proposition–actor interaction, so the same durable actor can therefore produce opposed responses without the model averaging the difference into noise.[^cheat-e9]

<a id="cheat-gids-e11"></a>
**GIDS–11 — Slow and fast operational state.**[^cheat-e11]

\[
\widehat s_{i,t} =
\left(
\widehat{\mathbf t}_{i,t},
\mathbf z_{i,t},
\mathbf c_{i,t},
\mathbf w_t
\right).
\]

The slow vector approximates what the actor is generally like now; the fast state carries recent chronology; context and world remain explicit. A founder does not become another founder because one email arrived, although the local state may change completely because of that email.

The fast state is aligned to decision-time through [GIDS–22](07_Evaluation_and_Ontology_Growth.md#gids-e22):

\[
N(t)=\#\{n:\operatorname{time}(\mathsf h_n)<\operatorname{time}(\mathsf d_t)\},
\qquad
\mathbf z_{i,t}=\mathbf z_{i,N(t)}.
\]

Only records available before the decision may enter the state used to predict that decision's consequences.[^cheat-e22]

## 4. Predictive State and Approximation

<a id="cheat-gids-e13"></a>
**GIDS–13 — Predictive equivalence.**[^cheat-e13]

\[
\mathsf I\sim\mathsf I'
\iff
\mathscr R^{(H)}(\cdot\mid\mathsf I,\mathbf x,\boldsymbol\xi) =
\mathscr R^{(H)}(\cdot\mid\mathsf I',\mathbf x,\boldsymbol\xi).
\]

Two information states count as the same predictive state when every admissible proposition sequence, horizon, and scenario in the declared family produces the same response law. If a biographical detail never changes any future we care about, it does not belong in the minimal state; if one forgotten event changes a single response ten steps later, it does.

<a id="cheat-gids-e14"></a>
**GIDS–14 — State error plus model error.**[^cheat-e14]

\[
\mathcal R_{\log}(P_{Y,\theta}\circ\widehat s)
-
\mathcal R_{\log}^{\star} =
\epsilon_{\tau,\Delta}^{\mathrm{state}}(\widehat s)
+
\epsilon_{\theta,\tau,\Delta}^{\mathrm{model}}(\widehat s).
\]

The first error means the state discarded predictive information; the second means the predictor failed to use retained information. Better training cannot recover what the representation threw away, while a richer representation does nothing if the predictive head cannot use it.

Memory is modeled as weighted traces whose availability changes under the present proposition; categorical evidence remains typed by source, role, and regime, because biography, stated language, observed behavior, and third-party inference are not interchangeable merely because they share a label.

## 5. Composite Actors and the World Model

A corporation is not an average of employees; the canonical [composite institutional state](05_Composite_Actors.md#gids-e16) preserves information access, authority, memory, incentives, coalitions, vetoes, missingness, and the way private organization-objects become coupled into one persistent higher-order actor.[^cheat-e16]

<a id="cheat-gids-e17"></a>
**GIDS–17 — Filtered dyadic state.**[^cheat-e17]

\[
\widehat D_{ab,t} =
\left(
\widehat s_{a,t}^{\mathrm{person}},
\widehat S_{C_a,t},
\widehat s_{b,t}^{\mathrm{person}},
\widehat S_{C_b,t},
\widehat\Gamma_{ab,t},
\mathbf w_t
\right),
\]

containing two people, two institutions, one relationship, and a shared world.

<a id="cheat-gids-e20"></a>
**GIDS–20 — Simulation and filtering.**[^cheat-e20]

\[
\widetilde D_{ab,t+1}
\sim
K_{D,\theta}
\!\left(
\cdot
\mid
\widehat D_{ab,t},
X_t=x_t,
\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right),
\]

\[
\widehat D_{ab,t+1} =
\mathcal F_\theta
\!\left(
\widehat D_{ab,t},
 x_t,
\mathcal H_{(t,t+1]}
\right).
\]

The first equation simulates what might happen; the second filters after real evidence arrives. Never write a model-generated future back as though it were evidence, and never mistake one observed trace for the whole state.

Proposition search evaluates candidate futures under the canonical [predictive proposition value](06_World_Models_and_Proposition_Search.md#gids-e21); the regime fixes continuation policy, future candidate sets, external scenario, and outcome convention, while forecasting predicts observed propositions, ranking compares simulated alternatives, and causal control requires intervention-grade evidence.[^cheat-e21]

## 6. Ontology Growth and the Four Tests

<a id="cheat-gids-e7"></a>
**GIDS–7 — Cross-task ontology retention.**[^cheat-e7]

\[
\Delta_{\mathrm{tr}}(\delta) =
\sum_{\tau\in\mathcal T_{\mathrm{held}}}
\omega_\tau
\left[
\mathcal R_\tau(M_{\mathfrak R_n})
-
\mathcal R_\tau(M_{\mathfrak R_n\oplus\delta})
\right]
-
\lambda_C C(\delta)
-
\lambda_S S(\delta).
\]

A candidate distinction may split, merge, rotate, weaken, retire, restrict, or create a coordinate family; it survives when it improves held-out tasks after complexity and instability are counted, and—where intervention data exists—participates in the predicted change. The typed registry is chosen for debuggability, while reparameterizable latent blocks preserve continuity where discrete boundaries become artificial.

The four essential tests are:

1. **Slow versus fast:** remove each and verify that failure appears on the timescales its meaning predicts.
2. **Actor, relationship, and institution:** replace the structured state with flat metadata and shallow history.
3. **Explicit construction versus monolithic sequence:** provide the same information and comparable capacity; if the monolith wins everywhere, the ontology was a story.
4. **Cross-task transfer:** discover actor structure on one outcome family and test another without rebuilding the actor from scratch.

The program-level stopping rule is simple: after three materially different implementations across at least two independent temporal corpora, if the explicit decomposition cannot match a capacity-comparable monolithic model and produces no repeatable cross-task transfer, retire the decomposition for that actor class; keep the useful event discipline, stop calling the latent construction a stable ontology.

## 7. The Shortest Accurate Summary

External reality is larger than any actor's experienced reality; actors inherit and develop methods for carving that reality into usable distinctions; a proposition is reconstructed inside the receiving actor; traces provide evidence about slow organization and fast state; memory, role, relationship, institution, and world alter the interaction; the interaction changes a recursively usable predictive state; visible behavior is one residue of that transition; candidate propositions can be compared by predicted trajectories; a new axis earns its place through cross-task transfer; the ontology is the product.

---

[^cheat-e1]: Canonical explanation: [GIDS–1](00_Opening.md#gids-e1), developed in [God's Infinite Dimensional Space](02_Gods_Infinite_Dimensional_Space.md#gids-e1-recap). Repeated here for reference; no new claim is introduced.
[^cheat-e4]: Canonical explanation: [GIDS–4](02_Gods_Infinite_Dimensional_Space.md#gids-e4).
[^cheat-e5]: Canonical explanation: [GIDS–5](02_Gods_Infinite_Dimensional_Space.md#gids-e5).
[^cheat-e2]: Canonical explanation: [GIDS–2](00_Opening.md#gids-e2), restated in [The Chimera](03_The_Chimera.md#gids-e2-recap-chimera).
[^cheat-e9]: Canonical explanation: [GIDS–9](03_The_Chimera.md#gids-e9).
[^cheat-e11]: Canonical explanation: [GIDS–11](03_The_Chimera.md#gids-e11).
[^cheat-e13]: Canonical explanation: [GIDS–13](04_Predictive_Actor_State.md#gids-e13).
[^cheat-e14]: Canonical explanation: [GIDS–14](04_Predictive_Actor_State.md#gids-e14).
[^cheat-e16]: Canonical explanation: [GIDS–16](05_Composite_Actors.md#gids-e16). The equation is not repeated here because the cheat sheet uses the dyadic composite as its one displayed actor-scale interface.
[^cheat-e17]: Canonical explanation: [GIDS–17](05_Composite_Actors.md#gids-e17).
[^cheat-e20]: Canonical explanation: [GIDS–20](06_World_Models_and_Proposition_Search.md#gids-e20).
[^cheat-e21]: Canonical explanation: [GIDS–21](06_World_Models_and_Proposition_Search.md#gids-e21).
[^cheat-e22]: Canonical explanation: [GIDS–22](07_Evaluation_and_Ontology_Growth.md#gids-e22).
[^cheat-e7]: Canonical explanation: [GIDS–7](02_Gods_Infinite_Dimensional_Space.md#gids-e7), applied in [Evaluation and Ontology Growth](07_Evaluation_and_Ontology_Growth.md#gids-e7-recap-evaluation).