# God’s Infinite Dimensional Space — Complete Audited Manuscript

## God's Infinite Dimensional Space

<div class="paper-title-block">
  <h1>God's Infinite Dimensional Space</h1>
  <p class="paper-subtitle">Transcendental Embeddings as a Way to Mathematically Express Reality, Predictive Actor-State, and the Next Phenomenal Transition of Observers</p>
</div>

<div class="paper-epigraph">
  <p>
    "Und mich ergreift ein längst entwöhntes Sehnen<br />
    Nach jenem stillen, ernsten Geisterreich,<br />
    Es schwebet nun, in unbestimmten Tönen,<br />
    …<br />
    Was ich besitze seh' ich wie im weiten,<br />
    Und was verschwand wird mir zu Wirklichkeiten."
  </p>
  <p>
    "What I possess appears as if far away;<br />
    And what has vanished becomes real to me."
  </p>
  <p class="paper-attribution">Goethe, <em>Faust I</em>, “Zueignung” (Dedication) · <em>Translation mine</em></p>
</div>

**How does reality appear to you?**

_Reality is too large to be experienced all at once, so organisms inherit a finite way of carving it up. A person then becomes a specific realized version of that inherited structure through language, history, memory, culture, and repeated events. For prediction, I do not need the whole 'soul' in some mystical sense, I just need a task-relevant approximation of the person: a slow representation of what they are generally like now, a fast representation of what is currently active in them, their present role and world-state, and a representation of the 'proposition' hitting them now. Then I model the interaction, predict the next task-relevant state (for the observer), decode visible outcomes from it, and update the system under error. The categorical part matters because a lot of what we observe about people is discrete, repeated, and role-dependent._

> Actors embedded in environments express persistent structure and transient state through the traces they leave behind. Those traces can be used to estimate the features by which the actor understands and operates in the world. Once those features are approximated, a proposition can be introduced into the model and the actor's next state and behavior can be forecast. Humans are the first research target. Sales is the first laboratory. Corporations come later as composite actors built out of people, institutional structure, facts, statistics, memory, and environment.

You can model your current mental interior, everything that you are experiencing now, as a small slice of reality that your genetic lineage allows you to experience, that can be traced by a series of state transitions up until this moment in time. The mind is an evolved, structured projection system that turns input into a lived state, and behavior is downstream of transitions in that state. Predicting what your next state will be is _not_ an impossible task: in this work I am attempting to formalize a standard algebra to make this easier and tractable.

I'll be blunt, this work is a monster, and it is, in essence, autobiographical of the mental state of the author who wrote it, representing the debauched & tortured way in which these 'discoveries' were made:

-As philosophy: this work is ambitious but undisciplined.

-As math: this work is mostly formal packaging around these undisciplined assumptions.

-As ML research propositions: this work is _potentially_ worthwhile if you squint at it.

-As a finished research article: I fear it cannot be completed with a lifetime of work.

However, the goal is to examine this. First, external conditions are registered and restricted to what a lineage can access:

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

Then inherited structure becomes one realized actor, and the inaccessible ideal is connected to a finite predictive model:

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

The measurable system estimates a state from records available before the proposition, simulates a recursively closed next state, and separately predicts delayed outcomes:

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

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

\[
(\widehat D_{ab,t},X_t=x_t)
\longmapsto
P_{Y,\theta,\tau,\Delta}
\!\left(
\cdot\mid\widehat D_{ab,t},X_t=x_t
\right).
\]

That is a roadmap, not a claim that the external world, a Hilbert space, a person, and an outcome are literally the same mathematical type.

> skip to the end if you are impatient and want a definition now

If you accept my starting assumptions, you can apply this framework to your own projects and start estimating task-relevant state transitions and outward behavior for 'agent observers' (people); the next phenomenal state remains the motivating ideal rather than a directly observed target. If you do not believe my assumptions this paper will be useless to you (but I swear to entertain, nonetheless).

Almost every serious attempt to formalize mind or behavior ends up either:

Waiting on neuroscience: "once we map the 'connectome' (or whatever the new limitation is) we'll understand behavior," a promise that has repeatedly been pushed into an indefinite future

Staying purely behavioral: black box input/output with no theory/framework of internal structure

Getting lost in phenomenology: Husserl, Heidegger, etc. philosophical but computationally intractable, thus mostly pointless.

This paper attempts a fourth path: take the structure of experience seriously as a mathematical object without needing to know its physical underpinnings. The machinery of the brain is deliberately abstracted away from the formalism. You could run the same formal program on an octopus, a human, a corporation, or a hypothetical silicon agent, but this does not mean these are all the same kind of object. A corporation is not an oversized person-vector. It is an amalgamation of people, institutional memory, incentives, facts, statistics, rules, and the environment it exists inside. Once that composite system has persistence and a characteristic way of taking propositions and producing responses, it can be treated as a higher-order actor. The outer grammar remains the same; the construction of the actor does not.

The closest intellectual ancestors are probably:

Friston's free energy principle (I legitimately didn't read this guy until well after part 2 was written, avoiding this line of thinking earlier would have been great): similar ambition of substrate-independence, but Friston goes deep into neuroscience anyway and the math becomes almost deliberately obscure; I could not extract the engineering program I wanted from the concept itself.

Marr's levels of analysis: the idea that computational and algorithmic descriptions can be analyzed without reducing them immediately to the physical implementation, while the levels remain complementary

Early Dennett: intentional stance as a legitimate predictive state without committing to substrate

But this paper is more engineering-forward than any of those. It's not asking "what is mind," which at this point is a stupid question to ask, instead we ask: "assuming mind has structure, what's the minimal formal system that lets us estimate its predictive state and forecast its outward behavior from observations alone."

This paper aims to formalize several disparate fields into a single, coherent whole. We'll begin with the tragic story for this exploration (which has to do with Kant), then go down the rabbit hole of theory together and come out the other side with a fundamental theory of 'reality' that can be applied across some fields. First, we'll discuss how the appearance of reality is constructed and how organisms parse out their version of reality. Next comes how organisms perceive state (state being the appearance of reality at that instant), and what the organism is biased to do next. Afterwards, we'll discuss how to compute memory and learning, and apply this to our understanding of state and decompose the philosophical proposition into a register that can be understood by engineers. If the program works, we will be able to progressively estimate an individual's task-relevant psychology with greater precision, test which inferred features transfer, and use those estimates to forecast behavior under propositions.

I am proposing that there is a universal way to decompose these questions through a single mathematical arena, and I will illustrate that proposal here. Let us take the measure of reality and examine God's infinite dimensional space!

Lastly, here are the differences between GIDS and standard ML:

Standard ML: construct a model → optimize for task performance → latent representations are a byproduct.
GIDS: construct a stable latent object over the observer → task performance is a probe that tells you if your latent object is good → the ontology is the product.

One dominant paradigm in modern machine learning is:

-Collect massive undifferentiated data
-Train a general model on reconstruction or next-token prediction
-Hope that task-relevant structure emerges in the latent space
-Fine-tune or probe for specific applications afterward

This is one version of the GPT/BERT/foundation-model playbook. It works extraordinarily well for language and, in adapted forms, for vision. In many such systems, the latent geometry is largely a consequence of the training objective and scale rather than an explicitly specified ontology of the actor.

GIDS inverts this completely and deliberately. The latent space is the goal, this is the product we will define, the inputs and outputs are just discovery probes. Scale is just a mechanism to get there; GIDS is a research program on how to bootstrap itself.

**Preface**

First, we're going to talk about Kant (German philosopher, hugely important), don't worry about the exact details of his works, I'm just going over the first handful of sections in his main book, Critique of Pure Reason, and using that as a jumping off point to how your reality can be represented using embeddings and states. Next, we'll use the embedding concept to examine and measure bias when an individual is interpreting reality. And finally, we'll talk about applications using this technique. Apologies for using philosophy as a segue into math; however, the pill is easier to swallow if the source of all this is adequately explained. I've kept the terms restricted to what you can find in a modern dictionary, so don't worry about converting from some esoteric nonsense to English.

I read Kant directly while going through the Western canon. Unfortunately, Kant is the worst person to represent his own ideas, so you'll have to bear with my fundamental misunderstanding of the source material. This is good news, however, as my misunderstanding of Kant is more useful than getting a 'correct' interpretation from most commentators. If you want to save yourself a year of your life, you can skip The Critique – and ignore all of the requisite readings – and try ​​Wolff's class ([Link](https://youtube.com/playlist?list=PLo0o3xtOPNLgnl2CtaxNHzie1TUWt_bp4&si=xxnQ7496XheYI5k4)). Kant uses a lot of dated terminology and systems that are only relevant to the era he wrote in (a reason why you should always start with the Greeks). I prefer the first edition, but the first and second editions differ materially, and a serious reading should consult both rather than pretending one simply replaces the other. Also, just skip Kant's stupid moral system and the categorical imperative altogether. The Critique of Pure Reason rips itself to shreds: Nietzsche was right, Kant became a coward before _his_ God.

Table of Contents:

The canonical symbol definitions are in [Canonical Notation and Mathematical Conventions](08_Canonical_Notation.md).


**[Part 0: Background](02_Part_0_Background.md):**
"Kant From An Evolutionary Perspective"
"A Fucking Table"

**[Part 1: Specifying the Area of Interest](03_Part_1_Specifying_the_Area_of_Interest.md):**
"Vectors Are All You Need"
"The Nature of Phenomenal Reality: What are we trying to measure?"
"The Evolutionary Mechanism for Encoding Transcendental Embeddings"

**[Part 2: Deriving the Transcendental Embedding](04_Part_2_Deriving_the_Transcendental_Embedding.md):**
"The Technical Scope (because otherwise I'll accidentally lie to you)"
"Behold; You! The Chimera"
"Psychology and Factor Analysis"
"Dimensionality Reduction, Attention, and Relevance"
"The Notion of State"
"Observable Predictive State"
"Memory as a Series of Vectors"
"Categorical Trace Pooling as an Operational Memory Estimator"
"Minimality, Identifiability, and Slow/Fast Factorization"
"Deriving the Transcendental Embedding"

**[Part 3: Application — Predicting How People Behave](05_Part_3_Predicting_Human_Behavior.md):**
"Towards a Universal State Transition Function"
"God's Infinite Dimensional Space: Making All Realities Composable"
"Creating the World Model"
"From Forecasting to Proposition Search"

**[Part 4: Benchmarking the World Model](06_Part_4_Benchmarking_the_World_Model.md):**
"Operational Definition of State"
"Event Time and Dataset Construction"
"The Benchmark"
"The Proposed Latent-State Model"
"Training Objective, Update Loop, and Intervention"
"Temporal Split, Evaluation, and Drift"

**[Appendix A: Study Guide / Cheat Sheet](07_Appendix_A_Study_Guide.md)**

---


---

# Part 0: Background

## Kant From An Evolutionary Perspective

> This idea, transcendental idealism, is a man walking on phantom legs: locomotion is achieved via selective understanding. Evolutionary theory is the cure for this leap from logic into one man's faith; we merely need to describe how we can go from noumena to phenomena and from purely unfiltered 'reality' external to the mind to 'Transcendental Idealism.' Fortunately, evolution is entirely capable of explaining how rocks can become delusional enough to think they are alive. The core of this (my) concept is that you are not evolving a proximal understanding of reality that is only mildly different from 'noumenal reality.' No: your perceptions of reality were progressively built up, specialized, and abstracted from 'older' realities. I use 'reality' in the solipsist sense, i.e., reality exists for an individual organism in a very particular way, and organisms have no access to what the 'actual' universe looks like. Let's get into it.

After reading Critique of Pure Reason, I was struck with the urge to align the Kantian worldview with evolutionary theory; he was so close to having a useful account of reality as it appears to you that, by a little modification and a little extension, we could build a system of reality construction. The basic thought I had was that the Transcendental Aesthetic—space and time as forms of intuition—and the Transcendental Logic—the concepts and rules through which appearances become objects of thought—needed to be placed inside an account of how minds originate and vary. That was not Kant's project. His project was to explain the conditions under which experience and knowledge are possible, not the evolutionary origin of those conditions. Posterity did not simply prove this entire apparatus correct; later work made parts of the constructive insight look fruitful while leaving the full transcendental system contested. But we can do better for the engineering problem.

To quickly get you up to speed on Kantian theory, he was against the empiricist picture in which all cognitive structure is passively received from sense experience, while also rejecting dogmatic rationalist claims to knowledge beyond the bounds of possible experience. Kant's central idea is that experience already depends on forms and concepts supplied by the cognizing subject. Space and time are forms of sensible intuition; the categories are concepts of the understanding used to organize appearances into objects of possible experience. That does not mean the anatomy of a sensory organ is itself a piece of a priori knowledge, or that the subject learns a rule before ever encountering anything. It means that some structure is presupposed by experience rather than copied out of experience. And that idea in particular stuck with me like a tick.

> "I think, therefore I am." Here is the phrase that has injured philosophy the most: a heady mixture of solipsism and egoism is all it took to shackle the minds of the great systematizers and force their expenditure on triviality. To make great systems and frameworks of mind means little, regardless of how correct one can be, unless there is practical application. We should measure our system's correctness via its potential for practical application. "I think, therefore I am" is an island of mathematical tautology only fit (and attractive) for the autistic mind; given the language of communication, definitions, the cultural setting, and the genetic predisposition of the individuals committing to dialectic, we can examine an infinite variety of logical games which can all be true. Kant inherits the post-Cartesian problem of how a subject can know anything at all, but he also attacks the rational-psychology leap from the bare unity of the “I think” to a substantial, immortal soul. More unfortunately, the whole project of "philosophy" is tainted by a series of logical games generated out of misunderstanding, and the subject should usually not be engaged with seriously.

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

My modification of the thesis is: "Our representation of reality presupposes the interpretive structures required to generate 'reality' for the organism. The origin of mind comes from a manifold combination of interpretive structures and efficient responses as a single unit encoded in a particular way via evolutionary pressures. E.g., some encoding process (evolution) hammered out a lower-dimensional version of reality which—when understood and decompressed—can inform you of the version of reality one is seeing; this can be represented mathematically."

For humans, we find the basis of our mind in the mental representation we have of reality, the ordered transmission of expedient data to the mind, the symbolic transmission between minds (conversation/writing), and the collective delusion we experience and infer to one another through symbolic transmission. Consciousness (this reality you are experiencing) is an individual projection, a group effort and affect, and is easier to think about as a series of ordered delusions imposed onto our mental interior than something entirely rational. We 'evolved' our 'Synthetic A Priori Judgments,' those mental structures that add contextual understanding to the subjects we are looking at, to react to increasingly complex streams of data gathered through sensory organs and processed into increasingly complex phenomena in our minds.

From simple rules, we can examine the development of complex systems. Imagine a group of organisms that can only see a 48x48 grid of pixels, they all 'see' a shape traveling across the pixel surface and the organism has three programmed responses: Attack, Mate, Run. Let's say one response in this scenario is correct: Attack. All organisms that didn't have a bias toward that action didn't pass on their genes. Natural selection refines the infinite data stream into biasing structures. Nothing new here. Let's examine whether we make this a little bit more complicated: how about if we had a vector for the size of the observer and a vector for the size of the shape? We can imagine an interaction in which the bias for attacking or running is size-dependent on relative size differences. However, size is a spatial phenomenon, so we need a vector for distance and position; and how do you compute the distance and position vectors? You'd need time and velocity vectors such that, over time, shapes' relative increase or decrease in size gives the position and distance to the observer. You'd also need causality to infer that shapes seen in this state are the same shapes seen in the previous state. Thus, we can see a sort of synergistic interrelatedness between space, time, causality, and other things that probably share some primitive in the mind's interpretive structures.

Mathematics is a great example of these interpretive structures operating in concert and cohesively developing a system of rules that seems to be universally correct, even though mathematics isn't part of the physical world. To count to 3, you need the a priori knowledge that: 1\. separation exists, 2\. that separations in a 'chunk' of data constitute an 'object,' 3\. that there can be more than a singular object. From this premise (and a bit of a logical leap), you can count to 3: object 1 is in my mental interior, object 2 is in my mental interior, object 3 is in my mental interior and the sum, an abstract 'de-separation,' is a variable representing all of the objects. This is Kant's 'synthesis', a single act of knowledge generated from observing what is 'manifold' in them. From this basis, you can abstract the object from the physical data that is being represented in your mental interior and treat it independent of physical reality, yet another interpretive structure. This function of adding together objects where the sum is represented by a value contains 'predicates,' which provide extra context to abstract objects. We evolved a huge array of predicates to deal with data that needed synthetic manipulation for efficient responses.

So, we see the basis of mathematics, a synthetic a priori judgment that cannot be understood by 'pure reason' alone. 'Math' is just the most efficient way to interact with your interpretation of abstractions. In fact, your notion of what 'reason' is comes from a biasing mechanism that assists you in understanding & parsing a very complex set of environmental data, all in an effort to assist you in reproduction.

A posteriori knowledge, knowledge through experience, is subsumed by the a priori, because our interpretive structures form the basis by which we can learn things. We are designed to learn in (generally, not pedagogically) specific patterns; otherwise, our development is stunted. Which is why we cannot separate understanding from interpretation; they are functionally the same thing. The practical application for pedagogy –to do is to learn–however, this idea is hugely extensible and can explain very complicated behaviors. Your politics is the interpretation of your reality with your biases, your understanding of violence is from your inbuilt interpretive structures, and your understanding of yourself is biased through your interpretive structures. Hence, a large part of psychology uses “constructs” to compress recurring patterns of mind and behavior into factors that can be measured. Psychiatry uses some of the same language, but it is not reducible to that one method.

However, all of the information I've just supplied to you is very Kantian in origin, and it doesn't explain the core idea we're tackling. What we've just talked about is pure conjecture on how the mind may work in an abstract way, and how reality is plausibly experienced, this has almost no practical application, it is merely a device to softly introduce the following: Exploring how the mind works from the point of view of a mind is extremely difficult and is not a first principles approach, by figuring out the primitives and then attempting to understand origin of mind we can get a more holistic understanding of what is happening to people.

From here, I'm going to make a critique of Kant, then we're going to talk about the nature of phenomenal representations (and how to represent them mathematically), then we're going to get into how those representations cascade into phenomenal structures, and then finally we're going to integrate it all together so we can see how minds work. Simple.

However, I'm going to preface the next section, which is full of vitriol, by stating that I believe Kant's childish moral philosophy has done more harm to the human race than any philosophy before it and should be exiled from our dialectic; we will not be touching on this as it is a complete distraction.

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

## A Fucking Table

> Kant explaining categories, Circa 1781 (colorized)

Kant's Table of Categories is key to the Critique of Pure Reason and lists the pure concepts he says the understanding uses to synthesize appearances into objects of possible experience. The categories are arranged under four headings: quantity (unity, plurality, totality), quality (reality, negation, limitation), relation (inherence and subsistence, causality and dependence, community or reciprocity), and modality (possibility–impossibility, existence–nonexistence, necessity–contingency). Kant does not present this as an empirical taxonomy of human traits. He argues that the table is derived from the logical forms of judgment and is therefore complete. While the derivation sounds architecturally neat, I still think the bridge from forms of judgment to a complete inventory of cognition is made to do far more work than it earns.

The worst part about reading German philosophers is the tour-de-force required for them to explain simple ideas (you'll find the irony in that statement soon, lmao); Kant explained his ideas so succinctly that his manifesto sprawled across hundreds upon hundreds of pages, with far fewer concrete examples than a modern reader would want and an almost continuous stream of architecture. The second-worst part of reading German philosophers is that they assume everyone reading it will be German. I know where Kant says the table comes from: the table of the logical forms of judgment. I still do not think that derivation is sufficient to establish a complete and universal inventory of the concepts required for every possible cognition, and I have no idea why I should accept the rest of the system merely because the table closes neatly. Kant couldn't contemplate the chaos of noumena becoming phenomena without his own mental instruments, but he cannot expect me to use the same flawed device as a way of coping with chaos.

Here are some quotes:

> "Without sensibility no object would be given to us, without understanding no object would be thought. Thoughts without content are empty, intuitions without concepts are blind. **Hence it is as necessary to make our concepts sensible, i.e., to add the object to them in intuition**, as to make our intuitions understandable, i.e., to bring them under concepts." — *Critique of Pure Reason*, A51/B75, translation wording adapted

> Kant argues that the categories "**prescribe laws** a priori to appearances," and therefore to nature understood as the sum of appearances. — *Critique of Pure Reason*, B163–B165, wording abridged and adapted

This is what happens when someone has a linguistic IQ of 150 and needs to explain away something so the rest of his theory works. Moreover, the Critique does not make cross-cultural or individual variation central to its account of the conditions of experience. Nietzsche's line that a philosophy is a kind of personal confession feels right again.

This is where I differ greatly from Kant, his assumptions leave too many gaps in his core philosophy and even if you were to understand The Critique Of Pure Reason in its totality, there would be _little_ practical application. I don't believe the categories are a necessary predicate of experience. With a slight modification to Kant's core, we can draw a vast ocean of wealth out of his philosophy, instead of a stupid table and everything that comes after it. Some aspects of the table _may_ be a linguistic description of what I describe later; however, as it stands now, it lacks composability, nuance, degrees of intensity, and it is not a universal key to understand any form of cognition.

---


---

# Part 1: Specifying the Area of Interest

## Vectors Are All You Need

> Earlier in my life, I wrote a chapter called 'chaos' in a book I was writing, it was essentially a long-winded deconstruction of light and matter into component datums that interact with each other, so light \+ a particular protein/molecule \+ interpretation from a brain \= the experiential feeling of, for example, seeing a color. I stretched that idea past its breaking point when I went on to describe a 'being' that can 'see' all light without interpretive structures as an illustrative tool. I imagined it would be like a chaotic static without form. The point of describing that background is because in the next chapter, I brought forward a nascent version of this Kantian theory I'm presenting now. We need to bias our interpretation of data to make heuristic sense of what's going on around us and produce optimal responses. Afterwards, I diverge from the Kantian line by proposing the evolutionary model of transcendental idealism and a far more nuanced tool to understand reality.

I want to build a system that has universal composability between something that is true at the biochemical level to the realm of mental interiority and then further to the realm of macroscopic behavioral analysis and still be flexible enough to be integrated into practical applications. As I've stated before, inside this framework, perception, interpretation, understanding, and action are treated as phases of one coupled state-transition process rather than as unrelated substances. This normalization greatly simplifies the mathematics because the implementation machinery can be abstracted behind a transition law: at the chosen level of analysis, a thing is characterized by the differences it makes. You can map states of phenomenal transition from one form to another and call that change the product of a 'transcendental function.' The molecules in a rod cell interacting with light, the resulting neural transformations, the experienced image, and the action recruited by that image can therefore be described at different resolutions inside one dynamical grammar without pretending that they are literally the same physical object.

Let me declare my priors before continuing; we will return to each of them in more precise form later.

> One warning before the list: a vector in this paper is a representation, not the thing itself. A photon, a table, a sentence, a person, and a corporation are not secretly little arrows hiding inside a Hilbert space. A category can be represented by a direction, a region, a family of activations, or a learned function over the space; an object of experience is a structured activation produced when an observer encounters something. GIDS is the arena in which those distinctions can be written and related. It is not an inventory in which every object has been assigned its own primitive shelf.

1. For the purposes of this framework, every observable change in matter can be given a vector representation at the level of analysis we care about.
   1. These vectors are meaningless to an organism until an evolutionary process encodes significance onto the observer.
   2. I will call the pre-interpreted side of this picture **noumenal distinctions**: the raw differences available prior to the organism's full phenomenal organization of them. They only become vectors after the framework registers them mathematically.
      1. These are not "seen" directly by consciousness; they are the bits of raw, unfiltered reality that first enter the system through physical interaction.
2. Reality arrives as a continuous stream, but for modeling purposes we can sample that stream into discrete state-snapshots. At any given snapshot, a local noumenal state contains the vector-representable material changes relevant at that moment and position.
3. An observer's qualia—phenomenal reality—can likewise be modeled as a state that updates over time. That state is the representation of feelings, visuals, memories, action-tendencies, and every other cognitive process available to the observer at that moment.
4. In this model, evolution supplies a **seed** for the organism's Transcendental Embedding. This is the inherited developmental template that constrains the repertoire of distinctions an organism can, in principle, acquire or express. The seed does not guarantee one identical experienced world for every member of a lineage; it supplies a lineage-compatible starting organization through which external conditions can be registered and recursively transformed into phenomenal states.
   1. There exists an ambient latent space—conceptually open-ended, and idealized here as infinite-dimensional—into which experienceable phenomena can be projected.
   2. A transformation function maps noumenal inputs together with the current phenomenal state into new phenomenal representations.
   3. This inherited seed represents the organism's interpretive lens: the lineage-fixed framework through which experience must first pass.
   4. At the resolution of this abstraction, the inherited seed is treated as fixed over an individual's life while development, gene regulation, plasticity, memory, and history determine how that starting organization is realized. This is a modeling separation, not a claim that biological development is genetically static.
5. The Transcendental Embedding transforms registered noumenal distinctions into phenomenal representations through two complementary processes:
   1. **Projection:** mapping raw inputs into meaningful coordinates within the organism's latent space.
   2. **Transformation:** combining the current phenomenal state with new inputs under the inherited rule-set.
   3. These phenomenal vectors are representations of reality as it appears to the organism: this is the framework in which perception, interpretation, and action are treated as one continuous process.
6. Paired with this inherited seed is a transition rule that maps one phenomenal state to the next. In the idealized version of the framework, that rule is treated as fixed with respect to the seed itself, while the organism's actual state supplies the changing input.
7. Taken together, the inherited seed and the transition rule generate phenomenal vectors. Those phenomenal vectors are reality as it appears to the organism.
   1. Noumenal conditions are invisible to consciousness; the model only represents the distinctions that physical interaction first registers.
   2. Everything available to conscious experience is already on the phenomenal side of the transformation.
   3. In principle, one can trace the transformation of a registered physical distinction into a phenomenal representation through the biochemical and computational chain that produces the experience.
8. The recursive mapping from one phenomenal state to the next is, for the organism, its lived reality.
   1. In the idealized explanatory version of this framework, I write the update deterministically: given the organism's inherited seed, the complete current state, and the complete incoming conditions, the next state follows. In the operational version, the universe and our knowledge of it may be stochastic, so the arrow becomes a conditional distribution over possible next states.

> WARNING: I'm using a simpler, incomplete and somewhat contradictory version of the term "Transcendental Embedding" so that it will be easier to grok now, but will be explained more fully later.

From your eyes to inside your mind, you are currently running extremely complex systems to organize these letters into discernible symbols and then translate that ordering of symbols into information you can grok. But let's pretend, for a moment, that you were much dumber than you are now—so dumb, in fact, that you are not even conscious. Things simply happen to you. You barely have a sense of time. Your vision is more like a flash of symbols to which you can only have strong affective reactions. Here is a story of your life:

`You're walking along and suddenly, out of nowhere:`

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

`Of course, in a moment of panic, you feel:`

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

`And instinctually you do:`

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

`Whew, thank God that's over; now you want to do:`

`D [0.27, 0.72, -0.09, 0.65, 0.41, -0.53] with the:`
`    A1 [0.54, -0.13, 0.75, 0.42, -0.26, 0.10]`

`    ...Kinda gross, but whatever.`

Notice the small change in the last coordinate between the first and last object-vector, `.87 → .10`. We can infer that most of the structure remains intact while one aspect of the represented object has changed. In this toy example, the system's bias structure turns

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

into

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

and everything in between is the vector representation of the internal phenomenal activity required to bring about that change. On this view, the feeling and the instinctive action can be written in the same general format.

For readability, I decomposed the previous series into separate time-steps. That is not how the process actually unfolds. In reality, these states overlap and bleed into one another. The point of the decomposition is only to show how one structured representation can recruit another by shared positions. If `A0` activates the system and `B` appears immediately afterward, you should imagine the coordinates of `A0` and `B` occupying the same larger state-space, with some regions active and others blank. In that more realistic presentation, the same story looks like this. Let `S(n)` denote the state at discrete modeling step `n`:

`S1[...0.54,-0.13,0.75,0.42,-0.26,0.87,0,0,0,0,0,0,0,0,0,0,0,0,0 ... 0]`
`            A0 alone`
`S2[...{0.54,-0.13,0.75,0.42,-0.26,0.87},`
`{0.32,0.69,-0.15,0.78,0.25,-0.44},`
`0,0,0,0,0,... 0]`
`A0 + B`
`S3[...{0.54,-0.13,0.75,0.42,-0.26,0.87},`
`{0.32,0.69,-0.15,0.78,0.25,-0.44},`
`{0.61,-0.33,0.48,0.91,-0.18,0.36},`
`... 0]`
`A0 + B + C`
`S4[...{0.54,-0.13,0.75,0.42,-0.26,0.10},`
`0,0,0,0,0,`
`0,0,0,0,0,`
`... 0]`
`A1`
`S5[...{0.54,-0.13,0.75,0.42,-0.26,0.10},`
`{0.27,0.72,-0.09,0.65,0.41,-0.53},`
`0,0,0,0,0,`
`... 0]`
`A1 + D`

I added the `{}` only to make the story more legible and to separate features of the represented state; they are not part of the formal system itself. Notice also that the `B` and `D` representations share positions. That suggests a region of the state-space associated with an affective or motivational response to the coordinates occupied by `A`.

I am using a more tangible action-based example here, but the same logic would apply to something as simple as light hitting a photoreceptor and the observer mapping color and position into a phenomenal representation. Depending on the application, you can average the state over time, or produce a higher-order vector representation of the whole sequence. That may be lossy, but that is acceptable: evolution does not need a perfect copy of reality. It needs an approximation of reality good enough to bring about adaptive state-transitions.

There are many ways to describe the world. People often say that functions describe the world, and that is true as far as it goes. But most approximations—including functions in isolation—describe only the world of appearances available to the model. The advantage of the vector-space picture is that it lets us describe many different levels of organization inside one composable framework. Neural networks are still useful here, but they are downstream processors of structured representations; they are not, by themselves, a theory of how those representations are made available to the organism.

So, in this section, I assume an open-ended nonlinear mapping rule capable of weighting and summing phenomenal vectors as they co-occur. That rule takes the current phenomenal state, maps it through the organism's inherited interpretive structure, and yields a new phenomenal state. Later, when we get to applications, the practical problem will be path-isolation and signal-processing: given the noise of a large state-space, which contributing factors produce the strongest signal for the transition we care about?

The broader claim is not that the organism first builds a neural network and only then acquires a world. It is the reverse. The organism inherits a structured way of carving reality into usable distinctions, and the processor it builds later operates within that inherited space. Evolution gives the observer a repertoire of vector-like distinctions—time, space, color, objectness, shape, bodily boundary, hierarchy, proportion, attention, lower-order affect, higher-order affect, symbolic meaning, interiority, and so on. Action is a byproduct of the continuous processing of these distinctions through the organism's interpretive structure.

It is important to understand the primitives before the abstractions. We want a representation of phenomenal life that is mathematically tractable without pretending that every observer receives the same world in the same way. The same broad stream of reality may confront multiple observers, but different inherited structures and different realized histories will transform that stream into different phenomenal outcomes.

If you want an intuitive picture, think of the inherited structure as a massive coordinate-ready template and of lived history as the process that determines how that template is realized in the individual. In principle, that structure constrains how you can react to the world; in practice, any model we build will only ever approximate that structure. Evolution is the encoding protocol, compression is part of the mechanism, and the point of the framework is to describe how reality is described for an organism—not to confuse the description with the thing itself.

Why use an embedding system rather than just talk about neural networks? Because I am not trying only to describe a processor. I am trying to describe the representational conditions that make processing possible in the first place. In that sense, what this section does is combine the space/time side of experience with the rest of phenomenal life into one common representational framework, rather than treating them as separate faculties that must later be stitched back together.

Note also that we are never really processing one isolated datum at a time. The phenomenal stream is continuous, even when the model samples it discretely. The inherited structure is relatively stable; your experiences, memories, and moods are not modifications to the seed itself so much as modifications to the stream it is processing and to the realized organization built on top of that seed. That is why this device can represent what Kant's table of categories cannot: not just a generic human mind, but, in principle, the different ways minds can be structured across organisms and across individuals.

So ends **Kant from the Evolutionary Perspective** and **Vectors Are All You Need**.

Next, we get into the real meat: what an application of this theory looks like, and how to do these calculations.

## The Nature of Phenomenal Reality: What are we trying to measure?

Most of the time, when mathematicians are using the 'infinite' it is to simplify a problem and also say something concrete about finite things. Thus, we are continuing the tradition by emphasizing the infinite ways in which reality can be understood so that we can isolate a finite representation relevant to us. To simplify further, I will often write the system as though it were deterministic: if the complete noumenal conditions, the complete phenomenal state, and the exact transition rule were known, then the next state would follow cleanly. This is an explanatory device, not a declaration that the actual universe must be deterministic. For real work, incomplete state and genuinely stochastic dynamics require probability distributions. The deterministic arrow is the silhouette; the stochastic kernel is the model we can actually train.

In the previous section, we went over the concept of Transcendental Embeddings and we assumed the finite vector space that the organism played on. In 'God's Infinite Dimensional Space' there are infinitely many available directions of representation, enough to leave room for distinctions that one actor can register and another cannot. Quantum mechanics offers a useful mathematical analogy, but only an analogy: pure quantum states are represented by rays in a **complex** Hilbert space and observables, in the standard formulation, by self-adjoint operators, whereas GIDS uses an idealized **real** Hilbert space as a representation arena for distinctions relevant to actors. Nothing here is a claim that psychology is quantum mechanical. The common point is that Hilbert-space structure lets us represent a state with as many coordinates as the model requires while still taking finite projections for actual calculation.

So what we are trying to find first is a general transition grammar that maps the current actor–world state and incoming conditions onto a distribution over what follows. Different organisms will instantiate different transition laws; the universal claim concerns the form of the description, not one literal function shared by every actor. Keep in mind that the inherited organization carries the residue of evolutionary history through bodies, sensors, developmental programs, and learning capacities—not as a complete record written directly into a few genes. So how do we reduce the complexity to something tangible? We represent those inherited developmental constraints as limiting and organizing the repertoire of distinctions the organism can acquire and use. The guiding question is how to operationalize that constraint mathematically so that we can construct, or at least approximate, the transition law for one actor class.

In the formal model, inherited developmental constraints are represented by a stable but extraordinarily high-dimensional accessible structure: the seed of the Transcendental Embedding. Within this idealized structure, an organism's capacities—whether sensory, interoceptive, motor, or eventually conceptual—are described in Hilbert-space coordinates and nonlinear transformations. Genes do not directly label coordinates such as fear, color, or hierarchy. Genes participate in developmental systems that build bodies, sensors, neural and biochemical machinery, and learning rules; the accessible structure is our mathematical description of the capacities that emerge.

This is not a one-gene/one-coordinate claim. Genes alter developmental machinery, bodies, sensors, and learning systems; the axes are an idealized description of the capacities and transformations that emerge from that machinery.

The trick is, no organism has conscious access to these developmental templates; instead, they appear as “normal” to the experiencing subject. Imagine a hypothetical lineage in which a particular chromatic signal reliably predicted danger and the developmental system made that signal strongly recruit threat response. For that organism, the surge would arrive as an immediate fact of experience rather than as a consciously selected inference. In actual humans, responses to red are not one genetically fixed universal reflex; biology, learning, culture, and context all contribute. The point of the example is only that inherited organization can make some transformations feel self-evident or inescapable to the subject.

Also, keep in mind that the ideal explanation is still deterministic because it is easier to see the machinery when one state points cleanly into the next. The operational framework is not allowed that luxury. It will use conditional distributions, calibration, and experiments wherever the real world refuses to collapse into one answer. The question is not whether probability can be banished; it is whether the underlying state can be made rich enough that the remaining uncertainty becomes tractable.

### The Evolutionary Mechanism for Encoding Transcendental Embeddings

It is useful to imagine humans as experiencing a small slice of all possible ways to experience reality. Within this small slice, there is enormous variation; however, we are still subject to an evolutionary inheritance that provides us with a predictable range of ways to perceive and structure reality. You could imagine the way humans experience reality is a series of dimensions. To be simple, we could have dimensions for the range of colors we see, as well as dimensions for how we position those colors in our mental interior when a photon hits our eyeball. We could build on this by adding other dimensions and associating groups of pixels. Then, we could create structures from those groupings and assign dimensions that associate meaning (or potential for meaning) with those superstructures. Finally—you could imagine—we have dimensions that represent placeholders for the superstructures humans expect in their lives. An example of this would be something like a mother figure, or what a place to sit looks like, or what violence is, or even something complex like a god object. These are complex phenomena that depend on humans holding a particular psychic position in reality; the god object, for example, could be represented as a structured pattern assembled from communal bonding, fatherhood, war, purity, spite, revenge, love, sacrifice, death, externalized meaning, care, fear, language, outsiders, and the couplings between them. A crude first implementation might pool or average some of those representations, but the category is not itself one primitive axis. Different weightings of the pattern, coupled with a supporting or dismissive society, can produce a range of outcomes when a human internalizes and practices the god object.

All of the sub-vectors of the god object are their own series of complex dimensions as well. The war object might require communal bonds, hunger, fear, negative ethnocentrism, concepts of ownership and land, hierarchy, etc. Let's get even simpler, the hunger object could be composed of the vectors of glucose saturation, fullness, thirst, fat concentration, presence of ghrelin, and other biochemical factors. The sleight of hand I performed was relating biochemical signals and things that exist purely as a concept in the mind of the user, like god and war. Ah, but that is the point! At the level of this formalism, hunger and the feeling that your group is at war both enter the phenomenal state and can recruit action. This does not mean they feel identical; it means the same representational grammar can carry both bodily and synthetic experience. Humans (to varying individual degrees of intensity and presence) have a space in their minds for both concepts.

All these complex traits, being the results of series of simpler objects, make their examination and deracination possible as well as, in principle, computable at an appropriate resolution. But how did humans get to become so complex? Why is the range of our experiences so large relative to other creatures? How did we obtain this Transcendental Embedding? To be overly simplistic, we can trace evolutionary lineages backward and imagine additions, losses, and reorganizations of accessible distinctions over time. This also gives us a way to ask which organisms have functional overlap. Some corvids, for example, have solved water-displacement tasks in laboratory settings. That is enough to motivate convergent functional structure; it is not evidence that corvids possess a human-like theory of buoyancy or experience the task in the same way. The framework needs room for both overlap and radically different internal construction.

**To be very clear:** some enormous number of interacting coordinates and transformations produces the expected reality available to a human. I picture this as radically larger for a human than for an earthworm, but the numbers here are an intuition pump, not an empirical estimate.

But let's start at the beginning and build a toy organism, not a literal history of the origin of life. Imagine a membrane-bounded protocell or minimal cell-like system with a rupture-sensitive response mechanism: changes in ionic balance alter whether the boundary remains intact. At the level of the toy representation, that gives us a first usable distinction—membrane integrity—whose states matter for persistence. We can call this a “membrane-integrity axis,” while remembering that the axis is our model of a coupled biochemical process, not a claim that the protocell possesses a conscious binary qualia.

Random copying errors then throw up tiny tweaks: a peptide that bends when it binds a proton, a chromophore that flips shape when hit by a photon, an ion channel that opens more readily in warmer fluid. In the toy geometry, each tweak proposes a new accessible distinction or transformation. Many changes are neutral, harmful, or useful only in combination with other changes; they do not pass through one clean immediate-payoff filter. Occasionally, however, a heritable change helps descendants move toward nutrients or away from damaging conditions and increases in frequency. The represented repertoire can then become richer, while drift, pleiotropy, developmental coupling, and later loss remain outside this simplified picture.

In this toy story, retained capacities can be reorganized into higher-order compound distinctions rather than simply discarded. Once a light-sensitive molecule exists, downstream mutations wire two such molecules together, letting the organism register differences in intensity rather than mere presence. That difference becomes a new vector—contrast. A later duplication introduces a second pigment shifted in wavelength; the comparator circuit now yields a chromatic axis. What began as a single light/no-light bit has unfolded into a color cube where the vectors of understanding reality become synergistically linked.

This simple logical mechanism is highly scalable as a story. Chemical gradients can become richer sensorimotor maps; distributed pressure signals can contribute to a body schema; socially relevant signals can participate in an emerging social manifold. In this simplified picture, selection tends to retain and organize capacities whose contributions to reproduction outweigh their costs in the environments where they matter. Real evolutionary change also includes neutral drift, constraint, exaptation, frequency dependence, and traits whose value appears only through interactions. Complexity is the residue of all of that history, not the result of an accountant adding one profitable axis at a time.

By the time we reach hominids, the embedding has accrued countless axes and transformations. Some are exteroceptive (hue, pitch, depth), others interoceptive (blood CO₂ and signals involved in hunger), and others still are synthetic patterns available only through learned conceptual organization: tool, ally, lover, or taboo. The aggregate is a pseudo-species-level template: an expected repertoire of distinctions a typical human can, in principle, develop. The proper resolution, however, is the individual organism. An individual inherits a developmental program for a species-compatible template; the realized embedding is not fully formed at birth. If an individual lacks a functional long-wavelength-sensitive cone pigment, or has an altered version of it, some chromatic discriminations are unavailable or changed even though the species-level model retains a place for long-wavelength contrast. Reality changes for that observer; it does not simply lose one mathematically isolated “shade.” Our concept of self-preservation appears complex, but it is built on vastly older mechanisms of persistence. Finally, vector norm has no intrinsic psychological meaning. A representation may use magnitude to encode intensity or importance only if the encoding rule or training objective gives the norm that interpretation; self-preservation is not automatically “far from the origin.”

Before reducing this into four steps, I need to mark the level of the claim. This is a toy construction of how new distinctions could become available to a lineage. Real evolution also contains drift, pleiotropy, correlated traits, exaptation, changing environments, loss, and reorganization of old capacities. I am keeping the simple axis story because it gives us a usable formal picture, not because biology literally appends one clean orthogonal coordinate at a time.

To be brief:

1. A mutation or developmental change proposes a new accessible distinction for the organism.
   1. In the toy geometry, this appears as a candidate coordinate or transformation of experience.
2. If the change contributes to reproductive success, it can spread through the lineage.
3. The capacity may stabilize, be modified, or later be lost.
4. Richer composite coordinates arise through interaction, recombination, and reorganization of older distinctions.

## Formalization

> _(Bold lowercase symbols denote finite coordinate vectors. Calligraphic symbols denote spaces, sets, or families of maps. Ordinary uppercase symbols may denote structured states, kernels, linear maps, or matrices; their type is declared where they appear.)_

Before we continue, the philosophical argument I gave was to ensure that we can simplify the essential elements down to the point where we can inject them into standard and proven mathematical frameworks. I have to reiterate: I am not inventing new math. I am trying to decide what kind of objects the old math should be attached to without quietly turning the metaphor into a fact.

Lastly, this work has been an absolute tour-de-force of effort; as a consequence, we should assume there can be logical gaps or errors in the formalization. Email me if you see anything wrong.

### 1\. The noumenal domain and God's Infinite Dimensional Space

Let \(\mathcal N\) denote the **noumenal domain**: whatever complete physical or otherwise external conditions exist prior to the observer's organization of them.

I do **not** need to assume that \(\mathcal N\) is itself a Hilbert space. Doing that would already smuggle the observer's mathematics into the thing I am claiming the observer cannot encounter directly. \(\mathcal N\) is therefore left deliberately under-specified.

Let

\[
\mathcal G
\]

denote **God's Infinite Dimensional Space**: an idealized real Hilbert representation arena rich enough to encode distinctions and combinations of distinctions that could, in principle, participate in the experience or response of possible actors. The construction does not require one orthogonal basis coordinate for every named distinction.

For the working formalism, take \(\mathcal G\) to be separable. Separability is a tractability convention, not a metaphysical discovery; it says the space admits a countable orthonormal basis. If the ideal arena required a nonseparable space, the outer construction could be widened without changing the finite models we can actually build.

For concreteness, one can think of

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

the space of square-summable real sequences, with an orthonormal basis

\[
(\mathbf e_k)_{k=1}^{\infty}.
\]

A modeled activation in GIDS can then be written

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

This is a capacity claim, not an ontology of objects. A person is not one basis vector. A proposition is not another basis vector. A corporation is not a special organizational subspace waiting next to the human subspace. The space contains possible distinctions and combinations of distinctions. Objects, categories, actors, propositions, and institutions are represented by structured patterns, regions, functions, and transformations built with those coordinates.

The distinction matters. "Fear" may eventually correspond to a family of related activations. "Corporation" may correspond to a category used by a human observer. Neither claim means that fear or a corporation is literally one primitive axis in the fabric of \(\mathcal G\).

To move from external conditions into a model-side representation, introduce a **registration map**

\[
\operatorname{Reg}:\mathcal N_{\mathrm{loc}}\to\mathcal G.
\]

For a local external condition \(\omega_t\in\mathcal N_{\mathrm{loc}}\),

\[
\mathbf g_t^{\mathrm{reg}}=\operatorname{Reg}(\omega_t)\in\mathcal G
\]

is the model's representation of distinctions available in that condition. It is not the noumenon itself. It is already a mathematical encoding chosen for the framework.

### 2\. Species-level accessible structure

Evolution does not give a lineage access to every coordinate in \(\mathcal G\). It preserves and organizes a finite or effectively finite repertoire of distinctions that the lineage can register and use.

Define the species-level accessible structure by a closed subspace

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

where the vectors \(\mathbf v_1,\ldots,\mathbf v_d\) are an idealized coordinate basis for the lineage's available distinctions. For convenience, assume they are orthonormal:

\[
\langle \mathbf v_i,\mathbf v_j\rangle=\delta_{ij}.
\]

The associated projection is

\[
P^{\mathrm{spec}}:\mathcal G\to\mathcal M^{\mathrm{spec}},
\qquad
P^{\mathrm{spec}}\mathbf y
=
\sum_{j=1}^{d}
\langle \mathbf v_j,\mathbf y\rangle\mathbf v_j,
\qquad \mathbf y\in\mathcal G.
\]

Applied to a registered local condition,

\[
\widetilde{\mathbf g}_t^{\mathrm{spec}}
=
P^{\mathrm{spec}}\mathbf g_t^{\mathrm{reg}},
\]

this yields the part of the model-side distinction field that the lineage can, in principle, use.

The orthogonal projection is a clean mathematical idealization. It says: under the norm we selected, retain the component that lies in the accessible structure and annihilate the rest. Biology need not literally perform an orthogonal projection, and psychological independence need not literally equal geometric orthogonality. Later implementations can replace this with a nonlinear access map. I am using the linear form because it makes the first version legible.

The same accessible slice can be expressed in coordinates through

\[
E^{\mathrm{spec}}:\mathcal G\to\mathbb R^d,
\qquad
E^{\mathrm{spec}}(\mathbf g_t^{\mathrm{reg}})
=
\begin{bmatrix}
\langle \mathbf v_1,\mathbf g_t^{\mathrm{reg}}\rangle\\
\vdots\\
\langle \mathbf v_d,\mathbf g_t^{\mathrm{reg}}\rangle
\end{bmatrix}.
\]

These coordinates are not yet phenomenal reality. They are the lineage-available material from which phenomenal organization can be built.

This is the sense in which GIDS makes room for experiences that humans cannot have. A coordinate can exist in \(\mathcal G\) while being annihilated by \(P^{\mathrm{spec}}\) for the human lineage. A hypothetical actor with a different access structure may retain it. Dark matter does not need to be consciously available to a human in order for the general space to leave room for an actor that can register some distinction associated with it.

### 3\. Objects of experience and categories

Now we can state the correction that keeps the framework from becoming stupid.

An **object of experience** is not generally a primitive element of \(\mathcal G\). It is a structured representation produced by an actor when external conditions encounter the actor's accessible structure and current state.

Let \(\omega_t\in\mathcal N_{\mathrm{loc}}\) denote an external configuration that may be experienced as an object, event, person, sentence, price, threat, chair, god, or anything else. Let \(\phi_{i,t}\) be the current phenomenal state and \(c_{i,t}\) the active context of actor \(i\). Write the actor-relative representation as

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

where

\[
A_{i,t}:\mathcal G\to\mathcal V_{i,t}^{\mathrm{acc}},
\qquad
\mathcal I_{i,t}:
\mathcal V_{i,t}^{\mathrm{acc}}\times\Phi_i\times\mathcal C_i^{\mathrm{ctx}}
\to
\mathcal V_{i,t}^{\mathrm{obj}}.
\]

The first map controls access; the second is the nonlinear organization that turns accessible distinctions plus current state into a phenomenal representation. Neither actor-relative space needs to be a linear subspace of \(\mathcal G\), and \(A_{i,t}\) need not be linear or orthogonal. These slowly changing maps and organizations will be bundled into the realized individual structure \(T_{i,t}\) in Part 2.

The same external configuration can therefore produce different object-representations in different actors, or in the same actor at different times.

A **category** requires a declared representation domain. Write that domain as

\[
\mathcal V_{\kappa}^{\mathrm{cat}}.
\]

Depending on the question, \(\mathcal V_{\kappa}^{\mathrm{cat}}\) may be GIDS itself, an actor-relative object space such as \(\mathcal V_{i,t}^{\mathrm{obj}}\), or a finite learned feature space. Once the domain and its measurable structure have been declared, a category may be represented as:

- a measurable region \(\mathcal C_\kappa\subseteq\mathcal V_{\kappa}^{\mathrm{cat}}\);
- a prototype point \(\operatorname{proto}_\kappa\in\mathcal V_{\kappa}^{\mathrm{cat}}\);
- a probability measure \(\mathbb P_\kappa\) on \(\mathcal V_{\kappa}^{\mathrm{cat}}\);
- or a learned scoring function
  \[
  \operatorname{cat}_{\kappa}:\mathcal V_{\kappa}^{\mathrm{cat}}\to[0,1].
  \]

The framework does not force every category to be an axis, or even to be an element of \(\mathcal G\). Many categories will be conglomerations over subtler coordinates and their interactions. If \(\mathcal V_{\kappa}^{\mathrm{cat}}=\mathcal G\), a prototype point is an element of GIDS and a category region is a subset of it; a probability measure or scoring function is still an object defined **on** the space rather than a vector inside it. If the category is actor-relative, its natural domain may instead be \(\mathcal V_{i,t}^{\mathrm{obj}}\). This is precisely what happens in ordinary factor analysis: the named construct is often a rough statistical compression over a more complicated underlying organization.

A proposition follows the same rule. The email, offer, threat, price, person, or meeting exists first as an external or symbolic configuration. What enters the actor's transition is its actor-relative representation:

\[
\mathbf p_{i,t}(x_t)
=
\mathcal P_{i,t}
\!\left(
 x_t,
\phi_{i,t},
 c_{i,t}
\right)
\in\mathcal V_{i,t}^{\mathrm{obj}}.
\]

The proposition does not need a dedicated proposition subspace. It needs an actor-relative representation in the distinctions available to the actor it is confronting. Part 3 will split \(\mathcal P_{i,t}\) into an external proposition encoder and a learned actor-relative map.

### 4\. A toy evolutionary rule for accessible dimensions

Now let evolutionary stage be indexed by \(\nu\), and let \(\mathcal M_\nu^{\mathrm{spec}}\) denote the lineage's current accessible structure at that stage. Let \(P_\nu^{\mathrm{spec}}\) be the orthogonal projection onto \(\mathcal M_\nu^{\mathrm{spec}}\) in this toy geometry.

A mutation or developmental modification proposes a candidate direction

\[
\Delta\mathbf v\in\mathcal G.
\]

Remove what the current structure already captures:

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

If \(\Delta\mathbf v_{\perp}=0\), the candidate adds no new representational direction under this model. If it is nonzero, normalize it:

\[
\widehat{\Delta\mathbf v}_{\perp}
=
\frac{\Delta\mathbf v_{\perp}}
{\|\Delta\mathbf v_{\perp}\|}.
\]

Let

\[
\mathfrak F_\nu(\mathcal M)
=
\mathbb E_{e\sim\mathcal E_\nu}
[W(e,\mathcal M)]
-
C_\nu(\mathcal M)
\]

be a finite real-valued toy evolutionary objective combining expected reproductive value and the total cost of maintaining an accessible structure. Define the contribution of the proposed direction directly by

\[
\Delta\mathfrak F_\nu
=
\mathfrak F_\nu\!\left(
\mathcal M_\nu^{\mathrm{spec}}
\oplus
\operatorname{span}\{\widehat{\Delta\mathbf v}_{\perp}\}
\right)
-
\mathfrak F_\nu(\mathcal M_\nu^{\mathrm{spec}}).
\]

Then the simplified retention rule is

\[
\mathcal M_{\nu+1}^{\mathrm{spec}}
=
\begin{cases}
\mathcal M_\nu^{\mathrm{spec}}
\oplus
\operatorname{span}\{\widehat{\Delta\mathbf v}_{\perp}\},
& \Delta\mathbf v_{\perp}\neq 0
  \text{ and }\Delta\mathfrak F_\nu>0,\\[6pt]
\mathcal M_\nu^{\mathrm{spec}},
& \text{otherwise}.
\end{cases}
\]

So the toy story is:

1. a mutation proposes a candidate distinction;
2. subtract what the lineage already represents;
3. compare the total value of the expanded structure against the total value of the old one;
4. retain the direction if the difference is positive.

This is not a complete population-genetic model. It ignores drift, frequency dependence, pleiotropy, exaptation, correlated features, and the possibility that old coordinates are rotated, merged, weakened, or deleted. It is a compact explanation of how a repertoire of distinctions could become richer over time.

### 5\. Deterministic silhouette and stochastic operation

The clean philosophical transition for the phenomenal component is

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

Read this as the ideal claim that a complete state, complete input, and complete transition law determine what follows. Over this explanatory step, the slowly changing person structure and the exogenous world are treated as fixed unless their updates are written explicitly.

For the complete dynamics, define

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

Let \(\boldsymbol\Xi_{t+1}\) be a random exogenous innovation and \(\boldsymbol\xi_{t+1}\) a realized or supplied scenario value. The notation below evaluates a transition kernel at the supplied scenario value; it does not require the singleton event \(\{\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}\}\) to have positive probability. The operationally honest ideal is

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

where \(K_i^\star\) is the unknown transition kernel over complete next states. A deterministic system is the special case in which that kernel collapses to a point mass. The first equation is the phenomenal component of this fuller transition.

This convention will run through the rest of the paper. I will keep arrows when the arrow makes the explanation easier to see. When we train and evaluate the model, the arrow expands into a distribution because the universe may be stochastic and our state estimate is certainly incomplete.

### Species template versus individual realization

The species-level structure says what kinds of distinctions a member of the lineage can, in principle, host. It is not yet the person.

Write the lineage-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 species-compatible family of interpretive organizations.

The inherited seed of person \(i\) is an individual realization inside that lineage-compatible family:

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

Here

\[
A_i^{0}:\mathcal G\to\mathcal M_i^{0}
\]

is the inherited access map at the chosen resolution; it is not assumed linear or orthogonal. The component \(\mathcal I_i^{0}\) denotes the inherited initial organization that later development can modify.

This lets biological variation, developmental constraints, and pathology alter the inherited starting point without pretending every member of the species begins with an identical coordinate system.

Part 2 now asks how \(G_i\) becomes the slowly changing realized structure \(T_{i,t}\), how that person occupies a phenomenal state, and how a low-resolution estimate can be extracted from outward traces without pretending the estimate is the whole interior.


---

# Part 2: Deriving the Transcendental Embedding

Part 1 described how evolution can carve a finite repertoire of distinctions out of God's Infinite Dimensional Space. That account explains why an organism has a "world" at all. It does not yet explain why one human inhabits that world differently from another human, even when both inherit roughly the same species-level template. Part 2 answers that question.

## The Technical Scope (because otherwise I'll accidentally lie to you)

Before I keep descending into Kantian hell, I need to pin down the scope so I do not smuggle Kant into the parts that are supposed to be engineering.

There are really three layers running through the rest of this paper.

> First, there is the **interpretive layer**: the noumenal/phenomenal story that motivates why an observer should have structured experience at all.

> Second, there is the **predictive layer**: the formal object the mathematics is actually allowed to touch. That object is **not** the whole ineffable mush of phenomenal life in itself, but a predictive representation that preserves the parts of an actor's response surface needed for the tasks we care about.

> Third, there is the **control layer**: once such a state can be estimated, we can rank or search over candidate propositions by their predicted effect on the actor's next state and downstream objective. That is the whole point. But causal claims there require intervention-grade data, not just retrospective logs and me getting excited.

For this formalism, an **actor** is a bounded system for which we can specify persistent state, channels through which the environment reaches it, a rule or distribution by which state changes, and outward traces by which those changes become observable. Consciousness is not required for operational actorhood. It matters to the philosophical claim about experience; it is not required to ask whether a system has enough persistence and response structure to be predicted.

The draft up to this point used one name, _Transcendental Embedding_, for several different things at once. From here onward I separate them. The transition is easier to read under somewhat dubious pretension than by dumping everything on you, the reader, at once:

> First, there is the **inherited template**: the repertoire of distinctions a human organism can in principle host inside their mental interior.

> Second, there is the **realized individual embedding**: the weighting, coupling, and organization of that template in one person after development, language, culture, memory, and repeated experience.

> Third, there is the **phenomenal state at a time**: the full lived condition of the organism now.

> Fourth, there is the **general predictive actor-state**: the response structure that determines how the actor is expected to change under a family of possible propositions.

> Fifth, there is the **task-conditioned predictive state**: the projection of that general response structure needed for one task and horizon.

> Sixth, there is the **estimated state**: the low-resolution object we can compute from outward traces.

Part 2 is the transition from the first object to the other five.

One more boundary matters. The first actor class in this research program is the individual human. A corporation can later be treated as an actor, but it is not generated by copying the human equations and changing a label. A corporation has to be assembled from the people inside it, the organization of authority between them, its institutional memory, its incentives, its recorded facts and statistics, and the environment in which it exists. The same outer logic of state, proposition, transition, and trace can apply after the actor has been constructed. The construction itself is different.

## Behold; You! The Chimera

Call the person-in-role object a **Chimera**. The term is mnemonic only. The theoretical work is done by the fact that a person is never encountered in the abstract, but always as a person under a role, inside an institution, in a regime, in a place, at a time. It would be computationally challenging if we did not introduce this categorization now, even though it amounts to a shortcut and a bastardization of the philosophical thesis.

> If an alien biologist watched human outputs only, human life would look repetitive. Much of what humans do can be reduced, at the level of gross behavior, to self-preservation, courtship, reproduction, kin-bonding, status competition, alliance formation, and resource control. The outer patterns recur. The difficulty lies elsewhere. Human beings often arrive at similar outputs by different internal routes.

Practically, for what I, the author, am interested in—GTM engineering—imagine one founder rejects a product because of caution. Another rejects it because of fear. Another because the price signals weakness. Another because the pitch activated a prior bad memory. Another because the role they occupy requires public skepticism. Same output, different internal geometry.

That difference is the point of this section.

Let \(G_i\) denote the inherited seed available to person \(i\). This is an individual realization inside the species-compatible repertoire of possible distinctions.

Let \(T_{i,t}\) denote the realized individual Transcendental Embedding of person \(i\) at time \(t\): the relatively durable, slowly changing organization produced when the inherited seed develops through one life.

Let \(\phi_{i,t}\) denote the total phenomenal state of person \(i\) at time \(t\).

Let \(c_{i,t}\) denote the active role-and-institution context at time \(t\): founder, buyer, parent, employee, soldier, friend, plus the relevant company, market, group, and local demands.

We can then define the person-in-role object as

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

This says something simple. The same person can yield different outputs across settings not because the person changes species, but because a different context activates a different organization of salience, inhibition, and available action. The person remains one person. The local geometry changes.

Those role-specific masks are highly specific to the individual and the regime rather than generic archetypes. They can preserve opposed dispositions long enough for the model to ask whether the opposition is a true contradiction or whether it resolves along a deeper axis activated by the regime.

## Psychology and Factor Analysis

Psychology already contains rough tools for decomposition. Factor analysis took piles of correlated observations and produced constructs that were easier to name, compare, and compute. In a standard common-factor model,

\[
\mathbf x_i
=
\boldsymbol\mu_x
+
\boldsymbol\Lambda_{\mathrm{FA}}\mathbf f_i
+
\boldsymbol\varepsilon_i,
\qquad
\mathbb E[\mathbf f_i]=\mathbf 0,
\qquad
\operatorname{Cov}(\mathbf f_i)=I,
\qquad
\mathbb E[\boldsymbol\varepsilon_i\mid\mathbf f_i]=\mathbf 0,
\]

where \(\mathbf x_i\) is a vector of observed measurements, \(\boldsymbol\mu_x\) is the observed-variable mean, \(\mathbf f_i\) is a lower-dimensional latent-factor vector, \(\boldsymbol\Lambda_{\mathrm{FA}}\) contains factor loadings, and \(\boldsymbol\varepsilon_i\) is residual variation. If the measurements are centered, \(\boldsymbol\mu_x=\mathbf 0\).

Under the conventional standardized-factor identification above, the factor coordinates are not uniquely oriented. For an orthogonal matrix \(R\) satisfying \(R^{\top}R=I\),

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

The common component is unchanged, and the standardized factor covariance remains the identity because \(R^{\top}R=I\). So the same fitted covariance structure can survive a rotation of the latent coordinates. Named constructs require identification conventions and interpretation; the mathematics does not hand us one sacred psychological axis.

This was useful. It also produced crude conglomerations.

A named personality construct is generally not one primitive feature of the mind. It is a statistical compression over many subtler tendencies: what the person notices, what they count as threatening, how quickly they generalize, how they respond to uncertainty, how they imagine other minds, how they trade status against safety, how strongly they carry prior events into the present, and so on. The factor is real in the practical sense that it summarizes repeatable covariance. It is not necessarily a fundamental axis of experience.

That is the basic way to understand what GIDS is trying to do here. We are doing something similar to the factor-analytic method, but we are trying to push beneath the crude named constructs and recover more granular vectors of how people understand and operate in the world. We also refuse to freeze the whole person into one timeless score. Some structure is durable. Some is activated only in one role. Some changes after one event. Some exists only as an interaction between the person and the proposition.

Write a psychometric proxy for person \(i\) as

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

where the coordinates may include IQ-like measures, psychometric traits, moral scales, behavioral factors, or related standardized summaries.

This object is useful, but it is not the Transcendental Embedding itself.

A factor score is not a memory field.  
It is not a role.  
It is not a present state.  
It is not a proposition.  
It is not a transition rule.

What it does provide is a **coarse prior**. It places a person inside a region of likely behavior. That is enough to matter, but not enough to solve micro-interaction. Factor analysis may tell us that a person is threat-sensitive, novelty-seeking, rigid, verbal, impulsive, or dutiful. It does not tell us whether those coordinates are active now, under this framing, with this memory already cued, inside this role.

A candidate feature deserves to be called more fundamental only after it repeatedly pays rent. It should remain useful across time when the person is stable, transfer across related tasks and contexts, interact sensibly with fast state, improve prediction beyond simpler constructs, and—when intervention data exists—respond to propositions in the direction the model predicts. Even then, the coordinate chart is not sacred. Rotations and other reparameterizations may preserve the same predictive information.

So psychology enters Part 2 as one source of approximation, not as the final ontology. To make the problem tractable from an engineering standpoint, we also need to append source-aware categorical traces to the estimation of a person's Transcendental Embedding: role labels, objection classes, recurring topics, counterpart identities, action-types, firm-state tags, and other discrete observations that accumulate over time.

Those categorical traces are not the Transcendental Embedding in itself. They are an operational bridge from outward history to the estimated observer-side object. The important engineering choice is to avoid collapsing biography, stated language, observed behavior, and third-party inference into one immediate average merely because they sound semantically related. Some channels later become comparable; they should not be assumed comparable at ingestion.

> this is very up to you on specific implementation details, tbh, be creative

## Dimensionality Reduction, Attention, and Relevance

The issue is not that every decision has one permanent principal axis in the person. The problem compounds in difficulty because the individual organization contains many coordinates and relations, while any given transition is usually governed by a weighted subset of them.

Let \(x_t\) be the present proposition or stimulus. At the ideal level, define a task-conditioned relevance map

\[
\Lambda_{\tau}
\!\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
 x_t
\right)
\in\mathbb R^{d_\tau},
\]

where \(\tau\) indexes the task under study. The symbol \(\pi\) is deliberately not used here because it later denotes a decision policy.

Let attention or salience be represented by

\[
\boldsymbol\lambda_{i,t}^{(\tau)}(x_t)
\in[0,1]^{d_\tau}.
\]

Then the ideal active proposition-conditioned slice is

\[
\mathbf r_{i,t}^{(\tau)}(x_t)
=
\boldsymbol\lambda_{i,t}^{(\tau)}(x_t)
\odot
\Lambda_{\tau}
\!\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
 x_t
\right),
\]

where \(\odot\) denotes elementwise weighting.

This is the conceptual claim: the person may occupy a large space, but the next transition often depends on a smaller weighted slice of that space. The task is therefore not to discover one universal "main factor" of the person. The task is to discover which coordinates and interactions carry signal for a transition under a task, a proposition, and a context.

Attention matters because not all dimensions are weighted equally at every moment. The environment does not strike the whole embedding uniformly. A sentence, a person, a price, or a memory cue activates some coordinates and leaves others inert. That is why identical prompts can produce different outputs at different times.

The ideal expression is not directly computable because \(T_{i,t}\) and \(\phi_{i,t}\) are not observed. Later in this part, the learned approximation replaces them with the slow vector \(\widehat{\mathbf t}_{i,t}\), the fast vector \(\mathbf z_{i,t}\), and the explicit context carried by the operational state. The symbol \(\mathbf z_{i,t}\) is reserved for the fast state only; it is not reused for this proposition-conditioned slice.

## The Notion of State

Philosophically, everything belongs inside state.

The phenomenal state includes what is perceived, what is remembered, what is felt, what is attended to, what is being done, what bodily changes are underway, and what action tendencies are presently live. In that sense, the state is total.

Let that total state be

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

That is the full object.

If I leave the formal section aimed directly at \(\phi_{i,t}\), however, I start overclaiming almost immediately. So from here on I distinguish the motivating object from the objects the mathematics is actually allowed to touch.

Define the ideal pre-proposition information state

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

where \(\mathcal H_{i,<t}\) contains the history available before the proposition at decision time \(t\), \(T_{i,t}\) is the slowly changing realized person structure, \(c_{i,t}\) is active context, and \(w_t\) is the relevant external world state. This is the ideal information bundle relative to which the prediction problem is defined; it is not the complete actor–world state \(\Sigma_{i,t}^{\star}\), because it does not grant the predictor direct access to the full phenomenal state. Assume the relevant state and outcome spaces are standard Borel spaces so that the conditional laws below exist in the ordinary regular sense.

The first predictive object is a **general predictive actor-state**, denoted

\[
Q_{i,t}.
\]

Intuitively, \(Q_{i,t}\) represents the actor's response surface: everything in \(\mathsf I_{i,t}\) that still matters for how the actor would change under a family of admissible future propositions.

For a finite decision horizon \(H\), let

\[
\mathbf x_{t:t+H-1}
=
(x_t,x_{t+1},\ldots,x_{t+H-1})
\]

be an admissible open-loop proposition sequence. Let \(O_{i,t+1:t+H}\) denote the corresponding future observable trace sequence. Let \(\boldsymbol\Xi_{t+1:t+H}\) denote the random exogenous path and \(\boldsymbol\xi_{t+1:t+H}\) a supplied scenario value, including market, personnel, company, and role changes.

Choose a version of the observational regular conditional response kernel

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

This conditional kernel is defined only up to almost-sure equality and is operationally relevant on the support of the proposition and scenario process. If future propositions are adaptive, the factual joint law under the historical policy must include the policy and candidate-set process that generated them. Merely inserting a new policy into the conditioning index does not identify its counterfactual response law from observational data, even when its individual actions have support. A counterfactual adaptive policy requires controlled transition kernels plus an identification argument, or an explicit extrapolation assumption. A causal response surface uses the corresponding interventional laws under proposition interventions rather than merely conditioning on historically selected propositions.

Two ideal information states are equivalent when they induce the same declared family of response laws for every admissible proposition sequence or policy, scenario regime, planning horizon, and measurable future event in the family being studied. This equivalence relation defines the ideal predictive information object abstractly. It is **not** automatic that the quotient by this relation inherits a convenient standard-Borel measurable structure. Whenever a usable measurable realization exists, denote that realization by \(Q_{i,t}\). Otherwise, the indexed family of response kernels itself is the predictive object, and no finite or conveniently measurable quotient is claimed. The scenario path is explicit because a recursive model does not earn the right to freeze the future environment merely for mathematical convenience.

This object may be infinite-dimensional. That is fine. The ideal response surface can be much larger than the finite approximation we learn.

The second predictive object is a task-conditioned summary of that state. For task \(\tau\) and elapsed-time outcome horizon \(\Delta\), write

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

when a measurable map with the required sufficiency property exists. \(\Pi_{\tau,\Delta}\) need not be linear or orthogonal. It is called a projection only in the loose sense that it keeps what one task and horizon require while discarding the rest.

Let

\[
Y_{i,t}^{(\tau,\Delta)}
\]

denote the outcome associated with the decision at time \(t\), evaluated at or by horizon \(\Delta\): reply, meeting, objection class, delay bucket, stage advance, sentiment shift, or whatever the task actually cares about. Writing the outcome this way prevents a multi-horizon decision row from pretending that every target occurs at one common timestamp.

For every measurable outcome set \(B\), the task-conditioned state is sufficient when, for each admissible proposition value \(x\) in the predictive support,

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

Read that in English: once the same proposition value is conditioned on, everything in the ideal information state that still matters for this future has already been compressed into \(q_{i,t}^{(\tau,\Delta)}\). The causal version replaces this observational conditioning statement with an intervention-indexed one.

At this point the expression is predictive, not automatically causal. If propositions in the historical data were selected in a biased way, the learned conditional law is a forecasting law. A causal predictive state would have to preserve the corresponding interventional laws under \(\operatorname{do}(X_t=x)\), which requires the experimental machinery introduced later.

## Observable Predictive State

For engineering work, the key move is to define the formal state in terms of observable consequences rather than inaccessible total interiority.

The hierarchy is:

\[
\phi_{i,t}
\quad\text{full phenomenal state},
\]

\[
Q_{i,t}
\quad\text{general predictive response state},
\]

\[
q_{i,t}^{(\tau,\Delta)}
\quad\text{task-and-horizon summary},
\]

and

\[
\widehat s_{i,t}
\quad\text{measurable operational approximation}.
\]

The paper gets much more honest the second these stop being treated as the same thing.

The operational approximation I actually want to estimate is

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

where \(\widehat{\mathbf t}_{i,t}\in\mathbb R^{d_T}\) is the slow estimated person vector, \(\mathbf z_{i,t}\in\mathbb R^{d_Z}\) is the fast latent vector inferred from records available before \(t\), and \(\mathbf c_{i,t}\) and \(\mathbf w_t\) are finite encodings of context and world state.

The old notation \(\widehat s_{i,t}\approx q_{i,t}^{(\tau,\Delta)}\) was too vague. Approximation needs an operational meaning, and representation error must be separated from model-fitting error.

If \(\widehat s_{i,t}\) is a measurable function of \(\mathsf I_{i,t}\), define the state-compression gap

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

Equivalently,

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

where \(Y=Y_{i,t}^{(\tau,\Delta)}\). This term is zero exactly when the operational state is sufficient under the joint observational law and its supported propositions, up to null sets. Uniform or causal sufficiency across interventions requires the corresponding family of interventional laws.

A fitted conditional law \(P_{Y,\theta,\tau,\Delta}(\cdot\mid\widehat s,X)\) introduces a separate model-estimation gap:

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

To state the log-score identity without quietly assuming a discrete outcome, suppose the relevant conditional laws are dominated by one fixed reference measure on the outcome space. Let

\[
p^{\star}_{\tau,\Delta}(y\mid\mathsf I_{i,t},X_t)
\]

be a version of the full-information conditional density or mass function, let

\[
p_{\widehat s,\tau,\Delta}(y\mid\widehat s_{i,t},X_t)
\]

be the true conditional density or mass function after compression, and let

\[
p_{Y,\theta,\tau,\Delta}(y\mid\widehat s_{i,t},X_t)
\]

be the fitted density or mass function. Define the full-information Bayes log risk

\[
\mathscr R_{\log}^{\star}
:=
\mathbb E\!\left[
-\log p^{\star}_{\tau,\Delta}
\!\left(
Y_{i,t}^{(\tau,\Delta)}
\mid
\mathsf I_{i,t},X_t
\right)
\right].
\]

Whenever the displayed expectations and divergences are finite, the exact decomposition is

\[
\begin{aligned}
&\mathbb E\!\left[
-\log p_{Y,\theta,\tau,\Delta}
\!\left(
Y_{i,t}^{(\tau,\Delta)}
\mid
\widehat s_{i,t},X_t
\right)
\right]
-
\mathscr R_{\log}^{\star}
\\[4pt]
&\qquad=
\epsilon_{\tau,\Delta}^{\mathrm{state}}(\widehat s)
+
\epsilon_{\theta,\tau,\Delta}^{\mathrm{model}}(\widehat s).
\end{aligned}
\]

For a discrete outcome with counting measure, \(\mathscr R_{\log}^{\star}=H(Y_{i,t}^{(\tau,\Delta)}\mid\mathsf I_{i,t},X_t)\). For continuous outcomes it is a conditional expected negative log density, not an invariant “entropy of the state.” The first gap is information discarded by compression; the second is error in the fitted law conditional on that compressed state. Ordinary held-out log loss estimates total predictive risk, which also contains the irreducible full-information Bayes risk; differences against sufficiently strong references and targeted ablations are needed to diagnose whether failure came from state compression or from the predictor built on top of it.

The general state \(Q_{i,t}\) is supposed to support many task summaries. That is how the claim that "the ontology is the product" becomes testable. A person representation that predicts only one narrow label may simply be a task shortcut. A better actor-state should transfer across related outcomes, horizons, roles, and proposition families.

## Memory as a Series of Vectors

For present purposes, memory need not be treated as narrative first. It can be modeled as a field of traces with weights.

Let

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

where \(\mathbf h_{ij}^{\mathrm{mem}}\) is a stored trace representation and \(\varpi_{ij,t}\) is its weight at time \(t\).

Some traces are weak. Some are strong. Some decay. Some reactivate under similarity, emotion, role, or repetition. The point is not that memory is literally a vector sum in the brain; the point is that weighted trace structure gives us a tractable model of persistence and retrieval.

A present proposition does not encounter the whole memory field evenly. Before the response is observed, proposition-conditioned retrieval can be written schematically as

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

The proposition-conditioned retrieved memory is then

\[
\widetilde{\mathbf m}_{i,t}^{\mathrm{mem}}(x_t)
=
\sum_{j=1}^{N_i}
\widetilde\varpi_{ij,t}(x_t)
\mathbf h_{ij}^{\mathrm{mem}}.
\]

After the actual trace \(o_{i,t+1}\) arrives, the memory state can be updated by

\[
\mathbf m_{i,t+1}^{\mathrm{mem}}
=
U_{\mathrm{mem},\theta}
\!\left(
\mathbf m_{i,t}^{\mathrm{mem}},
 x_t,
 o_{i,t+1},
\mathbf c_{i,t}
\right).
\]

This gives memory two jobs.

First, it stores prior traces.  
Second, it changes which parts of the person-space are active now.

This also explains why repeated prompts can produce different outputs. The second encounter is not with the same person-state as the first. The first encounter has already changed the trace structure. A prior positive or negative experience can therefore make the second prediction harder, not easier, if the model fails to represent that update.

Recommendation systems provide a useful analogy here. A view history is not a mind. But a recommender does show the core move: repeated traces can be compressed into a latent representation that improves prediction. In the present framework, biography, language, preferences, recurrent actions, role history, and prior interactions play the role of trace data from which a person-level state can be estimated.

## Categorical Trace Pooling as an Operational Memory Estimator

A large part of what we observe about people arrives in categorical form: role labels, recurring topics, objection classes, counterpart identities, action-types, domain tags, product themes, price postures, and other discrete markers. Rather than treat these as dead one-hot tables or discard them into prose, we can embed them and pool them over time.

Operationally, the implementation assumes a fixed global registry of categorical families, source channels, and admissible regimes, with learned null vectors and mask bits for absent cells so the resulting representation stays fixed-width across people and time.

Let \(f\in\{1,\ldots,F\}\) index categorical families and let \(\sigma\in\mathcal S_{\mathrm{src}}\) index source channels such as biography, stated language, observed behavior, and third-party or inferred traces. For person \(i\), the notation \(r<t\) below means that record \(r\) was timestamped and available before decision epoch \(t\). Let

\[
\mathcal B_{i,r}^{(f,\sigma)}
\]

be the multiset of raw category tokens in family \(f\) from source \(\sigma\).

Before pooling, surface labels should be contextually typed. A token that looks contradictory in the raw may become perfectly consistent once we mark whether it concerns self versus other, in-group versus out-group, own-interest versus third-party interest, formal stance versus enacted stance, or another asymmetry carried by the regime. Write this contextual lifting as

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

Only the opposition that remains after this lifting deserves to be treated as a genuine contradiction.

Let

\[
E_{f,\sigma}:\mathcal V_{f,\sigma}\to\mathbb R^{d_{f,\sigma}}
\]

be the embedding table for that family and source, and let \(\mathbf e_{\varnothing}^{(f,\sigma)}\in\mathbb R^{d_{f,\sigma}}\) be a learned empty-bag vector. The event-level pooled representation is

\[
\mathbf u_{i,r}^{(f,\sigma)}
=
\begin{cases}
\displaystyle
\frac{1}{|\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}|}
\sum_{\upsilon\in\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}}
E_{f,\sigma}(\upsilon),
& |\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}|>0,\\[10pt]
\mathbf e_{\varnothing}^{(f,\sigma)},
& |\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}|=0.
\end{cases}
\]

Define the binary availability mask

\[
M_{i,r}^{(f,\sigma)}
=
\mathbf 1\{|\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}|>0\}.
\]

Let \(W_{f,\sigma}^{\mathrm{cat}}:\mathbb R^{d_{f,\sigma}}\to\mathbb R^{d_*}\) be a learned alignment map that gives every slot a common output width while preserving family and source identity. Concatenate the cells:

\[
\mathbf e_{i,r}^{\mathrm{cat}}
=
\big\|_{(f,\sigma)}
\left[
W_{f,\sigma}^{\mathrm{cat}}\mathbf u_{i,r}^{(f,\sigma)},
 M_{i,r}^{(f,\sigma)},
 \log\!\left(1+|\widetilde{\mathcal B}_{i,r}^{(f,\sigma)}|\right)
\right].
\]

This is the recommender move in its simplest form: sparse categorical IDs are mapped to dense vectors and multivalent bags are pooled into fixed-width representations. The count term prevents an event containing one token from becoming automatically identical to an event containing the same token many times. But I do not want the next step to be a naive global average across every source and every role. Categories that arrive through speech, biography, and behavior are not automatically the same thing just because they share a label. They can later become comparable; they should not be forced into comparability at ingestion.

Let

\[
\rho_r:=\operatorname{role}(c_{i,r})
\]

denote the regime or role at event \(r\). Let the slow weights satisfy \(\beta_{i,r,t}^{\mathrm{slow}}\ge 0\), and define the slow evidence mass

\[
\beta_{i,\rho,+,t}^{\mathrm{slow}}
:=
\sum_{r<t}
\mathbf 1\{\rho_r=\rho\}
\beta_{i,r,t}^{\mathrm{slow}}.
\]

The slow categorical memory for regime \(\rho\) is

\[
\mathbf g_{i,\rho,t}^{\mathrm{slow}}
=
\begin{cases}
\displaystyle
\frac{
\sum_{r<t}
\mathbf 1\{\rho_r=\rho\}
\beta_{i,r,t}^{\mathrm{slow}}
\mathbf e_{i,r}^{\mathrm{cat}}
}{\beta_{i,\rho,+,t}^{\mathrm{slow}}},
& \beta_{i,\rho,+,t}^{\mathrm{slow}}>0,\\[14pt]
\mathbf e_{\varnothing,\rho}^{\mathrm{slow}},
& \beta_{i,\rho,+,t}^{\mathrm{slow}}=0,
\end{cases}
\]

where \(\mathbf e_{\varnothing,\rho}^{\mathrm{slow}}\) is a learned empty-regime vector. Define the availability mask

\[
M_{i,\rho,t}^{\mathrm{slow}}
=
\mathbf 1\{\beta_{i,\rho,+,t}^{\mathrm{slow}}>0\}.
\]

The complete slow bank retains the pooled content, explicit availability, and total evidence mass:

\[
\mathbf g_{i,t}^{\mathrm{slow}}
=
\big\|_\rho
\left[
\mathbf g_{i,\rho,t}^{\mathrm{slow}},
 M_{i,\rho,t}^{\mathrm{slow}},
 \log\!\left(1+\beta_{i,\rho,+,t}^{\mathrm{slow}}\right)
\right].
\]

For task-conditioned retrieval from recent categorical history, let the nonnegative relevance weights satisfy \(\alpha_{i,r,t}^{(\tau)}\ge 0\), and define

\[
\alpha_{i,+,t}^{(\tau)}
=
\sum_{r<t}\alpha_{i,r,t}^{(\tau)}.
\]

The normalized content summary is

\[
\overline{\mathbf g}_{i,t}^{\mathrm{fast},\tau}
=
\begin{cases}
\displaystyle
\frac{1}{\alpha_{i,+,t}^{(\tau)}}
\sum_{r<t}
\alpha_{i,r,t}^{(\tau)}
\mathbf e_{i,r}^{\mathrm{cat}},
& \alpha_{i,+,t}^{(\tau)}>0,\\[12pt]
\mathbf e_{\varnothing,\tau}^{\mathrm{fast}},
& \alpha_{i,+,t}^{(\tau)}=0.
\end{cases}
\]

Then retain both content and accumulated evidence:

\[
\mathbf g_{i,t}^{\mathrm{fast},\tau}
=
\left[
\overline{\mathbf g}_{i,t}^{\mathrm{fast},\tau},
\log\!\left(1+\alpha_{i,+,t}^{(\tau)}\right)
\right].
\]

This is a task-conditioned retrieval from recent categorical history, not the fast state itself. The shared fast state \(\mathbf z_{i,t}\) is updated chronologically from observed events. The retrieved pool may then enter the task-relevance map that asks which parts of that shared state matter now. Keeping the relevance mass separate is necessary because a normalized average alone cannot distinguish one weak exposure from the same weak exposure repeated twenty times.

The weighting laws should not treat every trace equally. They may depend on recency, task relevance, regime relevance, action intensity, repeated weak exposure, susceptibility, and source reliability. Decisive action traces should often outrank passive exposure traces, while repeated weak exposures should still accumulate over time according to the person's susceptibility.

This is also where strategic self-presentation enters the model: stated concern, biographical prior, and observed behavior are allowed to disagree without being collapsed at ingestion.

The slow categorical bank is therefore best read as what the person is generally like **now**, at this stage and across regimes, rather than as a timeless essence. The fast pool captures what is currently active for a task. Neither is the phenomenal state itself.

## Minimality, Identifiability, and Slow/Fast Factorization

Now comes the part that keeps the whole thing from dissolving into vibes.

A predictive state is not interesting merely because it is sufficient. A gigantic archive is sufficient too. The point is to retain the information the future still cares about without dragging the whole archive behind every prediction forever.

Say that a task-conditioned predictive state \(q_{i,t}^{(\tau,\Delta)}\) is **minimal** when, for any other sufficient state \(r_{i,t}^{(\tau,\Delta)}\), there exists a measurable map \(h\) such that

\[
q_{i,t}^{(\tau,\Delta)}
=
h\!\left(r_{i,t}^{(\tau,\Delta)}\right)
\]

almost surely.

This is the right level of humility. I am not claiming there is one mystical coordinate chart for the soul. I am claiming that, for a task, there may be a smallest predictive information object. Any two minimal sufficient states generate the same predictive sigma-field up to null sets; under standard Borel regularity they can be related by measurable inverse maps on full-measure subsets. The manuscript does not assume that such a finite-dimensional minimal state exists for every possible task; that is an empirical and structural question.

The operational state is split into slow and fast pieces:

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

Let \(\mathbf u_{i,t}^{\mathrm{dur}}\) collect durable person-side proxies, including the slow categorical bank. The slow estimate is

\[
\widehat{\mathbf t}_{i,t}
=
E_{T,\theta}\!\left(
\mathbf u_{i,t}^{\mathrm{dur}},
\mathcal H_{i,<t}^{\mathrm{durable}}
\right).
\]

Let \((\mathsf h_n)_{n\ge1}\) be the chronological record sequence, with \(v_n:=\operatorname{time}(\mathsf h_n)\). When the stream is shared across actors, let \(\mathbf d_{i,n}^{\mathrm{rec}}\) record whether record \(n\) pertains to actor \(i\) and which actor-specific fields are available. Starting from \(\mathbf z_{i,0}\), the fast state after processing record \(n\) is

\[
\mathbf z_{i,n}
=
U_{z,\theta}\!\left(
\mathbf z_{i,n-1},
\widehat{\mathbf t}_{i}(v_n^-),
\mathbf c_i(v_n^-),
\mathbf w(v_n^-),
\mathsf h_n,
\mathbf d_{i,n}^{\mathrm{rec}}
\right),
\qquad n\ge1.
\]

The applicability vector is not a memory vector and not an outcome-availability mask. It contains a binary applicability flag, and when that flag is zero the update is required to leave the actor's fast state unchanged.

At decision epoch \(t\), let

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

This converts the asynchronous record clock into the decision-aligned state used by the model.

The reason for the split is not aesthetic. Different things change on different timescales. A founder does not become a different founder because of one email. But their local state can absolutely change because of one email. The slow term is supposed to carry durable person structure. The fast term is supposed to carry within-window state needed to preserve the predictive content of recent history.

If this split is real, removing \(\mathbf z_{i,t}\) should hurt short-horizon prediction. Removing \(\widehat{\mathbf t}_{i,t}\) should hurt cold-start performance and cross-context generalization. If neither happens, I do not get to pretend the decomposition was profound. Part 4 will force that issue.

Identifiability remains limited. A learned latent space can be rotated, rescaled, or otherwise reparameterized without changing its predictions. The empirical target is therefore not one sacred coordinate system. It is stable predictive information, recoverable structure, transfer, and intervention behavior.

## Deriving the Transcendental Embedding

We can now write the distinction that Part 1 left implicit.

Let \(G_i\) be the inherited seed: the individual lineage-compatible repertoire and organization available at the beginning of development.

Let

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

be the realized individual Transcendental Embedding at time \(t\), where \(\mathbf u_{i,<t}^{\mathrm{lang}}\) denotes language, culture, and socialization recorded before decision \(t\), and \(\mathcal H_{i,<t}^{\mathrm{life}}\) is life history recorded before that decision.

Life history is fundamentally a sequence, not a bag. For a crude summary, one may write

\[
\overline{\mathbf h}_{i,t}
=
\sum_{r<t}
\beta_{i,r,t}\mathbf v_{i,r}^{\mathrm{event}},
\qquad
\beta_{i,r,t}\ge 0,
\]

where \(\mathbf v_{i,r}^{\mathrm{event}}\) is an event representation and \(\beta_{i,r,t}\) is its later weight. The sequence model should still retain order when order matters.

Some events contribute little.  
Some events bend the later space of response.

This is the theoretical object.

But we do not observe \(T_{i,t}\) directly. What we observe are traces and proxies. Let

- \(\boldsymbol\psi_i\) be psychometric and cognitive summaries;
- \(\mathbf b_i\) be biography and background;
- \(\mathbf u_{i,t}^{\mathrm{lang}}\) be the language and cultural-position features available before decision \(t\);
- \(\mathbf v_{i,t}^{\mathrm{role}}\) be role and institution history features;
- \(\overline{\mathbf h}_{i,t}\) be a weighted summary of observable life-event structure;
- \(\mathbf g_{i,t}^{\mathrm{slow}}\) be the slow source-aware and regime-aware categorical trace bank.

Collect these in

\[
\mathbf u_{i,t}^{\mathrm{dur}}
=
\left[
\boldsymbol\psi_i,
\mathbf b_i,
\mathbf u_{i,t}^{\mathrm{lang}},
\mathbf v_{i,t}^{\mathrm{role}},
\overline{\mathbf h}_{i,t},
\mathbf g_{i,t}^{\mathrm{slow}}
\right].
\]

Then a first operational estimate is

\[
\widehat{\mathbf t}_{i,t}^{(0)}
=
E_{T,\theta}^{(0)}\!\left(
\mathbf u_{i,t}^{\mathrm{dur}}
\right).
\]

At first pass, this initial encoder can be additive:

\[
\widehat{\mathbf t}_{i,t}^{(0)}
=
W_\psi\boldsymbol\psi_i
+
W_b\mathbf b_i
+
W_{\mathrm{lang}}\mathbf u_{i,t}^{\mathrm{lang}}
+
W_{\mathrm{role}}\mathbf v_{i,t}^{\mathrm{role}}
+
W_h\overline{\mathbf h}_{i,t}
+
W_g\mathbf g_{i,t}^{\mathrm{slow}}.
\]

Each matrix maps its input into the common slow-state dimension \(d_T\). If one channel suppresses or inverts another, that sign belongs inside the learned operator rather than being hard-coded as a minus sign on the whole source.

This weighted estimate is **not** the Transcendental Embedding itself. It is the tractable object from which we begin. It is a low-resolution approximation built from standardized signals because those are the signals computation can access at scale.

As chronological records become available, update the record-indexed fast state by the rule above and use

\[
\mathbf z_{i,t}=\mathbf z_{i,N(t)}
\]

as the state available at decision epoch \(t\).

Once the slow estimate and fast state are available, the local operational object is

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

Given a proposition \(x_t\), the learned proposition-conditioned active representation is

\[
\widehat{\mathbf r}_{i,t}^{(\tau)}(x_t)
=
\boldsymbol\lambda_{\theta,\tau}
\!\left(
\widehat s_{i,t},x_t,\mathbf g_{i,t}^{\mathrm{fast},\tau}
\right)
\odot
\Lambda_{\theta,\tau}
\!\left(
 \widehat s_{i,t},
 x_t,
\mathbf g_{i,t}^{\mathrm{fast},\tau}
\right),
\]

where the learned salience map returns a vector in \([0,1]^{d_\tau}\) and the learned relevance map returns a vector in \(\mathbb R^{d_\tau}\). This active slice is an interaction feature. It is not the fast state \(\mathbf z_{i,t}\), and it is not the theoretical sufficient statistic \(q_{i,t}^{(\tau,\Delta)}\).

Let \(\boldsymbol\Xi_{t+1}\) denote the random exogenous change between decision epochs and \(\boldsymbol\xi_{t+1}\) a realized or supplied scenario value. Evaluating the parameterized transition kernel at \(\boldsymbol\xi_{t+1}\) defines a scenario forecast; it does not require a positive-probability singleton event. An unconditional forecast must instead integrate the conditional transition kernel against a declared law for \(\boldsymbol\Xi_{t+1}\). The recursively usable learned model evolves the operational state:

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

A delayed task outcome has a separate predictive law,

\[
P_{Y,\theta,\tau,\Delta}
\!\left(
\cdot
\mid
\widehat s_{i,t},
X_t=x_t
\right).
\]

For a one-step task this law may be read from the next state and immediate observation. For \(\Delta\) extending beyond the next decision epoch, the law is defined only relative to a declared continuation policy, future candidate-set process, exogenous-path regime, and censoring convention, unless those variables are conditioned on directly. It may be implemented by a rollout or by a direct head calibrated to that same regime. It must not be presented as if a 90-day outcome were an immediate emission from one next-state sample.

Part 2 stops here on purpose.

We cannot yet claim to know the true form of the transition kernel.  
In no way am I claiming to have solved qualia or consciousness; I have just, maybe, created a representation that keeps me from talking nonsense while trying to predict human transition.  
It does not claim that the estimate and the reality itself are identical.

What it does claim is smaller and enough for the next step: a person can be represented as a latent structure derived from an inherited seed, development, language, culture, memory, and repeated life events; that structure can be estimated from outward traces; and the estimate can serve as the person-side state in a transition model whose recursively modeled target is future operational state and whose task heads predict observable behavior.

Part 3 can now ask the narrower question: once a person has been represented by \(\widehat{\mathbf t}_{i,t}\), once recent dynamics have been represented by \(\mathbf z_{i,t}\), and once the person-in-role state has been represented by \(\widehat s_{i,t}\), how do we represent the proposition without pretending the proposition is the same kind of thing as the person, and how do we learn a transition map that predicts what happens next with increasing fidelity?


---

# Part 3: Application — Predicting How People Behave

Part 2 established the person-side object. It distinguished the inherited seed from the realized individual structure, the realized structure from the momentary phenomenal state, the momentary phenomenal state from the general predictive response state, and that ideal response state from the finite object the model can actually estimate. Part 3 now asks what can be done with it.

The claim of this section is narrower than a final proof about mind. It is not that we already possess the exact universal transition law for human beings. It is that we can define a mathematical program in which an actor, a proposition, the actor's environment, and the history between them can be represented through a shared coordinate arena, and that inside this program we can iteratively improve our estimate of how one state gives rise to the next.

In other words: this section does not complete the science; it specifies the playground in which the science can be built.

Everything the model computes on eventually becomes vectors, tensors, distributions, or tuples of those things. Raw inputs can still be text, categories, histories, metadata, prices, organizational charts, people, products, and context objects. A vector is the representation entering an operation. It is not the claim that the original thing was secretly a vector before we touched it.

The canonical symbol definitions are in [Canonical Notation and Mathematical Conventions](08_Canonical_Notation.md).

## Towards a Universal State Transition Function

In the ideal philosophical form of the theory, the object of interest is still the next phenomenal state.

Let \(\phi_{i,t}\) denote the full phenomenal state of person \(i\) at time \(t\), let \(T_{i,t}\) denote the slowly changing realized Transcendental Embedding of that person, let \(c_{i,t}\) denote the active role-and-institution context, let \(w_t\) denote world state, and let \(x_t\) denote the proposition confronting the observer at decision time \(t\).

Here "proposition" is meant broadly. It may be a sentence, an email, a product, a person, a meeting, a threat, a market signal, a price, or a whole local arrangement of circumstances. For the observer, what matters is not bare matter but presented structure.

Define the complete ideal actor–world state

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

The familiar deterministic silhouette

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

is the phenomenal component of a fuller transition in which the slow person, context, and world may also change. Let \(\boldsymbol\Xi_{t+1}\) denote the random exogenous change between decision epochs and \(\boldsymbol\xi_{t+1}\) a realized or supplied scenario value. The notation below evaluates a transition kernel at the supplied scenario value; it does not require the singleton event to have positive probability. The operationally honest ideal law is

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

I keep the deterministic equation because it is the motivating silhouette. It says that if the complete actor–world state and exact transition rule were known, the next state would follow. The stochastic kernel permits genuine randomness; our operational uncertainty is larger still because the complete state is not observed. A deterministic system is the special case in which the kernel places all of its mass on one next state.

The learned model therefore works with the operational state

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

Its recursively usable prediction is a distribution over the next operational state:

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

where \(\boldsymbol\Xi_{t+1}\) is the random exogenous change and \(\boldsymbol\xi_{t+1}\) is the realized or supplied scenario value between decision epochs.

The outcome associated with decision \(t\) and horizon \(\Delta\) has a separate predictive law:

\[
P_{Y,\theta,\tau,\Delta}
\!\left(
\cdot
\mid
\widehat s_{i,t},
X_t=x_t
\right).
\]

For a one-step task, this law may be induced by the next state and its immediate observation. For a longer horizon, it is defined relative to a declared continuation policy, future candidate-set process, exogenous-path law, and censoring convention unless those variables are conditioned on explicitly. It may be induced by a rollout or represented by a direct head calibrated to that same regime. A 90-day close is not an immediate emission one step after an email.

The reply, the purchase, the rejection, the delay, the meeting, or the concession is not the state itself. It is a visible residue of a transition and, for longer horizons, of a trajectory. The model is not fundamentally about "will they buy?" It is about "what state will this actor enter next, and what future traces follow, at the resolution relevant to the task?"

It is useful to decompose the learned transition into four operations.

First, encode the current individual actor-state estimate:

\[
\mathbf h_{i,t}^{(s)}
=
E_{s,\theta}(\widehat s_{i,t}).
\]

Second, encode the external or symbolic proposition without yet pretending that this encoding is how the actor experiences it:

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

Third, produce the actor-relative proposition representation for the individual-state model:

\[
\mathbf p_{i,t}^{(s)}(x_t)
=
\mathcal P_{s,\theta}
\!\left(
\mathbf e_t^{x},
 \widehat s_{i,t}
\right).
\]

Fourth, form the individual interaction representation:

\[
\mathbf h_{i,t}^{(s,\mathrm{int})}(x_t)
=
\Psi_{s,\theta}
\!\left(
\mathbf h_{i,t}^{(s)},
\mathbf p_{i,t}^{(s)}(x_t)
\right).
\]

The state transition can then be written

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

The individual actor-relative map \(\mathcal P_{s,\theta}\) need not be linear. It may contain projection, attention, retrieval, and nonlinear transformation. It stands for the fact that the same proposition is not available to every actor in the same way. The proposition is first represented, then reorganized through the distinctions, salience, language, role, relationship, and current state of the actor who receives it.

This factorization asserts that the fitted transition depends on the proposition through the joint interaction representation:

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

That is a model restriction, not a metaphysical theorem. It is what makes equality of actor-relative interaction representations imply equality of the fitted model's predictions.

## God's Infinite Dimensional Space: Making All Realities Composable

"To make all realities composable" does not mean every object is one interchangeable vector or that every category receives its own permanent part of the Hilbert space.

GIDS is the common arena of possible distinctions. The things represented through it are constructions.

A person is represented through an inherited and developed organization of those distinctions. A proposition is represented through the features it presents to an actor. A category may be a region, prototype, distribution, or classifier over those coordinates. A corporation is assembled from people, internal structure, memory, facts, statistics, incentives, and environment. None of these needs to be a primitive type inside the space.

The composability claim is therefore about a common formal grammar:

1. encode relevant distinctions;
2. project or gate them through an actor's available structure;
3. combine them with the current state;
4. apply a transition law;
5. read out observable consequences;
6. update from what actually happened.

Some operations will be linear. Most interesting interactions will not be.

A dot product between two vectors is meaningful only when the encoders and training objective have made it meaningful. Adding two representations is legitimate only when the result has a defined interpretation. The fact that two objects occupy the same Hilbert arena gives us the capacity to relate them; it does not grant every possible piece of vector arithmetic philosophical significance.

If two external propositions differ physically but produce the same task-relevant joint interaction representation, then they are equivalent for that actor and task **under the fitted model**. Formally, if

\[
\mathbf h_{i,t}^{(s,\mathrm{int})}(x_t^{(1)})
=
\mathbf h_{i,t}^{(s,\mathrm{int})}(x_t^{(2)}),
\]

then, holding the operational actor-state and exogenous input fixed,

\[
K_{s,\theta}
\!\left(
\cdot
\mid
\widehat s_{i,t},X_t=x_t^{(1)},\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right)
=
K_{s,\theta}
\!\left(
\cdot
\mid
\widehat s_{i,t},X_t=x_t^{(2)},\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right).
\]

This is a statement about the fitted model because the factorization above makes the actor-relative proposition representation the only proposition-side input to the transition kernel. It is not a metaphysical theorem that two physically distinct propositions must have the same real-world effect.

This captures a number of cases at once. A distinction associated with dark matter may be absent from unmediated human perception even though humans can infer dark matter through instruments, theory, and its observable effects. A table made of wood and a table made of stone may differ physically, yet remain interchangeable for a task in which material never enters the observer's next state. A price stated as a percentage and the same price stated as a dollar amount may be economically equivalent while producing different human representations because framing activates different coordinates.

The world is therefore not composable because everything is the same. It is composable because differences can be represented, transformed, ignored, or made consequential under one formal program.

### Composite actors

A corporation requires a separate construction.

Let \(C\) be a corporation and let \(\mathcal J_C(t)\) be the set of relevant people participating in it at time \(t\). For each member \(j\), define the abstract person-state available to the conceptual construction by

\[
\varsigma_{j,t}^{\mathrm{person}}
:=
\bigl(T_{j,t},\phi_{j,t},c_{j,t}\bigr).
\]

Let

- \(\mathsf{Facts}_{C,t}\) denote facts and statistics about the corporation;
- \(\mathsf{Org}_{C,t}\) denote role, authority, governance, and communication structure;
- \(\mathsf{Mem}_{C,t}\) denote institutional memory;
- \(\mathsf{Inc}_{C,t}\) denote incentives, constraints, and active objectives;
- \(\mathsf{EnvHist}_{C,t}\) denote historical environmental exposure already absorbed into the corporation's organization.

Define the conceptual composite state by

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

\(\mathcal A_{\mathrm{corp}}\) is not an average. It is an aggregation mechanism that should respect authority, information flow, coalition structure, veto rights, incentives, and the fact that most employees do not contribute equally to every corporate action.

Once this amalgamation has persistence, memory, a boundary, channels through which propositions enter, and a repeatable way of producing outward responses, it acquires a characteristic—almost feel-like—way of dealing with the world. For a corporation, "experience" means the state made institutionally available through its people, records, systems, incentives, and communication paths; it does not automatically mean phenomenal consciousness. The construction only needs to justify treating the corporation as an actor at the level of prediction.

The model never observes the complete conceptual state. Let \(\mathcal J_C^{\mathrm{obs}}(t)\subseteq\mathcal J_C(t)\) be the usable member subset, let \(\mathbf u_{C,t}^{\mathrm{corp}}\) collect measurable company-side evidence, and let \(\mathbf m_{C,t}^{\mathrm{miss}}\) encode missing membership, authority, and company fields. The learned estimate is

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

Let \(\boldsymbol\Xi_{C,t+1}\) denote the random exogenous corporate change before the next decision epoch and \(\boldsymbol\xi_{C,t+1}\) a realized or supplied scenario value. Its fitted transition law can then be written in the same outer form:

\[
\widetilde S_{C,t+1}
\sim
K_{\mathrm{corp},\theta}\!\left(
\cdot
\mid
\widehat S_{C,t},X_t=x_t,\mathbf w_t,\boldsymbol\Xi_{C,t+1}=\boldsymbol\xi_{C,t+1}
\right).
\]

The outer algebra is shared. The internal construction is not.

For the present paper, however, humans remain the first target. Corporations enter the first application as structured context around the people acting through them. Later, the same data can be reorganized so that the corporation itself becomes the center of prediction.

## Creating the World Model

A world model, in the present sense, is not a copy of physical reality. It is a learned simulator of actor, company, and relationship transitions under propositions.

The first application domain is go-to-market interaction because it produces repeated transitions, clear timestamps, and measurable outcomes.

Consider the actual problem.

Salesperson A, working at Company A, is trying to sell a product to Executive B, working at Company B. To keep the formal labels from eating each other, I will write the salesperson as \(a\), the executive as \(b\), the salesperson's company as \(C_a\), and the executive's company as \(C_b\).

The system must rank a sequence of propositions using the history of both people, the features of both companies, the relationship between them, and the environment in which the transaction is unfolding.

For each human \(i\in\{a,b\}\), define the estimated person-side substate by

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

The current world vector will appear once at the dyadic level rather than being copied into each person tuple.

Let

\[
\widehat S_{C_a,t},
\qquad
\widehat S_{C_b,t}
\]

be the relevant company-state representations, and let

\[
\widehat\Gamma_{ab,t}
\]

denote the estimated relationship state between the two sides: prior contact, familiarity, trust, perceived authority, commitments, objections, response history, sender reputation, channel history, and prior commercial interaction.

Define the filtered operational dyadic state by

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

This is the state type the first world model should operate on. The hats matter: this is the estimate available to the system before proposition \(x_t\), not the unknowable complete state of either person or corporation.

When a focal person also contributes to the learned company summary, the dyadic tuple contains a deliberate redundancy. An implementation should either exclude that focal person from the company-context summary, or estimate the person and company states jointly so the duplicate path is not treated as independent evidence.

The proposition is also structured. Write

\[
x_t
=
\left(
\text{content},
\text{offer},
\text{price},
\text{framing},
\text{evidence},
\text{channel},
\text{timing},
\text{sender},
\text{requested action}
\right)_t.
\]

Let \(\boldsymbol\Xi_{t+1}\) denote random exogenous changes between decision epochs and \(\boldsymbol\xi_{t+1}\) a realized or supplied scenario value: market movement, personnel change, company change, or anything else that is not generated by the dyadic action alone. Expressions of the form \(K(\cdot\mid\boldsymbol\Xi=\boldsymbol\xi)\) mean kernel evaluation at a supplied scenario value, not conditioning on a necessarily positive-probability point event. The notation \(K_{D,\theta}\) below is shorthand for the typed encoder–interaction–transition factorization defined explicitly in Algorithm 2. The single-step transition is

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

The immediate observable trace is

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}\!\left(
\cdot
\mid
\widetilde D_{ab,t+1}
\right).
\]

This readout makes the usual state-space assumption that the simulated next state is sufficient for the immediate trace. If that assumption is too strong for an implementation, use the more general kernel \(R_{O,\theta}(\cdot\mid\widetilde D_{ab,t+1},\widehat D_{ab,t},X_t=x_t)\) and test whether the extra conditioning is needed.

The trace variable \(O_{t+1}\) summarizes the records attributed to the interval \((t,t+1]\); several timestamped records may contribute to one trace bundle. It may include reply, silence, objection, meeting acceptance, forwarding behavior, a change in deal stage, a change in sentiment, or an internal action at Company \(C_b\).

The output of the transition is the same state type needed by the next transition. This is what makes the model recursively usable rather than a one-step decoder that cannot be composed.

Before an outcome 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 actual records arrive, the model **filters** or updates its estimate:

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

where \(\mathcal H_{(t,t+1]}\) is the collection of response, company, relationship, and environmental records that became available between the two decision epochs. Actual exogenous changes are included in this record set rather than passed to the filter a second time.

Simulation and filtering are different operations. The first asks what might happen. The second changes what we believe after something actually happened.

For a sequence of propositions \(x_t,\ldots,x_{t+H-1}\), and an exogenous scenario path \(\boldsymbol\xi_{t+1:t+H}\), repeated application of the transition and readout kernels defines the trajectory law

\[
\mathbb P_\theta
\!\left(
\widetilde D_{ab,t+1:t+H},\widetilde O_{t+1:t+H}
\mid
\widehat D_{ab,t},
\mathbf X_{t:t+H-1}=\mathbf x_{t:t+H-1},
\boldsymbol\Xi_{t+1:t+H}=\boldsymbol\xi_{t+1:t+H}
\right).
\]

Exogenous changes in the market, personnel, company condition, or world state must be supplied or modeled along the way. We do not get to pretend the future environment freezes merely because the recursion is convenient. When candidate propositions are compared under one ranking regime, they must be evaluated against the same declared exogenous scenario or the same action-invariant exogenous law. Otherwise a change in the assumed environment can be mistaken for an effect of the proposition. Any future variable that is itself caused by the proposition belongs inside the state transition, not inside an exogenous law that is held fixed across candidates.

For a task outcome measured at horizon \(\Delta\), write the fitted predictive law as

\[
P_{Y,\theta,\tau,\Delta}
\!\left(
\cdot
\mid
\widehat D_{ab,t},X_t=x_t
\right).
\]

This law may be implemented by a direct head or induced by rolling the state model forward and applying an outcome functional to the resulting trajectory. For a horizon extending beyond the next decision epoch, it is defined only relative to a declared continuation policy, future candidate-set process, exogenous-path law, and censoring convention unless those quantities are conditioned on explicitly. A direct head must be calibrated to the same regime. The notation does not pretend that a 90-day close is literally an immediate emission one step after an email.

Nothing in this framework commits us to one architecture. The transition kernel may be implemented by a linear state-space model, a recurrent model, an attention-based model, a structured probabilistic model, or a neural system that combines several of these. Sequence models in the Mamba family are one candidate because they compress long histories into an evolving state, but they do not define the theory. They are implementation options inside it.

What matters most at the outset is not architectural ambition but a working procedure.

### Algorithm 1: Estimate the actor and relationship state

Choose a task \(\tau\) and a horizon \(\Delta\).

Estimate the slow and fast person-side states for salesperson \(a\) and executive \(b\). Construct company representations from available organizational facts, statistics, roles, incentives, member states, and histories. Construct relationship state from all pre-proposition interactions.

This produces

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

### Algorithm 2: Encode the proposition and predict the transition

Encode the structured proposition:

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

Compute its actor-relative representation for executive \(b\) inside the full dyadic context:

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

Encode the current dyadic state and form the interaction:

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

Then predict the next state distribution:

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

The shorthand

\[
K_{D,\theta}\!\left(
\cdot\mid\widehat D_{ab,t},X_t=x_t,\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right)
:=
K_{D,\theta}^{\mathrm{int}}\!\left(
\cdot\mid\mathbf h_{ab,t}^{(D,\mathrm{int})}(x_t),\boldsymbol\xi_{t+1}
\right)
\]

will be used whenever the encoder–interaction factorization is not the point.

Decode immediate traces from the next state:

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}
\!\left(
\cdot\mid\widetilde D_{ab,t+1}
\right).
\]

For delayed primary outcomes and auxiliary probes that must be used jointly, define a coherent regime-specific law

\[
(\widetilde{\mathbf Y}_t,\widetilde{\mathbf Z}_t)
\sim
P_{YZ,\theta,\mathfrak e}
\!\left(
\cdot\mid\widehat D_{ab,t},X_t=x_t
\right).
\]

The horizon-specific primary and probe heads

\[
P_{Y,\theta,\tau,\Delta},
\qquad
P_{Z,\theta,m,\Delta_m}
\]

are marginals or conditionals of this joint law when a joint law is modeled. Separate marginal heads are adequate for separate losses and marginal expectations; they do not define cross-head dependence unless a coupling or conditional-independence assumption is declared.

### Algorithm 3: Update from error

Observe the matured primary outcomes and probe labels. Minimize

\[
\mathscr J(\theta)
=
\sum_{\ell=1}^{L_Y}
\lambda_{\ell}^{Y}
\mathscr J_{\ell}^{Y}(\theta)
+
\sum_{m=1}^{M_Z}
\lambda_m^{Z}
\mathscr J_m^{Z}(\theta)
+
\lambda_{\mathrm{reg}}\Omega(\theta),
\]

with

\[
\lambda_{\ell}^{Y},
\lambda_m^{Z},
\lambda_{\mathrm{reg}}\ge 0,
\qquad
\Omega(\theta)\ge 0.
\]

Masks, censoring weights, or survival likelihood contributions belong inside the relevant head loss. The positive signs matter. These are losses and penalties being minimized. Writing negative signs would reward the model for making them larger.

Update the parameters by gradient descent,

\[
\theta_{k+1}
=
\theta_k
-
\alpha_{\mathrm{opt}}\nabla_\theta\mathscr J(\theta_k),
\]

or update a posterior distribution over parameters in a Bayesian implementation.

Then update the fast person states and relationship state using the actual observed records. Refresh slow person and company states only when durable evidence accumulates.

### Algorithm 4: Detect drift and reopen discovery

No implementation should be assumed stable forever. If the market changes, the product changes, the organization changes, incentives shift, or a once-inert distinction becomes active, model quality will decay.

Compare recent loss and calibration with reference windows. When degradation persists, reopen the discovery process: add candidate features, revise the company aggregation, reweight existing coordinates, expand the proposition representation, or rebuild the task projection.

This is the correct sense in which the framework is open-ended. The framework is not rescued from every failure by saying the first implementation was weak. Specific implementations are falsified when they fail. The broader program survives only if it keeps generating better testable representations rather than excuses.

## From Forecasting to Proposition Search

This is the place where I stop pretending the point of the machinery is merely to admire prediction metrics.

The practical purpose of the framework is not only to forecast outcomes, but to compare admissible candidate propositions by their expected effect on the next state and downstream objective. Otherwise why the hell are we building it.

Let \(\mathcal X_t^{\mathrm{adm}}\) denote the admissible proposition universe and let \(\varnothing\ne\mathcal X_t^{\mathrm{cand}}\subseteq\mathcal X_t^{\mathrm{adm}}\) be the set actually available for scoring at decision \(t\). Let \(U_\tau\) be a declared measurable utility over predicted trajectories, outcomes, probe variables, cost, and policy constraints, and assume it is integrable under every model and evaluation regime being compared.

A value over more than one future decision is undefined until the evaluation regime is declared. Let \(\mathfrak e\) contain at least:

- a continuation policy \(\pi^{\mathrm{cont}}\) after the candidate proposition;
- the process that supplies future candidate sets;
- an exogenous-path law \(\mathbb Q_{\Xi}\);
- and the outcome, censoring, and terminal-utility convention used by the score.

For model-based predictive ranking, define

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

The tilded outcome and probe bundles are either measurable functionals of the same simulated trajectory or draws from the coherent joint kernel \(P_{YZ,\theta,\mathfrak e}\). If only marginal heads exist, the utility uses marginal expectations or an explicitly declared coupling. They are not independent duplicate futures. Candidate ranking chooses

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

This argmax exists automatically for a nonempty finite candidate set. For a continuous candidate space, existence requires conditions such as compactness and upper semicontinuity; otherwise the mathematically correct target is a supremum or an approximate optimizer. In deployment the score should also be constrained by support, uncertainty, policy rules, and the cost of simulator exploitation; the highest unconstrained point estimate should not automatically win.

But the actual sales problem concerns a sequence of propositions. The best first move may not be the message with the highest immediate meeting probability. It may be the message that reveals uncertainty, establishes legitimacy, changes perceived risk, or makes a later proposition more effective.

Let the policy be

\[
\pi_t
\!\left(
 x
\mid
\widehat D_{ab,t},
\mathcal X_t^{\mathrm{cand}}
\right).
\]

Let \(\mathfrak P_{\mathrm{adm}}\) be the nonempty admissible policy class, let \(\gamma\in[0,1]\), and let \(\mathbb Q_{\Xi}\) be the declared exogenous-path law and \(\mathbb Q_{\mathcal X}\) the future candidate-set law. For a planning length \(H\), define a step utility \(u_\tau^{\mathrm{step}}\) and terminal value \(V_\tau^{\mathrm{term}}\). Then

\[
J_{\theta,\mathbb Q_{\Xi},\mathbb Q_{\mathcal X}}(\pi\mid\widehat D_{ab,t})
:=
\mathbb E_{\theta,\pi,\mathbb Q_{\Xi},\mathbb Q_{\mathcal X}}
\!\left[
\sum_{k=0}^{H-1}
\gamma^k
u_\tau^{\mathrm{step}}
\!\left(
\widetilde D_{ab,t+k+1},
X_{t+k},
\widetilde O_{t+k+1}
\right)
+
\gamma^H V_\tau^{\mathrm{term}}(\widetilde D_{ab,t+H})
\mid
\widehat D_{ab,t}
\right],
\]

where the expectation also covers the future candidate-set process declared by the planning regime. The policy-search problem is

\[
\pi^\star
\in
\arg\max_{\pi\in\mathfrak P_{\mathrm{adm}}}
J_{\theta,\mathbb Q_{\Xi},\mathbb Q_{\mathcal X}}(\pi\mid\widehat D_{ab,t})
\]

when a maximizer exists; otherwise write \(\sup_{\pi\in\mathfrak P_{\mathrm{adm}}}\). This avoids the common stupidity of copying one eventual close backward and awarding it independently to every prior message. Credit assignment belongs to the trajectory.

The first implementation does not need to solve unconstrained long-horizon persuasion. A reasonable progression is:

1. one-step ranking among controlled proposition families;
2. ranking short predefined sequences;
3. adaptive next-best-action after each observed response;
4. longer-horizon policy optimization after the transition model and intervention design are trustworthy.

This gives us three distinct epistemic regimes.

First, there is **forecasting**: estimate what is likely to happen after the proposition that was actually delivered.

Second, there is **model-based ranking**: use the forecasting model to simulate and order candidate propositions. This is useful, but it is still only as good as the model and the support of the data.

Third, there is **interventional policy improvement**: choose propositions and claim that choosing them causes better outcomes. This requires an experimental or otherwise defensible causal identification design. When logged data are used, it also requires the assignment probabilities or densities recorded at decision time, sufficient overlap, consistency, appropriate handling of interference and censoring, and an estimator matched to the one-step or sequential regime.

The predictive conditional quantity is

\[
\mathbb P
\!\left(
Y_t^{(\tau,\Delta)}\in B
\mid
\widehat D_{ab,t},
X_t=x
\right).
\]

The causal quantity is

\[
\mathbb P
\!\left(
Y_t^{(\tau,\Delta)}\in B
\mid
\widehat D_{ab,t},
\operatorname{do}(X_t=x)
\right),
\]

or equivalently the potential-outcome law

\[
\mathbb P
\!\left(
Y_t^{(\tau,\Delta)}(x;\mathfrak e)\in B
\mid
\widehat D_{ab,t}
\right),
\]

where \(\mathfrak e\) fixes the downstream continuation, candidate-set, exogenous, and censoring regime after the current proposition. The corresponding causal value under that same declared regime is

\[
V_{\mathfrak e}^{\mathrm{causal}}(x\mid\widehat D_{ab,t})
=
\mathbb E_{\mathfrak e}
\!\left[
U_\tau
\mid
\widehat D_{ab,t},
\operatorname{do}(X_t=x)
\right].
\]

These are not interchangeable without an identification argument. The causal ambition is to choose propositions that change behavior. The engine underneath that ambition is forecasting. The causal claim begins only when the data collection process and assumptions let the forecasting architecture estimate an interventional law.

Until then, leave the causal swagger out of it. The system may still be useful. It is useful as a forecasting, simulation, and ranking device rather than as a proven controller.

## Closing Part 3

Part 1 argued that reality, as it appears to an organism, is not a mirror of noumena but the output of an evolved representational structure. Part 2 showed how that structure can be individualized, historically deformed, estimated from observable traces, and split into slow and fast state. Part 3 completes the descent into an operational world model.

The ideal object remains the next phenomenal state. The practical object is a predictive actor-state. GIDS supplies coordinates of possible distinction; it does not contain a warehouse of primitive people, propositions, and corporations. Those objects are constructed from patterns and relations in the space. Their interaction is modeled as a state transition. The resulting state can be simulated, compared against outcomes, decoded into probes, and updated under error.

For the first domain, the whole ambition can be written as

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

\[
(\widehat D_{ab,t},X_t=x_t)
\longrightarrow
\left(
\widetilde{\mathbf Y}_t,
\widetilde{\mathbf Z}_t
\right),
\]

\[
(\widehat D_{ab,t},x_t,\mathcal H_{(t,t+1]})
\longrightarrow
\widehat D_{ab,t+1}.
\]

The three lines are simulation, delayed prediction, and filtering. They are related; they are not the same operation.

And if proposition search is turned on,

\[
\widehat D_{ab,t}
\longrightarrow
V_{\theta,\mathfrak e}^{\mathrm{pred}}(x\mid\widehat D_{ab,t})
\longrightarrow
x_t^\star
\longrightarrow
\text{experiment and update}.
\]

That is the whole ambition of this section. Not a complete algebra of mind, but a way to build one without lying about what has and has not been solved.

OK, it is time to get serious now.


---

# Part 4: Benchmarking the World Model

Part 3 supplied a framework. Part 4 makes it answerable to some sort of data.

We've argued so far that an observer can be represented through an estimated Transcendental Embedding, that this estimate can be split into slow and fast state, that propositions can be represented through the distinctions they present to the observer, and that corporations can enter either as structured context or, later, as composite actors. But any framework that does not specify what counts as state, what data instantiates that state, what task is being predicted, what stronger baselines it must beat, and what evidence would justify proposition optimization is nascent and not really worth anyone's time.

This section deals with benchmarking.

The purpose of Part 4 is not to prove the full metaphysical claim directly. It is to ask a narrower question: if we represent a salesperson–executive interaction as a dyadic state containing slow person structure, fast local state, two company contexts, relationship history, world state, and a proposition, do we predict observable transitions better than simpler models? And if we later use that model to choose propositions, do we have the intervention machinery to say something more than "the simulator liked this one"?

The canonical symbol definitions are in [Canonical Notation and Mathematical Conventions](08_Canonical_Notation.md).

## 4.1 Operational Definition of State

Philosophically, the state of an organism is total. It includes perception, interoception, memory, action tendency, and the actions already underway. But benchmarks do not get to be mystical.

For executive \(b\), let

\[
\phi_{b,t}
\]

denote the full phenomenal state. It is not directly observed.

Let

\[
Q_{b,t}
\]

denote the general predictive response state. When a task-and-horizon summary satisfying the required sufficiency condition exists, write it as

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

This \(q_{b,t}^{(\tau,\Delta)}\) is a current-time predictive summary. It is not the object recursively rolled through the simulator.

The benchmark operates on a measurable dyadic approximation. Let salesperson \(a\) work for company \(C_a\), and executive \(b\) work for company \(C_b\). Define

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

Let \(X_t\) be the random proposition selected at decision epoch \(t\), and let \(x_t\in\mathcal X_t^{\mathrm{cand}}\) be its realized value, where \(\varnothing\ne\mathcal X_t^{\mathrm{cand}}\subseteq\mathcal X_t^{\mathrm{adm}}\). The proposition remains separate because the benchmark asks what happens when a particular proposition encounters the current dyadic state.

Let \(\boldsymbol\Xi_{t+1}\) denote the random exogenous change between decision epochs and \(\boldsymbol\xi_{t+1}\) a realized or supplied scenario value. Expressions of the form \(K(\cdot\mid\boldsymbol\Xi=\boldsymbol\xi)\) mean evaluation of a parameterized kernel at the supplied scenario value, not conditioning on a necessarily positive-probability singleton. The notation \(K_{D,\theta}\) below is shorthand for the typed encoder–interaction–transition factorization specified in Section 4.4. The measurable one-step simulation is

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

and the immediate trace-bundle law for records attributed to the interval \((t,t+1]\) is

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}\!\left(
\cdot
\mid
\widetilde D_{ab,t+1}
\right).
\]

This readout assumes the next operational state screens off the previous state and proposition for the immediate trace. If an implementation rejects that Markov-style emission assumption, use \(R_{O,\theta}(\cdot\mid\widetilde D_{ab,t+1},\widehat D_{ab,t},X_t=x_t)\) and evaluate the added conditioning empirically.

For delayed decision-associated outcomes and probes used together, write the coherent regime-specific law as

\[
(\widetilde{\mathbf Y}_t,\widetilde{\mathbf Z}_t)
\sim
P_{YZ,\theta,\mathfrak e}\!\left(
\cdot
\mid
\widehat D_{ab,t},
X_t=x_t
\right).
\]

The horizon-specific kernels \(P_{Y,\theta,\tau,\Delta}\) and \(P_{Z,\theta,m,\Delta_m}\) are its marginals or conditionals when a joint law is modeled. A direct horizon-specific law is defined only relative to a declared continuation policy, future candidate-set process, exogenous-path regime, and censoring convention unless those quantities are conditioned on explicitly. It may instead be induced by a rollout. Independently trained marginal heads do not define cross-head dependence; a utility combining them must use marginal expectations, a declared coupling, or a joint model. A direct head and a rollout used in the same score must refer to the same evaluation regime rather than generating duplicate futures. Either way, a 90-day close is not represented as an immediate event emitted at \(t+1\).

The hierarchy is therefore explicit:

\[
\phi_{b,t}
\quad\text{full motivating state},
\]

\[
Q_{b,t}
\quad\text{general predictive response state},
\]

\[
q_{b,t}^{(\tau,\Delta)}
\quad\text{task-conditioned sufficient summary, when it exists},
\]

\[
\widehat D_{ab,t}
\quad\text{measurable filtered dyadic state}.
\]

The benchmark only has access to the last object.

## 4.2 Event Time and Dataset Construction

The clock has to be clean or the entire benchmark becomes leakage with equations around it.

Let \(t\) index decision epochs. Define the pre-decision history by

\[
\mathcal H_{<t}
=
\left(
\mathsf h_n:
\operatorname{time}(\mathsf h_n)
<
\operatorname{time}(\mathsf d_t)
\right),
\]

where each \(\mathsf h_n\) is a timestamped record that was fully available before proposition \(x_t\) was selected or delivered. Write \(v_n:=\operatorname{time}(\mathsf h_n)\) and define

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

Records arrive asynchronously, so \(N(t)\) need not equal \(t-1\). The decision-aligned latent states are the record-indexed states after exactly these \(N(t)\) records have been processed. Timestamp ties require an explicit logging order that preserves the actual decision-before-response sequence.

At decision time \(t\):

1. construct \(\widehat D_{ab,t}\) from \(\mathcal H_{<t}\) and all other information available at that moment;
2. construct or retrieve the logged candidate set \(\mathcal X_t^{\mathrm{cand}}\);
3. select and deliver \(x_t\in\mathcal X_t^{\mathrm{cand}}\);
4. record the decision and make the forecast before observing any response;
5. append response, meeting, company-change, and other observation records only when they actually occur;
6. attach delayed labels to the original decision row only after their horizons mature.

This avoids the old ambiguity in which a history through time \(t\) could already contain the response to the proposition being predicted. It also avoids pretending a 90-day outcome is known before the next sales touch.

Use two timestamped record types. A decision record is

\[
\mathsf d_t
=
\left(
\operatorname{id}_t,
\mathcal X_t^{\mathrm{cand}},
 x_t,
\delta_t,
\mathbf e_t^{\mathrm{cat}},
\eta_t,
\mathsf I_t^{\mu}
\right),
\]

where \(\delta_t\) is time since the prior decision or contact, \(\mathbf e_t^{\mathrm{cat}}\) is the source-tagged categorical representation available from the decision itself, and \(\mathsf I_t^{\mu}\) is the information actually available to the behavior policy. The logged assignment probability for a discrete proposition, or assignment density for a continuous proposition, is

\[
\eta_t
:=
\mu_t\!\left(
 x_t
\mid
\mathsf I_t^{\mu},
\mathcal X_t^{\mathrm{cand}}
\right).
\]

This distinction matters. The denominator in off-policy evaluation is the probability the historical policy actually assigned using the information it actually had. It is not a probability retroactively conditioned on a richer state reconstructed later.

An observation record arriving at clock time \(v\) is

\[
\mathsf o_v
=
\left(
\mathbf o_v^{\mathrm{obs}},
\mathbf m_v^{\mathrm{obs}},
\Delta\mathbf f_{C_a,v}^{\mathrm{corp}},
\Delta\mathbf f_{C_b,v}^{\mathrm{corp}},
\Delta\mathbf w_v
\right).
\]

Here \(\mathbf o_v^{\mathrm{obs}}\) is the observed interaction trace, \(\mathbf m_v^{\mathrm{obs}}\) is an observable memory proxy such as a resurfaced objection, and the remaining terms are newly observed changes in company or world state. The chronological history \(\mathcal H_{<t}\) contains every decision and observation record whose timestamp precedes decision \(t\).

### Decision-associated labels

A single decision can carry several outcomes at different horizons. Define the primary head index set

\[
\mathcal J_Y
=
\{(\tau_\ell,\Delta_\ell):\ell=1,\ldots,L_Y\}.
\]

The primary outcome bundle attached to decision \(t\) is

\[
\mathbf Y_t
=
\left(
Y_t^{(\tau_\ell,\Delta_\ell)}
\right)_{\ell=1}^{L_Y}.
\]

For censoring and label maturity, define

\[
R_{t,\ell}^{\mathrm{obs}}
=
\mathbf 1\{Y_t^{(\tau_\ell,\Delta_\ell)}
\text{ is observed by the analysis cutoff}\}.
\]

A label with \(R_{t,\ell}^{\mathrm{obs}}=0\) is unobserved, missing, or censored. It is not a negative. If censoring depends on variables related to the outcome, use a survival likelihood, inverse-probability-of-censoring weights, or a joint model rather than complete-case optimism.

Write the primary availability bundle as

\[
\mathbf R_t^{\mathrm{obs}}
=
\left(
R_{t,\ell}^{\mathrm{obs}}
\right)_{\ell=1}^{L_Y}.
\]

Define the auxiliary probe bundle as

\[
\mathbf Z_t
=
\left(
Z_t^{(m,\Delta_m)}
\right)_{m=1}^{M_Z},
\]

with its own availability masks when probes mature at different times. Write

\[
R_{t,m}^{Z,\mathrm{obs}}
=
\mathbf 1\{Z_t^{(m,\Delta_m)}\text{ is observed by the analysis cutoff}\},
\qquad
\mathbf R_t^{Z,\mathrm{obs}}
=
\left(R_{t,m}^{Z,\mathrm{obs}}\right)_{m=1}^{M_Z}.
\]

### Event-time dataset

For the first benchmark, define

\[
\mathcal D
=
\left\{
\left(
\mathbf u_{a,t},
\mathbf u_{b,t},
\mathbf u_{C_a,t}^{\mathrm{corp}},
\mathbf u_{C_b,t}^{\mathrm{corp}},
\mathcal H_{<t},
\mathcal H_{ab,<t}^{\mathrm{rel}},
\mathbf w_t,
\mathcal X_t^{\mathrm{cand}},
 x_t,
\mathbf Y_t,
\mathbf Z_t,
\mathbf R_t^{\mathrm{obs}},
\mathbf R_t^{Z,\mathrm{obs}},
\eta_t,
\mathsf I_t^{\mu}
\right)
\right\}_{t=1}^{N}.
\]

Here:

- \(\mathbf u_{a,t}\) and \(\mathbf u_{b,t}\) are person-side inputs available before \(x_t\);
- \(\mathbf u_{C_a,t}^{\mathrm{corp}}\) and \(\mathbf u_{C_b,t}^{\mathrm{corp}}\) are measurable company-side input bundles;
- \(\mathcal H_{ab,<t}^{\mathrm{rel}}\) is the pre-proposition relationship-event history;
- \(\mathbf w_t\) is market and world state;
- \(\mathcal X_t^{\mathrm{cand}}\) is the candidate set actually available at the decision;
- \(\mathbf Y_t\) is the primary multi-horizon outcome bundle;
- \(\mathbf Z_t\) is the auxiliary probe bundle;
- \(\mathbf R_t^{\mathrm{obs}}\) and \(\mathbf R_t^{Z,\mathrm{obs}}\) contain primary-label and probe availability indicators;
- \(\eta_t\) is the logged probability or density assigned to the realized action;
- \(\mathsf I_t^{\mu}\) records what the behavior policy knew when it made the decision.

Every feature in a row must be available before the proposition is chosen. Later CRM fields, transcript summaries produced from the response, post-meeting notes, eventual stage changes, and other downstream information cannot leak backward into \(\widehat D_{ab,t}\).

### Person-side inputs

For each person \(i\in\{a,b\}\), let

\[
\mathbf u_{i,t}
=
\left[
\boldsymbol\psi_i,
\mathbf b_i,
\mathbf u_{i,t}^{\mathrm{lang}},
\mathbf v_{i,t}^{\mathrm{role}},
\overline{\mathbf h}_{i,t},
\mathbf g_{i,t}^{\mathrm{slow}}
\right],
\]

where the components are psychometric or cognitive proxies, biography, language and discourse, role and institution history, observable life or professional history, and slow categorical trace memory.

If a coordinate cannot be inferred cleanly from real data, mask it. Do not invent variables because they sound sophisticated.

### Company-side inputs

For each company \(C\in\{C_a,C_b\}\), let

\[
\mathbf u_{C,t}^{\mathrm{corp}}
=
\left[
\mathbf f_{C,t}^{\mathrm{corp}},
\mathbf g_{C,t}^{\mathrm{org}},
\mathbf m_{C,t}^{\mathrm{inst}},
\boldsymbol\iota_{C,t}^{\mathrm{inc}},
\mathbf h_{C,t}^{\mathrm{env}}
\right],
\]

where the components contain measurable facts and statistics, authority and communication structure, institutional memory proxies, incentives and constraints, and environmental history already absorbed into the corporation. Examples include size, industry, growth, funding, revenue proxies, technology, hiring, leadership, governance, recent events, and account history.

These inputs are not the corporation-state by themselves. They are evidence supplied to the corporate aggregation mechanism.

### Relationship state

Let

\[
\mathbf v_{ab,t}^{\mathrm{rel}}
=
\left[
\text{touch count},
\text{reply history},
\text{meeting history},
\text{prior objections},
\text{commitments},
\text{trust proxies},
\text{sender familiarity},
\text{channel history},
\text{deal stage},
\text{time since last touch}
\right]_t.
\]

The relationship is not reducible to either person. It is an evolving object created by their prior interaction.

### Proposition representation

The proposition should be logged in structured form whenever possible:

\[
x_t
=
\left[
\text{text or content},
\text{offer family},
\text{price},
\text{framing},
\text{evidence},
\text{channel},
\text{timing},
\text{sender},
\text{requested action}
\right]_t.
\]

Free-form text can be embedded, but intervention and off-policy evaluation become far easier when the action space also contains controlled dimensions such as message family, proof type, call-to-action, offer, channel, and timing.

### Outcomes and probes

A first primary index set can be

\[
\mathcal J_Y
=
\left\{
(\mathrm{reply},7\mathrm d),
(\mathrm{meeting},21\mathrm d),
(\mathrm{stage},30\mathrm d),
(\mathrm{close},90\mathrm d)
\right\}.
\]

Thus

\[
\mathbf Y_t
=
\left[
Y_t^{(\mathrm{reply},7\mathrm d)},
Y_t^{(\mathrm{meeting},21\mathrm d)},
Y_t^{(\mathrm{stage},30\mathrm d)},
Y_t^{(\mathrm{close},90\mathrm d)}
\right].
\]

The auxiliary bundle may include

\[
\mathbf Z_t
=
\left[
\text{objection class},
\text{sentiment shift},
\text{urgency shift},
\text{next action type},
\text{reply delay bucket},
\text{forward or internal share}
\right]_t,
\]

with a declared horizon and availability mask for every component.

These probes are not meant to reveal the one true hidden motive of the prospect. They test whether the latent state carries reusable structure beyond one binary target.

One eventual close should not be copied backward and treated as though every preceding touch independently caused it. For one-step forecasting, each decision row receives only the outcomes defined for its own matured horizon. For proposition-sequence optimization, reward and credit assignment belong to the trajectory: use step rewards, sequence-level returns, time-to-event targets, or another explicit attribution rule rather than awarding the same terminal event to every message.

The first dataset should be built from CRM events, email logs, call transcripts, meeting records, sender metadata, account metadata, company descriptors, and known market state. If psychometric proxies or detailed corporate features are unavailable, run the benchmark without them first. The framework is supposed to discover what helps, not reward the imagination of the researcher.

## 4.3 The Benchmark

The benchmark is simple: does an explicit dyadic predictive-state model beat weaker baselines on future data, and does the explicit slow/fast and composite-company construction beat a generic sequence model that has enough capacity to absorb everything into one black box?

For each model, let

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
\left(
\widehat\vartheta_{t,1}^{Y},\ldots,\widehat\vartheta_{t,L_Y}^{Y}
\right)
\]

denote the predicted Bernoulli probabilities or, for non-Bernoulli heads, the corresponding declared distributional parameters for the \(L_Y\) primary heads. More formally, \(\widehat\vartheta_{t,\ell}^{Y}\in\Theta_\ell\), where \(\Theta_\ell=[0,1]\) for a Bernoulli head and may be a higher-dimensional parameter space for another outcome family. The same label masks and follow-up rules apply to every model.

The proposed system has to beat the following baselines. Let \(\mathcal T_{\mathrm{train}}\) denote the set of decision indices in the training window.

**Baseline 0: per-head empirical marginal**

For a Bernoulli head with at least one observed training label,

\[
\widehat\vartheta_{t,\ell}^{Y}
=
\frac{
\sum_{r\in\mathcal T_{\mathrm{train}}}
R_{r,\ell}^{\mathrm{obs}}Y_r^{(\tau_\ell,\Delta_\ell)}
}{
\sum_{r\in\mathcal T_{\mathrm{train}}}
R_{r,\ell}^{\mathrm{obs}}
},
\qquad \ell=1,\ldots,L_Y.
\]

For non-Bernoulli heads, use the corresponding empirical marginal distribution, mean, or survival baseline rather than forcing every task into a prevalence scalar.

**Baseline 1: current-proposition model**

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
h_1\!\left(
 x_t,
\mathbf w_t,
\mathbf v_{ab,t}^{\mathrm{rel}}
\right),
\]

using only the current proposition, current world state, and shallow relationship context.

**Baseline 2: static dyadic tabular model**

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
h_2\!\left(
\mathbf u_{a,t},
\mathbf u_{b,t},
\mathbf u_{C_a,t}^{\mathrm{corp}},
\mathbf u_{C_b,t}^{\mathrm{corp}},
\mathbf v_{ab,t}^{\mathrm{rel}},
\mathbf w_t,
 x_t
\right),
\]

with no explicit sequence state.

**Baseline 3: shallow-history model**

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
h_3\!\left(
\text{static inputs},
\operatorname{agg}(\mathcal H_{<t}),
 x_t
\right),
\]

where \(\operatorname{agg}(\mathcal H_{<t})\) contains hand-built summaries such as touch count, last-response delay, prior meeting count, reply rate, topic counts, and stage history.

**Baseline 4: recommender-style two-tower model**

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
h_4\!\left(
\widehat{\mathbf d}_t^{\mathrm{tower}},
\widehat{\mathbf x}_t^{\mathrm{tower}}
\right),
\]

where the dyad and proposition are embedded separately and scored through a dot product or shallow fusion, but no explicit recursive state is maintained.

**Baseline 5: monolithic sequence model**

\[
\widehat{\boldsymbol\vartheta}_t^{Y}
=
h_5\!\left(
\text{all pre-proposition inputs},
\mathcal H_{<t},
 x_t
\right),
\]

implemented by a generic recurrent, transformer, or state-space sequence model that sees the same event stream but does not enforce an explicit slow/fast or human/company/relationship decomposition.

This last baseline matters. If a monolithic sequence block with enough capacity eats my lunch, then the decomposition was just a story I told myself after the fact. If the explicit construction still wins or ties while transferring better, calibrating better, or requiring less data, then it has earned the right to stay.

## 4.4 The Proposed Latent-State Model

The proposed model constructs the state in stages.

First, estimate the slow person-side vectors with a shared human encoder:

\[
\widehat{\mathbf t}_{i,t}
=
E_{T,\theta}(\mathbf u_{i,t}),
\qquad i\in\{a,b\}.
\]

Role, seller/buyer position, and local context remain explicit inputs; they are not reasons to build two unrelated psychologies.

Second, maintain fast person-side states with a shared update family. For chronological record \(\mathsf h_n\) at time \(v_n\),

\[
\mathbf z_{i,n}
=
U_{z,\theta}\!\left(
\mathbf z_{i,n-1},
\widehat{\mathbf t}_{i}(v_n^-),
\mathbf c_i(v_n^-),
\mathbf w(v_n^-),
\mathsf h_n,
\mathbf d_{i,n}^{\mathrm{rec}}
\right),
\qquad i\in\{a,b\},\;n\ge1,
\]

where \(\mathbf d_{i,n}^{\mathrm{rec}}\) is the **record-applicability vector** for actor \(i\): it contains a binary applicability flag and records which actor-specific fields are available for the update. It is neither a memory vector nor a label-availability mask. When the applicability flag is zero, \(U_{z,\theta}\) is required to leave \(\mathbf z_{i,n-1}\) unchanged. A record may be a decision or a later observation; the update uses only what has become available by that point. At decision \(t\), set \(\mathbf z_{i,t}:=\mathbf z_{i,N(t)}\).

The seller state matters because the proposition actually delivered is partly a product of the seller, and the same nominal message can be presented differently by different people.

Define each person-side substate as

\[
\widehat s_{i,t}^{\mathrm{person}}
=
\left(
\widehat{\mathbf t}_{i,t},
\mathbf z_{i,t},
\mathbf c_{i,t}
\right),
\qquad i\in\{a,b\}.
\]

The shared world state \(\mathbf w_t\) stays outside these person-side tuples because it appears once in the dyadic state.

Third, construct company states:

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

Here \(\mathcal J_C^{\mathrm{obs}}(t)\) is the usable member subset. The first implementation may observe only a small subset of relevant people. Missing membership and authority information must be represented as missingness and uncertainty, not silently treated as zero. Historical environmental exposure is carried through company inputs and memory; the current shared world state remains explicit in the dyadic state.

Fourth, estimate relationship state:

\[
\widehat\Gamma_{ab,t}
=
E_{\Gamma,\theta}\!\left(
\mathcal H_{ab,<t}^{\mathrm{rel}},
\mathbf v_{ab,t}^{\mathrm{rel}}
\right).
\]

Here \(E_{\Gamma,\theta}\) is the batch/history realization of the relationship estimator. It may be implemented by initializing \(\widehat\Gamma_{ab,0}\) and applying the recurrent update \(U_{\Gamma,\theta}\) below to the same predecision relationship records. The batch and recurrent forms must therefore agree on record order, applicability, and the information available before the candidate proposition.

Then construct

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

Person context and company state can inform one another. If the focal salesperson or executive is included inside the corresponding company aggregator and also appears separately in \(\widehat D_{ab,t}\), the representation is redundant; exclude the focal person from the company-context summary or estimate the coupled states jointly so the duplicate path is not treated as independent evidence. An implementation may use a fixed number of message-passing updates, provided every input remains pre-proposition and no outcome information leaks backward.

Given candidate proposition \(x_t\), encode and contextualize it:

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

form the 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

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

The shorthand

\[
K_{D,\theta}\!\left(
\cdot
\mid
\widehat D_{ab,t},X_t=x_t,\boldsymbol\Xi_{t+1}=\boldsymbol\xi_{t+1}
\right)
:=
K_{D,\theta}^{\mathrm{int}}\!\left(
\cdot
\mid
\mathbf h_{ab,t}^{(D,\mathrm{int})}(x_t),\boldsymbol\xi_{t+1}
\right)
\]

will be used whenever the internal factorization is not the point.

Decode immediate traces from the next state:

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}\!\left(
\cdot\mid\widetilde D_{ab,t+1}
\right).
\]

Decode primary and auxiliary decision-associated outcomes through a rollout or a coherent joint direct head:

\[
(\widetilde{\mathbf Y}_t,\widetilde{\mathbf Z}_t)
\sim
P_{YZ,\theta,\mathfrak e}\!\left(
\cdot
\mid
\widehat D_{ab,t},X_t=x_t
\right).
\]

The individual kernels \(P_{Y,\theta,\tau_\ell,\Delta_\ell}\), \(\ell=1,\ldots,L_Y\), and \(P_{Z,\theta,m,\Delta_m}\), \(m=1,\ldots,M_Z\), are trained as head-wise marginals or conditionals. If the implementation trains only separate marginals, no cross-head independence claim follows.

The multi-head structure is there so the latent state does not remain a completely black box. If the state is real in the operational sense, it should carry reusable signal that helps decode more than one downstream observable.

The architecture of the encoders, aggregators, and transition kernel is not fixed by the theory. The first implementation may use gradient-boosted trees for static components, a GRU or state-space block for history, graph or set aggregation for company structure, and an attention or bilinear interaction for propositions. The theory requires explicit state and testable decomposition. It does not require blind loyalty to one named architecture.

## 4.5 Training Objective, Update Loop, and Intervention

For multiple primary outcomes and probes, minimize

\[
\mathscr J(\theta)
=
\sum_{\ell=1}^{L_Y}
\lambda_{\ell}^{Y}
\mathscr J_{\ell}^{Y}(\theta)
+
\sum_{m=1}^{M_Z}
\lambda_m^{Z}
\mathscr J_m^{Z}(\theta)
+
\lambda_{\mathrm{reg}}\Omega(\theta),
\]

with

\[
\lambda_{\ell}^{Y},
\lambda_m^{Z},
\lambda_{\mathrm{reg}}\ge0,
\qquad
\Omega(\theta)\ge0.
\]

All signs are positive because every term is a loss or penalty being minimized. Masks, censoring weights, or survival-likelihood terms are applied inside the corresponding head loss.



The model updates on different timescales.

Fast person states update after each relevant record according to

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

Relationship state updates on the same record clock. Let
\(d_{ab,n}^{\mathrm{rel}}\in\{0,1\}\) indicate whether record \(n\) pertains to the ordered dyad \((a,b)\); when it is zero, the update map is required to leave the relationship state unchanged:

\[
\widehat\Gamma_{ab,n}
=
U_{\Gamma,\theta}\!\left(
\widehat\Gamma_{ab,n-1},
\mathsf h_n,
\widehat s_{a}^{\mathrm{person}}(v_n^-),
\widehat s_{b}^{\mathrm{person}}(v_n^-),
d_{ab,n}^{\mathrm{rel}}
\right),
\]

\[
U_{\Gamma,\theta}(\gamma,h,s_a,s_b,0)=\gamma.
\]

At decision epoch \(t\), use \(\mathbf z_{i,t}=\mathbf z_{i,N(t)}\) and \(\widehat\Gamma_{ab,t}=\widehat\Gamma_{ab,N(t)}\).

The full filtered state update, after actual response and exogenous-change records have arrived, is

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

The slow person estimates update only when durable evidence accumulates. If \(\widehat{\mathbf t}_{i,t}^{\mathrm{new}}\) is a refreshed estimate expressed in the **same latent coordinate chart** as \(\widehat{\mathbf t}_{i,t}\), use

\[
\widehat{\mathbf t}_{i,t+1}
=
(1-\alpha_i)\widehat{\mathbf t}_{i,t}
+
\alpha_i\widehat{\mathbf t}_{i,t}^{\mathrm{new}},
\qquad
0<\alpha_i\le1,
\]

with \(\alpha_i\) typically chosen small. This arithmetic update is meaningful only if the encoder is fixed, the refreshed representation has been aligned to the old chart, or the relevant histories have been re-encoded into a common chart after an encoder change. Company states may contain both fast and slow components as well. A funding event or executive departure can move a company state quickly; culture and governance usually move more slowly.

In deployment, the loop is:

1. construct \(\widehat D_{ab,t}\) from pre-proposition information;
2. generate or retrieve \(\mathcal X_t^{\mathrm{cand}}\);
3. score candidates under the current forecasting model;
4. choose according to the current policy and exploration rule;
5. log the exact candidate set, chosen proposition, decision time, and assignment probability or density;
6. observe immediate response and delayed outcomes as they mature;
7. update fast and relationship state;
8. periodically refit parameters and refresh slow estimates.

### Predictive ranking

A score extending beyond the immediate next response requires a declared predictive evaluation regime \(\mathfrak e\). It contains the continuation policy, future candidate-set process, exogenous-path law, and outcome/censoring convention used by the score. Assume the declared utility is measurable and integrable under every fitted model and regime being compared. Define

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

The tilded outcome and probe bundles are either measurable functionals of the same rollout or draws from the coherent joint head calibrated to \(\mathfrak e\). If only marginals are fitted, the score uses marginal expectations or a declared coupling. This supports simulation and ranking. Useful, yes. Causal, not yet.

Because a learned simulator can be exploited by the optimizer, practical ranking should be conservative. One option is

\[
V_{\theta,\mathfrak e}^{\mathrm{cons}}(x\mid\widehat D_{ab,t})
=
V_{\theta,\mathfrak e}^{\mathrm{pred}}(x\mid\widehat D_{ab,t})
-
\lambda_{\mathrm{unc}}\,\operatorname{Unc}_\theta(x\mid\widehat D_{ab,t})
-
\lambda_{\mathrm{ood}}\,d_{\mathrm{support}}(x,\widehat D_{ab,t}),
\]

where \(\operatorname{Unc}_\theta\) measures predictive uncertainty and \(d_{\mathrm{support}}\) penalizes candidates far from the support of observed or randomized actions. The exact penalty is an implementation choice; the need to defend against model exploitation is not.

### Off-policy evaluation

Let

\[
\pi_t\!\left(
 x
\mid
\widehat D_{ab,t},
\mathcal X_t^{\mathrm{cand}}
\right)
\]

be the target policy. The logged behavior-policy probability for the realized action is the scalar

\[
\eta_t
=
\mu_t\!\left(
 x_t
\mid
\mathsf I_t^{\mu},
\mathcal X_t^{\mathrm{cand}}
\right).
\]

Let \(\mathcal T_{\mathrm{ope}}\) be the predeclared set of eligible decisions whose scalar utility \(u_t^{\mathrm{obs}}\) is mature and usable under the chosen censoring rule, and let \(N_{\mathrm{ope}}=|\mathcal T_{\mathrm{ope}}|>0\). If the utility extends beyond the immediate response, this one-step estimand changes the current proposition while holding the declared or logged continuation regime fixed. A basic inverse-propensity estimate is

\[
\widehat V_{\mathrm{IPS}}(\pi)
=
\frac{1}{N_{\mathrm{ope}}}
\sum_{t\in\mathcal T_{\mathrm{ope}}}
\frac{
\pi_t(x_t\mid\widehat D_{ab,t},\mathcal X_t^{\mathrm{cand}})
}{
\eta_t
}
u_t^{\mathrm{obs}}.
\]

The denominator is the logged probability or density assigned by the historical policy at the historical decision. It is not recomputed from \(\widehat D_{ab,t}\). The target-policy numerator must also be a function only of information available before the evaluated decision; a representation containing post-decision information would invalidate the ratio. For continuous actions, numerator and denominator must be densities with respect to the same dominating measure. Report the weight distribution and effective sample size. Clipping and self-normalization can reduce variance but introduce bias or change the finite-sample estimand, so report them as sensitivity analyses rather than invisible repairs. Doubly robust estimators are often preferable when the outcome and propensity nuisance models are credible. Informative censoring requires an additional censoring model or weighting term rather than complete-case deletion. A learned target policy should be frozen and evaluated on held-out data, or estimated with sample splitting or cross-fitting. Sequence policies require sequential off-policy estimators with products or per-decision products of importance ratios, or sequential doubly robust alternatives; the one-step expression is not applied blindly to an entire message path.

This formula only makes sense under an action representation with support. If every proposition is a unique free-form message, exact historical overlap is nearly nonexistent. The first intervention program should therefore randomize controlled proposition dimensions or families—framing, proof type, offer, call-to-action, channel, timing—rather than pretending every novel paragraph has a reliable counterfactual twin in the logs.

Off-policy or causal claims additionally require:

- **consistency:** the recorded treatment corresponds to the proposition definition used in the model;
- **overlap:** the behavior policy assigns nonzero probability to actions the target policy may choose;
- **identification:** randomization or a defensible sequential no-unmeasured-confounding argument based on the recorded decision information;
- **correct event ordering:** no post-treatment variable enters the state used to select the treatment;
- **handling of delayed outcomes:** later messages and events can mediate long-horizon labels;
- **interference assumptions:** several people inside \(C_b\) may influence one another, so executive \(b\) is not always an isolated unit;
- **stable treatment definition:** supposedly identical proposition families must be similar enough that the causal comparison is coherent;
- **censoring discipline:** outcome observation and attrition must be handled rather than assumed independent by convenience.

### Online policy improvement

If a controlled fraction of traffic can be randomized, proposition selection becomes a real policy-learning problem rather than retrospective ranking. At that point, the model can choose among admissible messages, offers, sequences, or timing policies, and value can be evaluated through live lift, regret, conversion, or long-run utility.

Until then, leave the causal swagger out of it.

## 4.6 Temporal Split, Evaluation, and Drift

The benchmark must be temporal. Random row splits can leak future information and can substantially overstate generalization in repeated-interaction data.

Use nonoverlapping rolling windows. Model, feature, threshold, and calibration choices are made with training and validation data only; the final test window remains untouched until the analysis is frozen:

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

Time alone is not enough. Report separate evaluation regimes:

1. future interactions with people and accounts seen during training;
2. new executives inside known companies;
3. known executives inside new selling contexts;
4. entirely unseen companies and people;
5. transfer to a new proposition family or outcome horizon.

This separates memorization from actor-state generalization.

Long-horizon outcomes require censoring discipline. A deal observed for only 40 days cannot be labeled "did not close within 90 days." Use complete follow-up windows or time-to-event and competing-risk methods where appropriate.

Rows are dependent within people, sellers, companies, campaigns, and time periods. Confidence intervals should use clustered or hierarchical resampling rather than pretending every event is independent.

For binary or probabilistic primary tasks, report

\[
\mathrm{LogLoss},
\qquad
\mathrm{Brier},
\qquad
\mathrm{PR\text{-}AUC},
\qquad
\mathrm{ECE}.
\]

Use survival metrics for censored time-to-event outcomes, ranking metrics for candidate-ordering tasks, and probe-appropriate metrics for auxiliary outputs.

Calibration matters because proposition ranking depends on differences between predicted values. A model that ranks adequately but is badly miscalibrated can still produce destructive utility estimates. ECE depends on the binning rule and should not be reported alone; include reliability diagrams and at least one complementary calibration summary or sensitivity analysis over binning choices.

### Ablations

The benchmark should force each major claim to either pay rent or be removed.

1. **Remove fast executive state \(\mathbf z_{b,t}\).** If short-horizon performance barely moves, the fast state is ornamental.
2. **Remove slow executive state \(\widehat{\mathbf t}_{b,t}\).** If cold-start and cross-context performance barely move, the durable embedding is ornamental.
3. **Remove salesperson state.** If nothing changes, seller-side psychology is not needed for this task once the proposition is fixed.
4. **Remove relationship state \(\widehat\Gamma_{ab,t}\).** If nothing changes, the interaction history is already captured elsewhere or was never useful.
5. **Replace company aggregation with flat company features.** If performance improves, the composite actor construction is premature.
6. **Replace salience-weighted categorical pooling with uniform averaging.** If nothing changes, the weighting story is decorative.
7. **Collapse source channels and role regimes before pooling.** If performance improves, the source-aware separation is unnecessary; if it hurts, the separation is buying signal.
8. **Shuffle recent within-dyad history while preserving static profiles.** If performance does not fall, the model was not using sequence in the way claimed.
9. **Remove probe heads.** If they contribute no stable transfer or regularization value, remove them.
10. **Replace the explicit architecture with the monolithic sequence baseline.** If the generic model dominates, the decomposition is not buying enough.
11. **Replace the shared GIDS interaction representation with late fusion of unrelated embeddings.** This directly tests whether mapping distinctions into a common interaction arena adds value.
12. **Evaluate across tasks.** If the learned person-state cannot support more than one narrow target, do not call it a stable actor ontology.

For any candidate feature family \(f\), define its held-out contribution under a lower-is-better validation risk by

\[
\Delta_{\tau,\Delta}^{\mathrm{val}}(f)
:=
\mathcal R_{-f}^{\mathrm{val}}
-
\mathcal R_{+f}^{\mathrm{val}}.
\]

A positive value means the feature family reduced held-out risk. This is an empirical contribution, not proof that the learned coordinate is uniquely real or causally fundamental.

### Drift

Use a proper scoring rule whose lower value is better, such as average negative log likelihood. Let

\[
\mathcal R_{\mathrm{recent}}(t,h)
\]

be recent risk on a rolling window and let \(\mathcal R_{\mathrm{ref}}\) be reference risk. Define degradation by

\[
\Delta_{\mathrm{drift}}(t,h)
=
\mathcal R_{\mathrm{recent}}(t,h)
-
\mathcal R_{\mathrm{ref}}.
\]

Trigger investigation when

\[
\Delta_{\mathrm{drift}}(t,h)>c_{\mathrm{drift}},
\]

or when calibration, support, or data-quality diagnostics cross their own predeclared thresholds.

Responses may include:

- refit parameters;
- refresh slow state estimates;
- revise company aggregation;
- expand or prune feature families;
- reopen the task projection;
- retrain calibration;
- reduce policy aggressiveness until overlap and support return.

A model like this is expected to become wrong. The point is to catch it when it does.

## 4.7 What Counts as Success

Success is not that the model sounds deep. Success is narrower.

For a declared Bernoulli primary head, the first forecasting system succeeds if, on temporally held-out data,

\[
\operatorname{LogLoss}(M_{\mathrm{GIDS}})
<
\operatorname{LogLoss}(M_{\mathrm{best\ baseline}})
-
\epsilon_1
\]

and

\[
\operatorname{Brier}(M_{\mathrm{GIDS}})
<
\operatorname{Brier}(M_{\mathrm{best\ baseline}})
-
\epsilon_2,
\]

with \(\epsilon_1,\epsilon_2>0\) chosen before inspecting the final test window and uncertainty intervals that exclude trivial gains. For multiclass, ordinal, continuous, survival, or joint heads, replace these two Bernoulli scores with the predeclared proper scoring rule or task-specific loss appropriate to that outcome; do not average incomparable head metrics without a declared normalization and weighting rule.

It succeeds more strongly if:

- gains survive time drift and entity holdouts;
- the slow/fast ablations behave as predicted;
- relationship state adds unique signal;
- the company aggregation improves over flat metadata;
- task transfer shows that the actor-state supports several outcomes;
- the shared interaction representation beats simpler late fusion;
- uncertainty and calibration remain usable for decision-making.

If intervention data exists, proposition selection succeeds when a policy based on the model produces higher off-policy value under defensible assumptions or higher live experimental utility than a baseline policy.

If intervention data does **not** exist, proposition search results must be described as simulated or observational rankings, not causal wins.

The framework fails its first empirical test if simpler models match or exceed it, if the monolithic sequence model dominates without a meaningful interpretability or data-efficiency tradeoff, if the learned state does not transfer, if gains disappear on future windows, or if proposition rankings fail under randomization.

In that case, either the decomposition is wrong, the data does not contain the signal I thought it did, the actor construction is wrong, or the task never needed this much machinery in the first place.

## End of Part 4

Part 4 is where the framework becomes verifiable and we can use data.

At this stage, the job is straightforward: define what state means in a form a dataset can carry, define the event clock, define what must be predicted, define which weaker and stronger models must be beaten, define what the probes are supposed to establish, and define what intervention evidence is required before proposition optimization can be called causal.

The operational arc is

\[
(\mathcal H_{<t},\text{people},\text{companies},\text{relationship},\mathbf w_t)
\longrightarrow
\widehat D_{ab,t},
\]

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

\[
(\widehat D_{ab,t},X_t=x_t)
\longrightarrow
(\widetilde{\mathbf Y}_t,\widetilde{\mathbf Z}_t),
\]

\[
(\widehat D_{ab,t},x_t,\mathcal H_{(t,t+1]})
\longrightarrow
\widehat D_{ab,t+1}
\longrightarrow
\text{benchmark}.
\]

And if we later choose messages from the model,

\[
\widehat D_{ab,t}
\longrightarrow
V_{\theta,\mathfrak e}^{\mathrm{pred}}(x\mid\widehat D_{ab,t})
\longrightarrow
x_t^\star,
\]

with the giant asterisk that ranking is not causality unless the data collection regime supports that claim.

This is the point where the theory becomes falsifiable. The question is no longer whether reality can be expressed through a common formal arena. The question is whether this construction yields better forecasts of observable human transition than models that ignore explicit actor-state, relationship, and composite context, and whether its proposition-search layer survives the much nastier standard of intervention.

---

## In Memory of Einar Kringlen.

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


---

# Appendix A: Study Guide / Cheat Sheet

This appendix is the shorter working guide to the manuscript. The complete source of truth is [Canonical Notation and Mathematical Conventions](08_Canonical_Notation.md). When the shorthand here and the canonical file appear to disagree, use the canonical file.

## The whole program in one view

The philosophical arc is

\[
\mathcal N
\xrightarrow{\;\operatorname{Reg}\;}
\mathcal G
\longrightarrow
G_i
\longrightarrow
T_{i,t}
\longrightarrow
\phi_{i,t}.
\]

Read this as:

1. there is an external domain larger than any actor's experience;
2. a model registers some external configuration in GIDS;
3. an actor inherits a restricted organization of possible distinctions;
4. development and history realize that organization in one actor;
5. the actor occupies a full phenomenal state at a time.

For complete one-step statements, use

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

The predictive arc is

\[
\mathsf I_{i,t}
\longrightarrow
Q_{i,t}
\xrightarrow{\;\Pi_{\tau,\Delta}\;}
q_{i,t}^{(\tau,\Delta)},
\]

where \(Q_{i,t}\) is the actor's general predictive response object and \(q_{i,t}^{(\tau,\Delta)}\) is a task-and-horizon summary when such a sufficient summary exists.

The operational arc for one person is

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

The first application uses a 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 learned transition and filtering loop is

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

\[
\widetilde O_{t+1}
\sim
R_{O,\theta}
\left(
\cdot
\mid
\widetilde D_{ab,t+1}
\right),
\]

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

The first equation simulates. The second emits an immediate trace. The third updates the estimate after real records arrive. They are not interchangeable.

---

## 1. What GIDS is

Let \(\mathcal N\) be the external or noumenal domain. The manuscript does **not** assume that \(\mathcal N\) itself is a vector space.

Let

\[
\mathcal G
\]

be God's Infinite Dimensional Space: an idealized real separable Hilbert space of possible distinctions that could enter the experience or response of an actor.

A registration map

\[
\operatorname{Reg}:\mathcal N_{\mathrm{loc}}\to\mathcal G
\]

turns a local external configuration into a model-side representation:

\[
\mathbf g_t^{\mathrm{reg}}=\operatorname{Reg}(\omega_t).
\]

The representation is not the external object. The map \(\operatorname{Reg}\) may lose information, identify physically different configurations, and depend on the modeling resolution.

### Why a Hilbert space is useful

A Hilbert space supplies:

- coordinates and inner products;
- limits of convergent sequences;
- finite-dimensional subspaces inside an open-ended ambient arena;
- orthogonal projection when that particular approximation is justified;
- room for distinctions unavailable to the actor currently being modeled.

It does **not** prove that cognition is linear, that psychology has one natural Euclidean chart, or that every meaningful category is a basis vector.

---

## 2. Lineage access and the inherited seed

A finite-dimensional lineage idealization is

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

If the basis vectors are orthonormal, the corresponding projection is

\[
P^{\mathrm{spec}}\mathbf g
=
\sum_{j=1}^{d}
\langle\mathbf v_j,\mathbf g\rangle\mathbf v_j.
\]

This is the closest point in \(\mathcal M^{\mathrm{spec}}\) under the chosen Hilbert norm. That geometric fact does not establish that an organism literally performs an orthogonal projection.

The inherited seed of individual \(i\) is

\[
G_i
=
\left(
\mathcal M_i^0,
A_i^0,
\mathcal I_i^0
\right),
\qquad
A_i^0:\mathcal G\to\mathcal M_i^0.
\]

Here \(A_i^0\) is an individual access map and \(\mathcal I_i^0\) is an inherited starting organization. The access map is not assumed linear or orthogonal.

---

## 3. Objects and categories are constructed through the space

A person, proposition, corporation, table, threat, or category is not automatically one primitive vector in GIDS.

For actor \(i\), an external configuration \(\omega_t\) can induce an actor-relative representation

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

The access codomain \(\mathcal V_{i,t}^{\mathrm{acc}}\) and object codomain \(\mathcal V_{i,t}^{\mathrm{obj}}\) are actor-relative spaces. Neither needs to be a linear subspace of \(\mathcal G\).

A category requires a declared representation domain \(\mathcal V_\kappa^{\mathrm{cat}}\), which may be \(\mathcal G\), an actor-relative object space, or a finite learned feature space. It may then be represented as:

- a measurable region \(\mathcal C_\kappa\subseteq\mathcal V_\kappa^{\mathrm{cat}}\);
- a prototype point \(\operatorname{proto}_\kappa\in\mathcal V_\kappa^{\mathrm{cat}}\);
- a probability measure \(\mathbb P_\kappa\) on \(\mathcal V_\kappa^{\mathrm{cat}}\);
- a score \(\operatorname{cat}_{\kappa}:\mathcal V_\kappa^{\mathrm{cat}}\to[0,1]\);
- or a learned relational object.

Only a prototype is literally a point of its declared domain. Regions are subsets; distributions and scores are mathematical objects defined on the domain. A category is therefore not required to be a vector or a primitive part of GIDS.

### Valid and invalid algebra

Valid operations are operations with a declared interpretation, such as:

- projecting a registered signal into an accessible subspace;
- calculating similarity after the encoders have been calibrated for that use;
- aggregating structured member states into a corporate state;
- applying a transition kernel;
- integrating expected utility over predicted futures.

Invalid by default:

- adding a person vector to an email vector and calling the sum a person;
- averaging employees and calling the result a corporation;
- treating every dot product as psychological compatibility;
- assuming that proximity in a learned chart is observer-independent truth.

A common arena gives us a grammar for constructing relations. It does not bless arbitrary arithmetic.

---

## 4. The person hierarchy

The slowly changing realized person is

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

The full phenomenal state is

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

The person-in-role object, the Chimera, is

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

The hierarchy matters:

- \(G_i\) is inherited starting structure;
- \(T_{i,t}\) is the realized and slowly changing person;
- \(\phi_{i,t}\) is the complete current lived state;
- \(Q_{i,t}\) is the general predictive response object;
- \(q_{i,t}^{(\tau,\Delta)}\) is a task summary when it exists;
- \(\widehat s_{i,t}\) is the finite state estimated by the model.

These are not six names for the same thing.

---

## 5. Factor analysis and “fundamental features”

Classical factor analysis writes an observed variable vector as

\[
\mathbf x_i
=
\boldsymbol\mu
+
\boldsymbol\Lambda_{\mathrm{FA}}\mathbf f_i
+
\boldsymbol\varepsilon_i,
\qquad
\mathbb E[\mathbf f_i]=\mathbf 0,
\qquad
\operatorname{Cov}(\mathbf f_i)=I,
\qquad
\mathbb E[\boldsymbol\varepsilon_i\mid\mathbf f_i]=\mathbf 0.
\]

Factor-analytic programs helped produce useful personality constructs by compressing correlations among many observations. Those constructs are often coarse conglomerations over subtler distinctions and response tendencies; their interpretation depends on the measurement design and identification or rotation convention.

GIDS uses the same broad discovery instinct but asks for finer and more operational structure:

- persistent factors;
- fast state;
- role and environment;
- proposition-conditioned activation;
- transition prediction;
- cross-task transfer;
- intervention tests.

A learned factor earns the word “fundamental” only provisionally. It should persist where persistence is expected, transfer across contexts, improve multiple predictions, survive ablation, and—when possible—participate in successful intervention. Rotation and reparameterization mean the coordinate label itself is rarely sacred; the predictive information is what matters.

---

## 6. General predictive response state

Define the ideal pre-proposition information state

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

This is the reference information bundle for prediction, not the complete actor–world state; direct access to \(\phi_{i,t}\) remains unavailable.

For a finite sequence horizon \(H\), proposition path \(\mathbf x_{t:t+H-1}\), and supplied exogenous scenario \(\boldsymbol\xi_{t+1:t+H}\), choose a version of the observational response kernel

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

This kernel is defined up to almost-sure equality and is operationally used on the support of the proposition and scenario process. For adaptive future propositions, the factual joint law must include the historical policy and candidate-set process. Merely indexing the observational kernel by a new policy does not identify that policy's counterfactual law; controlled kernels plus an identification argument, or an explicit extrapolation assumption, are required. A causal response object replaces historical conditioning with the corresponding interventional laws.

Two ideal information states are equivalent when they induce the same declared family of future response laws. This defines an abstract predictive information object. A convenient measurable quotient is not automatic: when one exists, denote it by \(Q_{i,t}\); otherwise, use the indexed response-kernel family itself. Either object may be infinite-dimensional.

For task \(\tau\) and elapsed-time horizon \(\Delta\), write

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

only when a sufficient task summary exists.

The decision-associated outcome is indexed by the decision, not by pretending the label is an immediate next state:

\[
Y_{i,t}^{(\tau,\Delta)}.
\]

For every measurable event \(B\), sufficiency means

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

This is observational predictive sufficiency unless the laws are explicitly interventional.

---

## 7. Learned slow and fast state

The operational person-state is

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

- \(\widehat{\mathbf t}_{i,t}\) is the slow learned person vector;
- \(\mathbf z_{i,t}\) is the fast latent state;
- \(\mathbf c_{i,t}\) is role and institutional context;
- \(\mathbf w_t\) is measured world state.

Let \(v_n=\operatorname{time}(\mathsf h_n)\). If the chronological record stream is shared across actors, let \(\mathbf d_{i,n}^{\mathrm{rec}}\) indicate whether record \(n\) pertains to actor \(i\) and which actor-specific fields are available. The fast state is updated after chronological record \(n\) becomes available:

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

The applicability vector is not a memory representation and not a label mask. It contains a binary applicability flag; when that flag is zero, the fast-state update is the identity.

At decision \(t\),

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

The response caused by \(x_t\) therefore cannot already be inside \(\mathbf z_{i,t}\).

### Relevance and salience

The ideal task relevance map is

\[
\Lambda_\tau
\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
 x_t
\right).
\]

Salience is

\[
\boldsymbol\lambda_{i,t}^{(\tau)}(x_t)
\in[0,1]^{d_\tau},
\]

and the active slice is

\[
\mathbf r_{i,t}^{(\tau)}(x_t)
=
\boldsymbol\lambda_{i,t}^{(\tau)}(x_t)
\odot
\Lambda_\tau
\left(
T_{i,t},
\phi_{i,t},
 c_{i,t},
 w_t,
 x_t
\right).
\]

The symbol \(\mathbf z_{i,t}\) is not reused for this active slice.

---

## 8. What approximation means

The manuscript no longer writes \(s\approx q\) and leaves the symbol unexplained.

If \(\widehat s_{i,t}\) is constructed from the full pre-proposition information state, define the information lost by compression as

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

This equals zero when the operational state retains all observational predictive information about the outcome that was available in the ideal information state, conditional on the proposition.

Separately, for a fitted conditional law \(P_{Y,\theta,\tau,\Delta}\), define model-fitting error by

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

The first error asks whether the state threw information away. The second asks whether the fitted predictor used the retained information correctly. A model can fail either way.

Under log score, these gaps add exactly to excess risk over the full-information Bayes log risk when the conditional laws admit densities or mass functions with respect to one fixed reference measure and the required expectations are finite. For discrete outcomes that Bayes risk is conditional entropy. For continuous outcomes it is expected negative log density, not an invariant entropy of the actor-state.

---

## 9. Memory and categorical traces

A tractable memory field is

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

This is a model of weighted traces, not a literal claim about storage in the brain.

Categorical observations are indexed by feature family \(f\), source channel \(\sigma\), and role or regime \(\rho\). Contextual lifting occurs before pooling so that surface contradictions can be retyped rather than averaged into nonsense.

Slow categorical memory summarizes durable evidence available before the current decision. Fast categorical memory emphasizes recent and task-relevant evidence. Missingness uses explicit masks and learned null representations rather than pretending that absence equals numerical zero.

---

## 10. Corporations as composite actors

A corporation is not a point obtained by averaging its employees.

Let \(\mathcal J_C(t)\) be the people relevant to corporation \(C\)'s decision at time \(t\), and define

\[
\varsigma_{j,t}^{\mathrm{person}}
=
\bigl(T_{j,t},\phi_{j,t},c_{j,t}\bigr).
\]

The conceptual composite is

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

The measurable company-side evidence is

\[
\mathbf u_{C,t}^{\mathrm{corp}}
=
\left[
\mathbf f_{C,t}^{\mathrm{corp}},
\mathbf g_{C,t}^{\mathrm{org}},
\mathbf m_{C,t}^{\mathrm{inst}},
\boldsymbol\iota_{C,t}^{\mathrm{inc}},
\mathbf h_{C,t}^{\mathrm{env}}
\right],
\]

and a learned estimate is

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

The terms encode relevant people, facts and statistics, authority and communication structure, institutional memory, incentives and constraints, environmental history, and explicit missingness.

The aggregator must be capable of representing veto rights, coalitions, unequal authority, and information bottlenecks. This supports treating the corporation as an actor for prediction. It does not, by itself, claim human-like consciousness.

---

## 11. The sales dyad

The first concrete state is

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

Here:

- \(a\) is the salesperson;
- \(b\) is the executive;
- \(C_a\) and \(C_b\) are their companies;
- \(\widehat\Gamma_{ab,t}\) is relationship state;
- \(\mathbf w_t\) is shared world state.

A proposition may contain content, offer, price, framing, evidence, channel, timing, sender, and requested action. The dyadic actor-relative proposition and interaction maps are

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

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

These are distinct from the individual-state maps \(\mathcal P_{s,\theta}\) and \(\Psi_{s,\theta}\). The recursively closed transition is

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

It is recursively closed because the output is the same state type required at the next step. Evaluating the parameterized kernel at a supplied \(\boldsymbol\xi_{t+1}\) gives a scenario forecast; this does not require a positive-probability singleton event. An unconditional forecast integrates the conditional transition against a declared law for \(\boldsymbol\Xi_{t+1}\).

---

## 12. Event time and labels

The pre-decision history is

\[
\mathcal H_{<t}
=
\left(
\mathsf h_n:
\operatorname{time}(\mathsf h_n)
<
\operatorname{time}(\mathsf d_t)
\right).
\]

The decision record is

\[
\mathsf d_t
=
\left(
\operatorname{id}_t,
\mathcal X_t^{\mathrm{cand}},
 x_t,
\delta_t,
\mathbf e_t^{\mathrm{cat}},
\eta_t,
\mathsf I_t^\mu
\right),
\]

with logged assignment probability or density

\[
\eta_t
=
\mu_t
\left(
 x_t
\mid
\mathsf I_t^\mu,
\mathcal X_t^{\mathrm{cand}}
\right).
\]

Primary outcomes are

\[
\mathbf Y_t
=
\left(
Y_t^{(\tau_\ell,\Delta_\ell)}
\right)_{\ell=1}^{L_Y},
\]

and auxiliary probes are

\[
\mathbf Z_t
=
\left(
Z_t^{(m,\Delta_m)}
\right)_{m=1}^{M_Z}.
\]

The label mask is

\[
R_{t,\ell}^{\mathrm{obs}}
=
\mathbf 1
\{Y_t^{(\tau_\ell,\Delta_\ell)}
\text{ is observed by the analysis cutoff}\}.
\]

A missing 90-day label is not a negative. It is censored or unavailable.

Probe labels need their own masks:

\[
R_{t,m}^{Z,\mathrm{obs}}
=
\mathbf 1\{Z_t^{(m,\Delta_m)}\text{ is observed by the analysis cutoff}\}.
\]

---

## 13. Forecasting, ranking, and causality

For a score extending beyond the immediate next response, declare an evaluation regime \(\mathfrak e\) containing the continuation policy, future candidate-set process, exogenous-path law \(\mathbb Q_{\Xi}\), and the outcome/censoring convention. The predictive value is

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

The outcome and probe bundles must be functionals of the same rollout or draws from a coherent joint law \(P_{YZ,\theta,\mathfrak e}\). If only separate marginal heads are fitted, the utility may use marginal expectations or an explicitly declared coupling; separate heads do not silently imply independence. When a maximum is attained over a nonempty available set \(\mathcal X_t^{\mathrm{cand}}\subseteq\mathcal X_t^{\mathrm{adm}}\), ranking chooses

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

A nonempty finite candidate set guarantees an attained maximum. For an infinite candidate set, use conditions such as compactness and upper semicontinuity, or replace the argmax with a supremum and an approximate optimizer. This is a forecast under the fitted model. The corresponding causal value under the same regime is

\[
V_{\mathfrak e}^{\mathrm{causal}}(x\mid\widehat D_{ab,t})
=
\mathbb E_{\mathfrak e}
\left[
U_\tau
\mid
\widehat D_{ab,t},
\operatorname{do}(X_t=x)
\right].
\]

The two values are equal only under a valid identification argument and a correctly estimated identified law.

A policy is

\[
\pi_t
\left(
 x
\mid
\widehat D_{ab,t},
\mathcal X_t^{\mathrm{cand}}
\right).
\]

Let \(\mathfrak P_{\mathrm{adm}}\) be the nonempty admissible policy class, \(\gamma\in[0,1]\), and let \(\mathbb Q_{\Xi}\) and \(\mathbb Q_{\mathcal X}\) be the declared exogenous-path and future candidate-set laws. For a sequence horizon \(H\), the planning objective is

\[
J_{\theta,\mathbb Q_{\Xi},\mathbb Q_{\mathcal X}}(\pi\mid\widehat D_{ab,t})
=
\mathbb E_{\theta,\pi,\mathbb Q_{\Xi},\mathbb Q_{\mathcal X}}
\left[
\sum_{k=0}^{H-1}
\gamma^k
u_\tau^{\mathrm{step}}
\left(
\widetilde D_{ab,t+k+1},
X_{t+k},
\widetilde O_{t+k+1}
\right)
+
\gamma^H
V_\tau^{\mathrm{term}}
\left(
\widetilde D_{ab,t+H}
\right)
\mid
\widehat D_{ab,t}
\right].
\]

This is the proper place to assign sequence credit. Do not copy the full eventual transaction reward backward onto every prior message. Use an argmax only when the maximum is attained; otherwise optimize the supremum.

---

## 14. Training and off-policy evaluation

For \(L_Y\) primary heads and \(M_Z\) probe heads, minimize

\[
\mathscr J(\theta)
=
\sum_{\ell=1}^{L_Y}
\lambda_\ell^Y
\mathscr J_\ell^Y(\theta)
+
\sum_{m=1}^{M_Z}
\lambda_m^Z
\mathscr J_m^Z(\theta)
+
\lambda_{\mathrm{reg}}\Omega(\theta),
\]

with nonnegative weights and \(\Omega(\theta)\ge0\).

Every sign is positive because each term is minimized. Masks, censoring weights, or survival likelihoods belong inside the corresponding head loss.

A slow-vector moving average is valid only when the old and refreshed vectors share one coordinate chart: the encoder is fixed, the new representation is aligned, or the relevant histories are re-encoded after an encoder change.

For one-step off-policy evaluation, let \(\mathcal T_{\mathrm{ope}}\) be the predeclared set of eligible decisions whose utility is mature under the chosen censoring rule, and let \(N_{\mathrm{ope}}=|\mathcal T_{\mathrm{ope}}|>0\). Then

\[
\widehat V_{\mathrm{IPS}}(\pi)
=
\frac{1}{N_{\mathrm{ope}}}
\sum_{t\in\mathcal T_{\mathrm{ope}}}
\frac{
\pi_t(x_t\mid\widehat D_{ab,t},\mathcal X_t^{\mathrm{cand}})
}{
\eta_t
}
u_t^{\mathrm{obs}}.
\]

Here \(u_t^{\mathrm{obs}}\) must be observed and mature, or handled with an appropriate censoring model or weight. If it extends beyond the immediate response, the one-step estimand changes the current proposition under a fixed declared or logged continuation regime. The target-policy numerator must use only pre-decision information. For continuous actions, numerator and denominator are densities under the same dominating measure. Report weight diagnostics and effective sample size. Clipping and self-normalization trade variance for bias or a changed finite-sample estimand; doubly robust methods are often preferable when nuisance models are credible. The estimator requires correct logged probabilities, consistency, overlap, a valid assignment or ignorability argument, correct event ordering, and appropriate treatment of delayed outcomes, censoring, and interference. Learned policies require held-out evaluation or cross-fitting. Sequence policies require sequential estimators; this one-step formula is not stretched across an entire conversation.

---

## 15. What the benchmark must establish

The proposed model must be compared against:

1. prevalence-only prediction;
2. current-proposition features;
3. static dyadic tabular features;
4. shallow history summaries;
5. a two-tower recommender-style model;
6. a monolithic sequence model with the same available data.

The evaluation must use future time windows and should separately test:

- known people and known companies;
- new people inside known companies;
- new companies;
- new proposition families;
- new outcome horizons.

Ablations should remove:

- fast state;
- slow state;
- salesperson state;
- relationship state;
- composite company aggregation;
- source-aware pooling;
- salience weighting;
- probe heads;
- the shared interaction representation.

For a feature family \(f\), under a lower-is-better held-out risk,

\[
\Delta_{\tau,\Delta}^{\mathrm{val}}(f)
=
\mathcal R_{-f}^{\mathrm{val}}
-
\mathcal R_{+f}^{\mathrm{val}}.
\]

A positive value means the feature family reduced held-out risk. It does not prove that one learned coordinate is the unique metaphysical feature of the person.

---

## 16. Symbols that must not be confused

| Symbol | Meaning |
|---|---|
| \(\nu\) | evolutionary stage |
| \(\mathbf e_{\varnothing}\) | learned empty-cell representation; not evolutionary stage |
| \(\tau\) | prediction task |
| \(H\) | planning length in decision steps |
| \(\mathcal H\) | history |
| \(\Pi_{\tau,\Delta}\) | task-summary map |
| \(\pi_t\) | decision policy |
| \(\mathbf z_{i,t}\) | fast latent state only |
| \(\mathbf r_{i,t}^{(\tau)}(x)\) | proposition-conditioned active slice |
| \(\mathcal A_{\mathrm{corp}}\) | corporate aggregation operator |
| \(\mathbf Z_t\) | auxiliary probe bundle |
| \(R_{O,\theta}\) | immediate trace law |
| \(P_{Y,\theta,\tau,\Delta}\) | delayed outcome law |
| \(\boldsymbol\Xi\), \(\boldsymbol\xi\) | random exogenous path and a realized/supplied path value |
| \(\mathbb Q_{\mathcal X}\) | future candidate-set law in planning |
| \(\mathfrak e\) | declared predictive evaluation regime |
| \(\mathfrak P_{\mathrm{adm}}\) | admissible policy class |
| \(\mathscr J(\theta)\) | training objective |
| \(\mathcal L(X\mid Z)\) | conditional probability law |
| \(\mathcal P_{s,\theta},\Psi_{s,\theta}\) | individual-state proposition and interaction maps |
| \(\mathcal P_{D,\theta},\Psi_{D,\theta}\) | dyadic proposition and interaction maps |
| \(\mathbf d_{i,n}^{\mathrm{rec}}\) | actor-specific record applicability; not memory or a label mask |
| \(d_{ab,n}^{\mathrm{rel}}\) | relationship-record applicability; zero means the relationship update is the identity |
| \(a,b\) | salesperson and executive |
| \(C_a,C_b\) | their corporations |

## Shortest usable summary

The framework says:

- reality contains more possible distinctions than one actor can use;
- an actor inherits and develops a way of organizing some of those distinctions;
- traces provide evidence about slow organization and fast state;
- a proposition is reorganized through the receiving actor rather than merely placed beside them;
- the interaction changes an operational state;
- immediate and delayed observables are consequences of that transition or trajectory;
- propositions can be ranked by forecasted utility;
- causal selection requires intervention-grade evidence.

That is the system in its shortest accurate form.


---

## Canonical notation

The canonical notation is maintained as a separate source of truth: [Canonical Notation and Mathematical Conventions](08_Canonical_Notation.md).