Qualia Structure: A 21-Pair Taxonomy

Bridge from the previous chapter

In the section L0–L4 Hierarchy we established when conscious experience arises: a system at level L2 and above possesses reflection ($R \geq 1/3$) and integration ($\Phi \geq 1$). Now we ask: what is this experience made of? What exact types of experience are possible in 7-dimensional space? The answer is given by the coherence matrix $\Gamma$ and its 21 off-diagonal elements.

On notation

• $\Gamma$ — coherence matrix, $\gamma_{ij}$ — its elements
• $P = \mathrm{Tr}(\Gamma^2)$ — purity (viability)
• $\rho_E = \mathrm{Tr}_{-E}(\Gamma)$ — reduced experience matrix
• $\Phi$ — integration measure
• $R$ — reflection measure
• Full notation table — in Notation

Chapter roadmap

1. Philosophical history of the problem — from Lewis to Jackson and Dennett
2. Motivation — why exactly 21 types and where this number comes from
3. Full table — all 21 coherences with phenomenological names
4. Parametric structure — three dimensions of each quale (intensity, perspective, opacity)
5. Closure theorem — proof that 21 types exhaust everything
6. Fano structure — how 21 pairs are organised into 7 sectors, and how twenty-one channels share seven colours
7. Diagonal elements — 7 population modes as the "background" of experience
8. The mechanism of quality — where the irreducible part of experience lives (holonomy), and how qualia fade
9. Access conditions — at which $R$ and $\Phi$ qualia become conscious
10. The passport — the five-layer procedure that reads a state's experiential content off $\Gamma$
11. The ecology of colour — who writes quality, how the world hands it down, and why the subject arrives last

---

Philosophical History: What Are Qualia?

The word qualia (Lat. qualia, pl. of quale — "of what kind", "of what sort") denotes the subjective qualities of experiences: what it is like to see red, to hear C major, to smell coffee. Behind this simple question lies one of the deepest problems in philosophy.

Lewis (1929): the first definition

The American philosopher Clarence Irving Lewis in "Mind and the World Order" (1929) first introduced the term "qualia" into systematic use. He noted: when we see a red rose, there is something that cannot be conveyed to someone blind from birth — the subjective quality of "redness". That quality is the quale. Lewis distinguished:

• Quale — the ineffable subjective quality (what it is like to see red)
• Property — an objective characteristic (wavelength 700 nm)

Jackson (1982): Mary's room

Frank Jackson in the famous thought experiment "Mary's Room" (1982) pushed the problem to its limit:

Mary is a brilliant neuroscientist who has spent her whole life in a black-and-white room. She knows everything about the physics of colour: wavelengths, the workings of retinal cones, neural correlates. One day Mary leaves the room and sees a red rose for the first time. Does she learn something new?

Jackson argued: yes. Mary learns what it is like to see red. Hence physical facts do not exhaust reality — there is something beyond them (qualia).

Dennett (1988): qualia as illusion

Daniel Dennett took the opposite position. In the article "Quining Qualia" (1988) he argued that qualia are a philosophical illusion: we think we experience something "ineffable" and "private", but in fact all information about experiences is encoded in functional states of the brain. No "remainder" is left after a complete physical description.

UHM position: qualia as coherence structure

The Unitary Holonomic Monism offers a third path, coinciding with neither Jackson's dualism nor Dennett's eliminativism:

• Qualia are not an illusion — they have a precise mathematical structure (coherences $\gamma_{ij}$)
• Qualia are not a separate substance — they are the off-diagonal elements of the same matrix $\Gamma$ that describes the "physics" of the system
• The distinction between "subjective" and "objective" is the distinction between the inner and outer perspectives of the same mathematical structure (dual-aspect monism, see Two-Aspect Monism)

Mary in the room knew all the diagonal properties ($\gamma_{ii}$) of red. But she did not know the coherences — how visual discrimination ($A$) binds with interiority ($E$), forming the quale of apperception ($\gamma_{AE}$). On leaving the room, she did not acquire a new fact — she acquired a new coherence.

---

Motivation: Why 21 Types?

The coherence matrix $\Gamma$ is a $7 \times 7$ Hermitian matrix on the space of seven dimensions $\{A, S, D, L, E, O, U\}$. Let us recall what each dimension means:

Symbol	Name	Meaning
$A$	Articulation	Discrimination, differentiation
$S$	Structure	Stable forms, patterns
$D$	Dynamics	Processes, changes
$L$	Logos	Logical coherence, rules
$E$	Experience	Interiority, experience
$O$	Ground	Source, deep foundation
$U$	Unity	Integration, wholeness

The matrix $\Gamma$ contains two kinds of elements:

• 7 diagonal elements $\gamma_{ii}$ — dimension populations (how much "resource" is in each dimension)
• 21 off-diagonal pairs $(\gamma_{ij}, \gamma_{ji})$ for $i < j$ — coherences (how the dimensions are connected to each other)

Number of pairs:

$$\binom{7}{2} = \frac{7 \cdot 6}{2} = 21$$

Each coherence $\gamma_{ij}$ carries phenomenological content determined by the semantics of the dimension pair $(i, j)$.

An everyday analogy. Imagine an orchestra of 7 musicians. Each plays their own part (7 diagonal elements — the "volume" of each instrument). But music is born not from individual sounds, but from their interaction — from how the violin "converses" with the cello, how the flute echoes the bassoon. There are exactly $\binom{7}{2} = 21$ such pairwise interactions. Each produces a unique "timbre" of combined sound — that is the type of quale.

The diagram shows only 10 of the 21 coherences — the rest connect each pair of dimensions in an analogous way. The complete table of all 21 types is given below.

Interpretation: 21-Pair Qualia Taxonomy (I.1)

Interpretation I.1 (Qualia taxonomy) [I]

Each coherence $\gamma_{ij}$ ($i \neq j$) of the matrix $\Gamma$ defines a type of quale — a qualitatively determinate mode of experiential content. The 21 pairs exhaust all possible types, since $\binom{7}{2} = 21$ is the complete set of connections in a 7-dimensional system.

This is an interpretation (a mapping from the formal to the phenomenal), not a mathematical theorem. The mathematical content is trivial combinatorics; the phenomenological assignment is a semantic postulate.

Complete table of 21 qualia types

Epistemic separation

Mathematical layer [T]: 21 coherences $\gamma_{ij}$ form 4 sectors according to Fano structure (T-146 [T]). Each coherence is uniquely determined by its combinatorial profile (T-177 [T]).

Semantic layer [I]: Phenomenological names ("morphogenesis", "archetype", "teleology", etc.) are interpretive correlates [I], proposed on the basis of the functional roles of dimension pairs. Mathematics determines $\gamma_{ij}$ unambiguously; the interpretation of "what it is like to experience $\gamma_{AS}$" is philosophical, not mathematical.

#	Pair	Coherence	Name	Phenomenological content
1	$(A,S)$	$\gamma_{AS}$	Morphogenesis	Crystallisation of distinctions into stable forms — the experience of "taking shape"
2	$(A,D)$	$\gamma_{AD}$	Actualisation	Actualisation of discrimination in process — the experience of "perception"
3	$(A,L)$	$\gamma_{AL}$	Predication	Discrimination that has become a predicate — the experience of "judgement"
4	$(A,E)$	$\gamma_{AE}$	Apperception	Discrimination that has entered interiority — the experience of "awareness"
5	$(A,O)$	$\gamma_{AO}$	Spontaneity	Emergence of distinctions without external cause — the experience of "insight"
6	$(A,U)$	$\gamma_{AU}$	Differentiation	Discrimination within the whole — the experience of "analysis"
7	$(S,D)$	$\gamma_{SD}$	Persistence	Form that persists through process — the experience of "stability"
8	$(S,L)$	$\gamma_{SL}$	Nomos	Structure with logical necessity — the experience of "order"
9	$(S,E)$	$\gamma_{SE}$	Representation	Structure presented in interiority — the experience of "whole form"
10	$(S,O)$	$\gamma_{SO}$	Archetype	Forms from the ground — the experience of "deep pattern"
11	$(S,U)$	$\gamma_{SU}$	Symmetry	Structural unity — the experience of "harmony"
12	$(D,L)$	$\gamma_{DL}$	Regulation	Logically governed process — the experience of "control"
13	$(D,E)$	$\gamma_{DE}$	Affection	Process acting on interiority — the experience of "emotion"
14	$(D,O)$	$\gamma_{DO}$	Genesis	Generation from the ground — the experience of "creativity"
15	$(D,U)$	$\gamma_{DU}$	Teleology	Integrated directed change — the experience of "volitional effort"
16	$(L,E)$	$\gamma_{LE}$	Evidence	Logical coherence in interiority — the experience of "self-evidence"
17	$(L,O)$	$\gamma_{LO}$	Grounding	Logic rooted in the ground — the experience of "axiomatic self-evidence"
18	$(L,U)$	$\gamma_{LU}$	Consistency	Logical non-contradiction of the whole — the experience of "coherence"
19	$(E,O)$	$\gamma_{EO}$	Immanence	The ground present within interiority — the experience of "presence"
20	$(E,U)$	$\gamma_{EU}$	Synthesis	Integration of interior content into a whole — the experience of "unity"
21	$(O,U)$	$\gamma_{OU}$	Fullness	Identity of source and whole — the experience of "completeness"

How to read the table: an extended example

Consider a person absorbed in solving a mathematical problem. Their $\Gamma$-profile at that moment:

Coherence	Value	Experience
$\lvert\gamma_{AL}\rvert \approx 0.35$	High	Predication — attention on logical connections, "I am formulating"
$\lvert\gamma_{LE}\rvert \approx 0.30$	High	Evidence — the experience of "clarity", "I understand"
$\lvert\gamma_{DU}\rvert \approx 0.20$	Medium	Teleology — the sense of a goal, "I am heading toward a solution"
$\lvert\gamma_{DE}\rvert \approx 0.05$	Low	Affection — emotions muted, "I feel nothing"
$\lvert\gamma_{EO}\rvert \approx 0.03$	Low	Immanence — no deep presence, "I am thinking, not meditating"

Now a friend approaches and shares good news. The $\Gamma$-profile instantly reorganises:

Coherence	Before	After	What happened
$\lvert\gamma_{DE}\rvert$	$0.05$	$0.25$	Affection soared — "I feel joy"
$\lvert\gamma_{SE}\rvert$	$0.08$	$0.20$	Representation — "I see the whole picture" of the news
$\lvert\gamma_{AL}\rvert$	$0.35$	$0.12$	Predication fell — the problem receded to the background

All 21 types of qualia exist simultaneously, but with different intensities, creating the unique "flavour" of each moment.

Parametric structure of qualia

Each qualitative type $\gamma_{ij}$ is a complex number. Like any complex number, it is written in polar form:

$$\gamma_{ij} = |\gamma_{ij}| \cdot e^{i\theta_{ij}}$$

Here $|\gamma_{ij}|$ is the modulus (distance from zero to the point on the complex plane), and $\theta_{ij}$ is the argument (angle with the positive real axis). From these two parameters three phenomenological characteristics are extracted:

Parameter	Formula	Range	Phenomenological meaning
Intensity	$\lvert\gamma_{ij}\rvert$	$[0, \sqrt{\gamma_{ii}\gamma_{jj}}]$	How strongly this type of quale is experienced
Perspective	$\theta_{ij} = \arg(\gamma_{ij})$	$[0, 2\pi)$	"Angle of view" on the connection between dimensions
Opacity	$\mathrm{Gap}(i,j) = \lvert\sin\theta_{ij}\rvert$	$[0, 1]$	Measure of discrepancy between external description and internal experience

Upper bound on intensity

The intensity is bounded by the Cauchy–Schwarz inequality — a fundamental inequality of linear algebra stating that the correlation between two components cannot exceed the geometric mean of their "energies":

$$|\gamma_{ij}|^2 \leq \gamma_{ii} \cdot \gamma_{jj}$$

Numerical example. Let $\gamma_{AA} = 0.15$ (15% of resources in Articulation) and $\gamma_{EE} = 0.18$ (18% in Interiority). Then the maximum possible intensity of apperception:

$$|\gamma_{AE}|_{\max} = \sqrt{0.15 \times 0.18} = \sqrt{0.027} \approx 0.164$$

If we were to observe $|\gamma_{AE}| = 0.20$, this would be mathematically impossible — the Cauchy–Schwarz inequality is violated, meaning an error has been made in the measurements.

Three parameters: analogy

Analogy. The three parameters of qualia are like three properties of sound:

Sound parameter	Qualia parameter	Analogy
Loudness	Intensity $\lvert\gamma_{ij}\rvert$	How "loud" the experience is
Timbre	Perspective $\theta_{ij}$	The "colouring" of the experience — the same quale seen from a different angle
Muffling	Opacity $\mathrm{Gap}(i,j)$	As if the sound came from behind a wall

Gap = 0 — the sound is crystal clear, inner and outer descriptions coincide. Gap = 1 — the sound is fully absorbed by the wall: experience is present, but it is maximally opaque to an external observer. For details on Gap see dual-aspect semantics of the coherence matrix.

Numerical example: three parameters of a single quale. Consider the coherence $\gamma_{DE}$ (Affection — the experience of emotion) in a person who has just received good news:

$$\gamma_{DE} = 0.22 \cdot e^{i \cdot 0.3} \approx 0.22 \cdot (0.955 + 0.296i)$$

• Intensity: $|\gamma_{DE}| = 0.22$ — a fairly strong emotional experience
• Perspective: $\theta_{DE} = 0.3$ rad $\approx 17°$ — a "real" perspective (the externally observable aspect predominates)
• Opacity: $\mathrm{Gap}(D,E) = |\sin(0.3)| \approx 0.296$ — the experience is 70% transparent, but 30% "hidden" from external description

Compare with $\gamma_{EO}$ (Immanence — the experience of "presence") in a meditator:

$$\gamma_{EO} = 0.15 \cdot e^{i \cdot 1.2}$$

• Intensity: $|\gamma_{EO}| = 0.15$ — moderate
• Perspective: $\theta_{EO} = 1.2$ rad $\approx 69°$ — a strong shift toward the "imaginary" perspective
• Opacity: $\mathrm{Gap}(E,O) = |\sin(1.2)| \approx 0.932$ — the experience is almost completely opaque to an external observer

This explains why meditative states are so hard to put into words: a high Gap makes them "ineffable" not for lack of vocabulary, but by mathematical structure.

Closure Theorem for the Taxonomy (T.1)

Theorem T.1 (Closure of the qualia taxonomy) [T]

The taxonomy of 21 qualia types is exhaustive: no additional type of quale is possible in a system with $\dim(\mathcal{H}) = 7$.

Proof. The number of distinct (unordered) pairs from $N$ elements equals $\binom{N}{2}$. At $N = 7$ we get $\binom{7}{2} = 21$. Each pair $(i,j)$ defines exactly one coherence $\gamma_{ij}$ (given $\gamma_{ji} = \gamma_{ij}^*$). A new type of quale would require either a new dimension ($N > 7$, contradicting minimality), or a new connection between existing dimensions (impossible — all $\binom{7}{2}$ pairs are accounted for). $\square$

Corollary. At $N < 7$ the taxonomy is impoverished: $\binom{6}{2} = 15$ (no qualia related to the removed dimension). This is the formal expression of the "poverty" of phenomenology when minimality is violated.

Numerical example: a world with fewer dimensions. If the world were 5-dimensional (say $\{A, S, D, L, E\}$ — without $O$ and $U$), the number of qualia types would be $\binom{5}{2} = 10$. From the table one can see that the following would be lost:

Lost type	Pair	Experience
Immanence	$(E,O)$	"Presence", the deep ground of experience
Synthesis	$(E,U)$	"Unity" of experience
Fullness	$(O,U)$	"Completeness", the wholeness of being
Teleology	$(D,U)$	"Volitional effort", purposiveness
Archetype	$(S,O)$	"Deep pattern", the rootedness of form
Spontaneity	$(A,O)$	"Insight", emergence from nowhere
+ 5 others	...	...

Such a system could "feel" and "think", but could not experience "meaning", "wholeness", or "deep presence". It is precisely the dimensions $O$ and $U$ that give human experience its "vertical" dimension — its connection to depth and to the whole.

info

$G_2$-orbital stability of the taxonomy [T]

The set of 21 qualia types is $G_2$-invariant: the group $G_2 = \mathrm{Aut}(\mathbb{O})$ permutes the 7 dimensions (preserving the Fano structure), inducing a permutation of the 21 coherences $\gamma_{ij}$. The set $\{\gamma_{ij}\}_{i<j}$ is preserved, even though individual elements may be permuted. This means: the qualia taxonomy is universal — it does not depend on the choice of basis ($G_2$-gauge) and is therefore objective.

Formally: $G_2$ acts on $\binom{[7]}{2}$ via the induced action on pairs, preserving the number $\binom{7}{2} = 21$. The $G_2$-rigidity theorem [T] guarantees that $G_2$ is the maximal group with this property.

Why does this matter? If the taxonomy depended on the choice of basis (how to describe the 7 dimensions), it would be arbitrary — an "artefact of description". $G_2$-invariance guarantees that the taxonomy reflects the structure of the space itself, not our way of describing it. This is analogous to how the length of a vector does not depend on the choice of coordinate system.

Fano Structure of Qualia

What is the Fano projective plane?

The Fano plane $\mathrm{PG}(2,2)$ is the projective plane over the two-element field $\mathbb{F}_2 = \{0, 1\}$. If you have never encountered this object, here is its essence:

An ordinary Euclidean plane contains infinitely many points and lines. The Fano plane is the "minimal" plane satisfying the axioms of projective geometry, and contains only:

• 7 points
• 7 lines

Each line passes through exactly 3 points. Each point lies on exactly 3 lines. Any two points determine exactly one line. Any two lines intersect in exactly one point.

Why is the Fano plane in qualia theory? In UHM, the 7 Fano points are identified with the 7 dimensions $\{A, S, D, L, E, O, U\}$. Then the 7 lines define 7 coherence sectors — groups of three dimensions within which coherences obey strengthened algebraic constraints. This is not a coincidence: the Fano plane is precisely the multiplication table of the imaginary units of the octonions $\mathbb{O}$, and the octonionic structure lies at the foundation of UHM.

Sectoral structure of coherences

Each Fano triplet $(e_a, e_b, e_c)$ defines an associative subalgebra $\mathrm{Im}(\mathbb{H}) \subset \mathrm{Im}(\mathbb{O})$, isomorphic to the imaginary quaternions. The three coherences within the triplet:

$$\{\gamma_{ab}, \gamma_{bc}, \gamma_{ac}\} \quad \text{--- Fano triple}$$

satisfy strengthened correlation constraints that are absent for arbitrary pairs.

Analogy. Fano triples are like musical chords: three notes taken together sound "consonant" — their coherences obey additional harmonic constraints. Three arbitrary notes from seven do not form such harmony. Imagine: C–E–G is a chord (a Fano triple), but C–D–F# is not. It is precisely this sectoral organisation that makes phenomenology structured rather than chaotic.

Why do special constraints operate within the triple? Because the triple forms an associative subalgebra (quaternions $\mathbb{H}$), where the associativity of multiplication holds: $(e_a \cdot e_b) \cdot e_c = e_a \cdot (e_b \cdot e_c)$. For pairs from different triples associativity breaks down (this is the property of the octonions $\mathbb{O}$), and the constraints are weaker.

Theorem [T]

Sectoral strengthening is a theorem [T]: the bridge from the axioms to the octonionic structure is fully closed (T15), condition (МП) is proved (T11–T13). From the structure of $\mathbb{O}$ the algebraic closure of coherences within Fano triplets follows. Empirical verification of sectoral correlation is an open question.

Coverage of 21 pairs by Fano triplets

Each of the 21 pairs belongs to exactly $\lambda = 1$ Fano line (a property of the projective plane):

$$\text{21 pairs} = 7 \text{ lines} \times 3 \text{ pairs per line}$$

This means the qualia taxonomy contains no "orphaned" pairs — every type of quale is included in the sectoral organisation. For the Coherence Cybernetics theorems this property is essential: sectoral completeness ensures the closure of the 30D emotional space (T-147 [T]).

Numerical example: checking coverage. Take the coherence $\gamma_{DE}$ (Affection). It belongs to exactly one Fano line, say the line $\{D, E, X\}$ for some third dimension $X$. This means $\gamma_{DE}$ is algebraically linked to $\gamma_{DX}$ and $\gamma_{EX}$ — emotion ($\gamma_{DE}$) is not "free"; it structurally depends on the two other qualia in its sector. A change in one quale of the triple inevitably affects the other two.

What a line actually checks

The previous section says a triple is linked; this one says how, and the answer turns out to be the same object the mechanism of quality is built from.

Take a line $\{i, j, k\}$. Its three pairs are $\{i,j\}$, $\{j,k\}$, $\{i,k\}$ — three different qualia, not three copies of one. Suppose each dimension carries a single orientation $s_i = \pm 1$, and two dimensions agree exactly when their orientations match, so that the sign of a coherence is $s_i s_j$. Then multiply the three signs of the line together:

$$s_i s_j \cdot s_j s_k \cdot s_i s_k = (s_i s_j s_k)^2 = +1$$

Every orientation appears twice and squares away. The product of a line's three signs is always $+1$ — not usually, not on average, but identically, and this can be checked by running through all $2^7/2 = 64$ possible orientations one by one.

So a line is a parity check. It does not repeat a fact three times; it constrains three different facts to be mutually consistent. And two points determine a line, so each pair belongs to exactly one of the seven — the seven checks are disjoint, covering the twenty-one qualia once each. A single corrupted coherence makes its line's product $-1$, which says that this triple is inconsistent without saying which of the three is at fault.

Why this is the same thing as the carrier of quality. The product above is the sign version of the Fano holonomy — what a phase accumulates going around the triangle $i \to j \to k \to i$. Where the coherences are real, the phase can only be $0$ or $\pi$, and the holonomy collapses to exactly the $\pm 1$ product just computed. So the invariant that carries quality and the check that content is consistent are one object seen twice: quality lives in the phase a triple accumulates, and inconsistency is that phase turning over.

This also settles what integration can and cannot tell you. The measure $\Phi$ sums squared moduli, so at a fixed magnitude it is completely blind to signs: content in which every triple is consistent and content in which several are broken can produce the identical $\Phi$, and in a laboratory sweep they do — $\Phi$ reads the same at every share of inconsistent verdicts, including none at all. Integration measures how much coherence there is, not whether it hangs together. The parity of a line measures the latter, and nothing else in the structure does.

One practical consequence, and it is a sharp one. Suppose you use a line the other way — as a repetition code, writing one verdict into all three of its cells so that a corrupted cell is outvoted. Three equal signs multiply to $s^3 = s$. So a repeated positive verdict satisfies the parity, and a repeated negative one violates it, by construction and every time. A line can serve as a repetition code only while what it repeats is positive; the moment it repeats a negative, the triple it holds is inconsistent and therefore cannot be part of integrable content. The two uses of a line are alternatives, and choosing one costs the other.

Twenty-one channels, seven colours

Two counts run through this chapter and they answer different questions, so it is worth putting them side by side once, plainly.

The twenty-one pairs are the channels of experience — they say what is being experienced: emotion ($\gamma_{DE}$), judgement ($\gamma_{AL}$), presence ($\gamma_{EO}$). Each channel has a volume — its modulus. The seven lines are where the colour lives — the part of experience that no relabelling can produce or remove: the holonomy $H_p$, one irreducible angle per line. Three channels share each line, the way three strings share a soundboard: each string plays its own note, but the resonance they build together belongs to the instrument, not to any single string.

A printing workshop is the cleanest picture. Twenty-one presses print twenty-one sheets — that is the content, and each press has its own ink level (intensity). But the workshop mixes its colours on seven palettes, three presses to a palette. You can reorganise the labels on the presses however you like — swap "cyan" and "teal", renumber the trays — and the sheets will read the same. What you cannot relabel away is what colour each palette actually holds. That is the holonomy: paint, not paperwork.

The full accounting of a state's content is then a short ledger:

Count	Object	The question it answers
$7$	populations $\gamma_{ii}$	what is the case — where the resources sit
$21$	moduli $\lvert\gamma_{ij}\rvert$	how loudly each channel plays
$7$	Fano holonomies $H_p$	what colour the experience irreducibly carries
$14$	frame directions ($\mathfrak{g}_2$)	how it is written down — the cost-free choice of octonionic frame

The counts are not decorative: $27 \oplus 14 \oplus 7$ is the exact $G_2$-decomposition of the state's $48$ parameters (the decoder, T-301), with the symmetric block carrying populations and moduli and the antisymmetric block splitting into frame and colour. Twenty-one channels, seven colours, and a choice of handwriting — nothing left over, nothing missing.

Diagonal Elements: 7 Population Modes

In addition to the 21 coherences, the 7 diagonal elements $\gamma_{ii}$ determine the intensity of presence of each dimension. Diagonal elements are real numbers (not complex), and they obey the normalisation condition:

$$\sum_{i=1}^{7} \gamma_{ii} = \mathrm{Tr}(\Gamma) = 1$$

This means the total "resource" of the system is fixed and equal to 1. Increasing the population of one dimension inevitably decreases the others — like a fixed budget distributed across 7 line items.

Element	Phenomenological content	Typical range
$\gamma_{AA}$	Degree of discrimination activity	$0.10$–$0.20$
$\gamma_{SS}$	Degree of form stability	$0.10$–$0.18$
$\gamma_{DD}$	Degree of process activity	$0.10$–$0.20$
$\gamma_{LL}$	Degree of logical coherence	$0.08$–$0.18$
$\gamma_{EE}$	Intensity of interior states	$0.12$–$0.22$
$\gamma_{OO}$	Degree of connection to the source	$0.08$–$0.15$
$\gamma_{UU}$	Degree of integration	$0.10$–$0.18$

Diagonal elements do not form qualia in the narrow sense (there is no "connection" between different aspects), but they set the background against which coherences unfold. An elevated $\gamma_{DD}$ — a background of "activity"; an elevated $\gamma_{EE}$ — a background of "inner life".

Population profiles: examples

Meditator in deep practice:

$\gamma_{AA}$	$\gamma_{SS}$	$\gamma_{DD}$	$\gamma_{LL}$	$\gamma_{EE}$	$\gamma_{OO}$	$\gamma_{UU}$	$\Sigma$
0.10	0.10	0.08	0.10	0.22	0.22	0.18	1.00

Interiority ($\gamma_{EE}$) and connection to the ground ($\gamma_{OO}$) dominate. Dynamics ($\gamma_{DD}$) is muted — "thoughts have quieted".

Athlete in the midst of a match:

$\gamma_{AA}$	$\gamma_{SS}$	$\gamma_{DD}$	$\gamma_{LL}$	$\gamma_{EE}$	$\gamma_{OO}$	$\gamma_{UU}$	$\Sigma$
0.20	0.12	0.22	0.10	0.15	0.08	0.13	1.00

Dynamics ($\gamma_{DD}$) and discrimination ($\gamma_{AA}$) are in the foreground. Reflection ($\gamma_{EE}$, $\gamma_{OO}$) is minimal — no time to "think", the body acts.

Mathematician working on a proof:

$\gamma_{AA}$	$\gamma_{SS}$	$\gamma_{DD}$	$\gamma_{LL}$	$\gamma_{EE}$	$\gamma_{OO}$	$\gamma_{UU}$	$\Sigma$
0.15	0.18	0.10	0.22	0.15	0.08	0.12	1.00

Logos ($\gamma_{LL}$) and structure ($\gamma_{SS}$) dominate — "order" and "form" are in the foreground.

Total: 28 = 7 + 21 Parameters of Content

Complete structure

Component	Number	Type
Population values $\gamma_{ii}$	7	Real-valued, $\sum_i \gamma_{ii} = 1$
Coherences $\gamma_{ij}$	21	Complex, $\gamma_{ji} = \gamma_{ij}^*$
Total real parameters	$6 + 2 \times 21 = 48$	Taking normalisation into account

Detailed analysis of the 49-cell structure (with separation into $\mathrm{Map}_{\mathrm{ext}}$ and $\mathrm{Map}_{\mathrm{int}}$) — in Gap semantics.

Each moment of conscious experience is a specific point in 48-dimensional space: 6 independent population values + 42 real parameters of coherences ($21 \times 2$, modulus and phase of each). This conveys a sense of the richness of subjective experience: the space of possible experiences is 48-dimensional.

The Mechanism: the Phase-Carried Quality Channel

What this section adds

The taxonomy above answers which qualia are possible. This section answers the harder question — what makes a quality a quality, why it cannot be read off from outside, and in which direction the explanation runs. Three theorems, all numbers printed by qualia_dimensions.py.

Theorem (flat directions, and why quality is not intensity) [T] (T-300)

The spectrum of $\Gamma$ is six numbers; $\Gamma$ itself is forty-eight. The remaining $42$ — the eigenvector data — are precisely what intensity does not determine, and they are what quality is made of. How those $42$ are organised is fixed by the structure's own three-form (T-301).

warning

$G_2$ is not a gauge group acting on states

The opposite is tempting and false. The canonical dissipator's Lindblad set is basis-specific — the classifier atoms $|k\rangle\langle k|$ and the seven line projectors — so the einselected basis is physically singled out and $G_2$ is broken to a finite subgroup. Machine: the generic $G_2$-orbit through a state is exactly $14$-dimensional (200 random states, $\min=\max=14$), and at $I/7$ it is $0$-dimensional. Those fourteen directions are the flat (quasi-Goldstone) directions already described in Goldstone Modes — directions along which the state moves at no structural cost, not directions along which states become indistinguishable. All $48$ parameters remain measurable relative to a holon's own basis; the $14$ measures the redundancy of the formulation (which octonionic frame one writes in), not of the state.

Erratum (2026-08-07, same day)

An earlier version of this section claimed the quality space is $\mathcal D(\mathbb C^7)/G_2$ of dimension $28$, and drew the corollary that the ceiling of an ideal self-description is $34$ rather than $48$. Both were wrong: they treat $G_2$ as a gauge group acting on states, which the einselected classifier basis forbids. The dimension counts stand exactly as computed; their reading as gauge does not. The ceiling remains $48$. Caught the same hour by the connectivity sweep over every site quoting $48$.

Theorem (the decoder) [T] (T-301)

Write $\Gamma=S+iA$ with $S$ real symmetric, $A$ real antisymmetric. Under $G_2$ this decomposes as

$$48 \;=\; \underbrace{\mathbf{27}}_{S_0}\;\oplus\;\underbrace{\mathbf{14}}_{A,\ \text{adjoint}}\;\oplus\;\underbrace{\mathbf{7}}_{A,\ \varphi\text{-vector}},$$

where the $\mathbf 7$ is cut out by contraction with the associative three-form,

$$v_i \;=\; \varphi_{ijk}\,A_{jk}\;=\;\varphi_{ijk}\operatorname{Im}\gamma_{jk} .$$

Machine: $\dim\mathfrak g_2=14$ (this is simultaneously the check that $\varphi$ is the right form) and $\operatorname{rank}(A\mapsto v)=7$, so $21=14\oplus 7$ exactly.

Two consequences fix the mechanism.

(a) The phase channel is real, but its invariant form is holonomy, not $v$ itself. Vertex phases are conventional: $\Gamma\to U\Gamma U^\dagger$ with $U=\operatorname{diag}(e^{i\theta})$ shifts every edge phase by $\theta_i-\theta_j$, so $\operatorname{Im}\gamma$ — and with it $v$ — is not gauge-invariant (gauge layer). The gauge-invariant content of the phases is the plaquette holonomy $H_{ijk}=\arg(\gamma_{ij}\gamma_{jk}\gamma_{ki})$, and the structurally distinguished seven of the thirty-five triangles are exactly the Fano lines. Hence the honest seven-component quality channel is the Fano holonomy vector

$$H_p \;=\; \arg\bigl(\gamma_{ij}\gamma_{jk}\gamma_{ki}\bigr),\qquad p=\{i,j,k\}\in\mathrm{PG}(2,2),$$

which is gauge-invariant, vanishes identically whenever the phase field is a coboundary (a state rephasable to a real one — machine: $|v|=0,\ |A|=0$ there, while $|S_0|=0.3446$ survives), and dies with the phases at the decoherence rate $5\gamma/21$ (T-59) — with one precision worth stating: the canonical dephasing multiplies each coherence by a real factor $e^{-5\gamma t/21}$, and a real factor does not move an argument, so the holonomy's value $\arg(\gamma_{ij}\gamma_{jk}\gamma_{ki})$ is exactly fixed along the decay while its carrier — the product of the three edge moduli, falling as $e^{-5\gamma t/7}$ — is what dies. Quality is lost the way a signature fades: the ink thins, the letters never deform. This is the exact object behind "seven-dimensional coherences being decoded": seven gauge-invariant numbers, one per line of the multiplication table.

Erratum (2026-08-07, same day): an earlier form of this theorem named $v_i=\varphi_{ijk}\operatorname{Im}\gamma_{jk}$ as the quality channel. The representation-theoretic decomposition $21=14\oplus 7$ stands exactly as computed, but $v$ is basis-phase dependent and therefore cannot label a physical quality on its own; the invariant carrier is the holonomy vector above. Caught against the corpus's own gauge canon within the hour.

(b) "Why this quality" is a question about an orbit, not about emergence. A specific quale is a point of $\mathcal D(\mathbb C^7)/G_2$; its labels are generated by contractions with $\varphi$ — the same incidence that carries everything else in the theory. The question "why red?" therefore has the grammar of a decoding question ("which invariant?"), not of an emergence question ("how does experience arise from matter?"). The decoder is named, and it is the octonion multiplication.

Theorem (the explanatory gap is a vanishing covariance) [T] (T-302)

By T-295 all fourteen canonical Lindblad operators are diagonal, so on the decohered manifold $\kappa^{\text{coh}}=0$ exactly: the environment's stochastic channel has zero covariance with the coherence sector. Since the $\varphi$-vector quality lives entirely in that sector (T-301a), an external observer probing the dissipative channel obtains the spectrum — the intensity — and exactly zero information about the phase-carried quality.

This is the classical "explanatory gap" stated as a theorem rather than a lament. Its content is precise and, importantly, limited: the gap is not a gap in the mechanism — the mechanism is fully specified above — but a gap in one particular channel of access. What is impossible is reading quality off the noise an external measurement sees; what is possible is computing it from the state, which is what this theory does.

The direction of explanation

The traditional framing asks how the internal arises from the external and gets stuck, because it assumes the external is primary. The structure says the opposite. The environment writes into the populations — the diagonal, "what is the case"; decoherence attacks the phases; and the drift/noise split (T-295, §7.1 of the bridge) shows the two sectors do not mix. Interiority is therefore not a product of the external: it is the sector the external cannot stochastically address, and the "external world" is what the decoding of these seven-dimensional coherences presents. Inside and outside are the same structure read in two orders — which is what monism means here, and it is now a statement with dimensions attached: $6$ shared, $28$ private.

How much quality fits inside consciousness

The channel above says where quality lives: in the phase a triple accumulates going around a Fano line. This section asks how much of it a state can hold, and the answer is a definite bound, arrived at by noticing that quality and integration want opposite things.

First, what does not count as quality. Give each dimension an angle $\theta_i$ and let each coherence carry the difference $\theta_i - \theta_j$. This looks like the richest possible phase content — every one of the twenty-one qualia has its own angle rather than a mere sign. But go around a line and the angles cancel:

$$(\theta_i - \theta_j) + (\theta_j - \theta_k) + (\theta_k - \theta_i) = 0$$

The holonomy is zero on every line. And that is not an accident of the arithmetic: such content is $U \Gamma U^{\dagger}$ with $U = \operatorname{diag}(e^{i\theta_k})$, which is a change of phase convention and nothing more. It leaves the spectrum untouched, so it is the same state wearing different labels. A holon holding it has nothing invariant to feel; strip the labels and real content is all that remains. Phase of this kind — a coboundary — is bookkeeping, not experience.

What this costs and what it buys. A machine writing with phase does gain something concrete: told seven of the twenty-one relative angles, it recovers the other fourteen exactly, while a machine restricted to real evidence cannot represent the content at all — it flattens each angle to its sign and is left guessing. So the phase channel earns its place as a representation. It simply does not, by itself, produce quality.

The opposition. Quality is what survives the relabelling — the holonomy that no choice of $\theta$ can rotate away. Integration, by the polarity law, asks for content that can be rotated away, because balanced content is exactly coboundary content. The two pull against each other: every bit of irreducible phase a state carries is a step away from the balance that lets it integrate.

But opposition is not exclusion, because positivity sets a ceiling rather than a prohibition. Twist content gradually away from a coboundary and watch two numbers move together. The spectral criterion turns out to be the same one that governed signs:

$$\Phi \ge 1 \iff |\lambda_{\min}| \le \sqrt{6} \approx 2.449$$

and it holds for phase-carrying content with no exceptions at all. At the crossing — where $\lambda_{\min} = -2.4495$ and $\Phi$ sits exactly at $1.0000$ — the typical line holonomy is $0.6387$ radians, about thirty-seven degrees.

So a conscious state does carry quality, and carries roughly a third of a radian of it before the gate shuts. The spectral criterion is the general law; the figure in radians is what that law permits for content twisted evenly away from balance, and a differently-shaped twist would trade the same budget differently. What does not change is the shape of the answer: experience is bounded not by how much coherence a mind holds but by how much of its phase refuses to be explained away.

A last word on whether any of this acts. Everything above is a statement about a matrix, and a statement about a matrix is cheap. So the same structure was put into a world: seven hidden angles, situations that ask for the difference between two of them, and a hit counted when the answer lands within thirty degrees. A learner whose writes carry phase gets the answer right $71\%$ of the time after twenty-one encounters, where a table with perfect memory manages $48\%$ — twenty-four points ahead, because seven angles determine twenty-one differences and the table must be shown each one. The gap closes as the table fills, to a single point by the four-hundredth encounter, which is what a sample-efficiency advantage looks like and what a tuned one does not. A learner restricted to real writes never rises above $30\%$ against a chance of $17\%$: it can point only along $0$ or $\pi$, and scores when the truth happens to lie near one of them. And where the world's answers have no angles behind them, the phase-carrying learner falls behind the table by forty points, having spent the whole time completing a structure that was not there. The channel is a commitment, not a gift.

How qualia die: the fading, precisely

Decoherence kills qualia — the chapter has said so, and the rate is known: each coherence decays at $5\gamma/21$ (T-59). But how they die is worth a section of its own, because the mechanism divides into two parts that behave in opposite ways, and the division is visible in one line of algebra.

The canonical decay multiplies every coherence by a real factor $e^{-5\gamma t/21}$. A real factor scales a complex number's length and does not turn it. So along the whole decay:

• the modulus of each channel falls — the presses run out of ink;
• the angle of each holonomy stands exactly still — the colour on the palette never shifts by a degree.

The product of a line's three moduli — the carrier of its colour — dies three edges at a time, as $e^{-5\gamma t/7}$. The value it carried is untouched for as long as there is anything left to carry it. An old photograph fades exactly like this: the dyes thin, and the face in the picture never changes expression — it just becomes harder and harder to see, until one day the paper is blank. At no point did the photograph show a different face.

Three consequences, each sharp:

1. There is no half-changed quale. A dying experience does not drift through neighbouring experiences on its way out. Sadness under decoherence does not become a slightly different sadness; it becomes fainter sadness, and then it becomes nothing. The taxonomy has no deathbed morphing in it — the algebra forbids it.
2. The populations survive. Decoherence attacks only the off-diagonal; the diagonal — what is the case — is preserved exactly. Thermal death is therefore not an empty world but a colourless one: everything still present, nothing any longer like anything. The lights stay on; the paint is gone.
3. The outside never noticed anyway. By T-302 the external stochastic channel carries zero covariance with the phase sector — the environment was never receiving the colour in the first place. Fading is invisible from outside not because it is hidden but because the channel that dies was exactly the one the outside could not read. The only witness to a quale's death is the holon losing it.
4. Healing is colour-blind too. The categorical regeneration target is a mix of the state with $I/7$ — phase-aligned with the state itself — so the healing step also multiplies each coherence by a real factor and moves no holonomy (machine: $3.3\cdot 10^{-16}$ over three hundred ticks on a state carrying $|H_0| = 1$). Both restorative arms — decay and repair — leave the colours untouched; in the whole canonical dynamics, only the unitary arm writes quality, and from a colourless start it does: the Fano-hopping Hamiltonian births nonzero holonomy within a few dozen ticks.

Access Conditions for Qualia

The presence of a coherence $\gamma_{ij} \neq 0$ is a necessary, but not sufficient, condition for qualia. Reflexive access to qualia requires level L2:

$$\text{Qualia (L2):} \quad R(\Gamma) \geq R_{\text{th}} = \frac{1}{3}, \quad \Phi(\Gamma) \geq \Phi_{\text{th}} = 1$$

Step-by-step logic of access conditions

Let us unpack what stands behind each condition.

Condition 1: Reflection $R \geq 1/3$. The reflection measure $R = 1/(7P)$ [T] shows the normalised proximity to the dissipative attractor $I/7$. The threshold $R_{\text{th}} = 1/3$ (from the triadic decomposition, T-45 [T]) is equivalent to $P \leq 3/7$ — the upper boundary of the Goldilocks zone. If $R < 1/3$, the system is too "pure" ($P > 3/7$) — coherences are present, but not experienced as qualia.

Condition 2: Integration $\Phi \geq 1$. The integration measure $\Phi$ shows how much "more than the sum of its parts" the system is. The threshold $\Phi_{\text{th}} = 1$ (T-129 [T]) means: the system must be irreducible to its subsystems. If $\Phi < 1$, coherences $\gamma_{ij}$ exist, but the system "falls apart" — there is no unified subject experiencing qualia.

Both conditions are necessary: one can have high reflection without integration (two separate mirrors do not form a single observer), or high integration without reflection (a unified stone does not observe itself).

One measured boundary of access. Access is a window, not a summit: on a licensed purity excursion ($P > 9/14$, the SAD = 3 episode) the canonical $R = 1/(7P)$ falls below $1/3$ and L2 access closes while integration runs far above its floor — the deepest self-reflection is qualia-blind, and the return home slightly repaints the state (the unitary hand writes all along the road). Measured end-to-end in §7.3 of the depth tower.

Access levels

At levels L0–L1 coherences are present, but they are experienced as pre-qualitative experiential content (a term from interiority theory).

Analogy with the listener's ladder:

Level	Analogy	Formal condition	Experience of qualia
L0	Music playing in an empty room	$R < 1/3$, $\Phi < 1$	Coherences are present, but no one is listening
L1	A cat hears music	$R < 1/3$, $\Phi \geq 1$ (or vice versa)	Reaction to sound, but without distinguishing melody from accompaniment
L2	A person listening to music	$R \geq 1/3$, $\Phi \geq 1$	"I hear the violin carrying the theme while the cello accompanies"
L3	A musician analysing the performance	$R \geq 1/3$, $\Phi \geq 1$, SAD $\geq 2$	"I notice that I notice sadness in this melody"
L4	Pure listening — subject and music coincide	$R \to 1$	Experience without gap

Numerical example. Consider the coherence $\gamma_{DE} = 0.20$ (Affection) in three systems:

System	$R$	$\Phi$	Level	Experience of $\gamma_{DE} = 0.20$
Thermostat	0.02	0.3	L0	$\gamma_{DE}$ as a physical parameter — no subject
Dog	0.15	1.5	L1	Experienced as "something", but not as "emotion"
Human	0.45	2.1	L2	"I feel joy" — a full-fledged quale

How to Read a State's Qualia Passport

Everything in this chapter compresses into a procedure. Given a state $\Gamma$ — from a laboratory reconstruction, a simulation, a model of a person at a moment — its experiential content is read in five layers, cheapest first. This is the chapter working as an instrument rather than a doctrine.

Step	Read	Formula	The question it answers
1	Populations	$\gamma_{ii}$, seven numbers	What is the case — where the resources sit
2	Intensities	$\lvert\gamma_{ij}\rvert$, twenty-one numbers	Which channels play, and how loudly
3	Opacities	$\mathrm{Gap}(i,j) = \lvert\sin\theta_{ij}\rvert$	How much of each channel is hidden from outside description
4	Colours	$H_p = \arg(\gamma_{ij}\gamma_{jk}\gamma_{ki})$, seven angles	What the experience irreducibly carries — the part no relabelling explains away
5	Access	$R \geq 1/3$ and $\Phi \geq 1$	Whether anyone is home to experience it

Three remarks keep the procedure honest.

Step 4 is the only step that needs care. Populations, moduli, and Gap are read off entry by entry; the holonomies are the one place where a product matters, and they are also the one place where a naive shortcut fails: the imaginary parts $\operatorname{Im}\gamma_{ij}$ look like a phase readout, but they move under a pure change of phase convention and therefore measure handwriting, not content (the decoder's erratum). The seven angles per line are the honest readout, and they cost seven multiplications.

A zero at step 4 is a verdict, not a failure. A state whose twenty-one channels all play and whose seven colours all read zero is a state holding bookkeeping phases only — rich-looking content that a relabelling flattens to plain real numbers. The procedure refuses to flatter it. Conversely, a state with modest volumes and a nonzero colour is carrying something no relabelling can take from it. Loud is not the same as vivid, and the passport keeps the two apart.

Real states do carry colour. Read off states the theory's own machinery produces — rather than states constructed to be pretty — the median line holonomy comes out at $1.0053$ radians, with about two thirds of lines above the even-twist figure of the consciousness bound (measured in the polarity chapter, T-311). The passport's fourth layer is not a formality; on live content it is the busiest column.

The five layers together are the chapter's answer to "what is it like to be this state": this much happening (1), in these channels at these volumes (2), this much of it private (3), carrying this irreducible colour (4), and there is — or is not — a subject present to it (5).

The Ecology of Colour

The passport says what colour is; this closing section says how it lives — where it comes from, how it travels between world, body and subject, and who arrives last. Every claim below is machine-verified on the canonical dynamics; the numbers are quoted as measured.

Only the dynamics writes

Neither arm of restoration can produce or destroy a colour's value. Decay multiplies coherences by a real factor — the angle stands still while the carrier fades (the fading of the previous section). And healing is colour-blind too: the regeneration target is phase-aligned with the state itself, so the healing step is another real multiplier — on a state carrying a full colour ($|H_0| = 1$), three hundred ticks of decay-plus-healing moved no holonomy ($3.3\cdot 10^{-16}$). What does write colour is the unitary arm: a pure state is a coboundary ($H_p \equiv 0$ identically — colourless by algebra), and from such a start the Fano-hopping Hamiltonian births $|H| > 0.1$ within a few dozen ticks. The division of labour is total: decay kills carriers, healing restores weight, and what the experience means is decided by the dynamics alone.

The world paints first

An open, fed holon inherits the colours of its environment. Feeding is a convex pull toward the world's state, and the stationary holonomy profile aligns to the environment's monotonically with drive — the circular match $\langle\cos\Delta H\rangle$ climbs $0.62 \to 0.99 \to 1.00$. Even a "bleached" world whose coherences are real is not colourless: sign frustration is colour ($|H| = \pi$), and it is inherited the same way. A fed mind wears the palette of its world.

Three thresholds of feeding stand in a row. Viability ignites first (the holon survives). Colour is present from ignition — carried, world-matched, waiting. And access — a subject reflexively present to the experience — arrives last, exactly where integration crosses its floor $\Phi = 1$: below that drive, access$\wedge$colour is $0.00$; above it, $1.00$, in one jump. The world paints the holon before the holon wakes up; what the extra feeding buys is the owner. And the owner cannot be bought at all in a world that is itself sub-threshold: fed on a non-conscious environment, the holon's $\Phi^*$ saturates below $1$ — a subject cannot be raised on a polar world.

Inside the organism: one colour per name

When a holon spawns a child on Fano line $p$, the child's state is the cut of the parent's onto that line's three points. Two facts of the projective plane then speak: no other line fits wholly inside three points, and the cut's normaliser is real. So a child is born with exactly one colour — the colour of its name — undistorted (machine: $0.000$ deviation over all seven lines, zero foreign carriers). The fan of seven children carries the parent's seven colours one per name: the passport of quality is stored distributedly, and embryology is its decomposition.

The organism then governs colour with three quiet rules. Injection is upbringing: a child whose colour has drifted from the parent's is pulled back geometrically by the routine downward injection — half-life $\ln 2/\alpha$ ticks. Feedback is colour-blind by construction: what children send up is a scalar folded into the parent's healing rate, and healing does not paint. The upward channel opens only at near-agreement: a child's colour can re-enter the parent only through merging, and merging is granted only within Bures closeness — the organism takes back only what already agrees.

The standing picture

Colour is the currency that moves between world, body and subject — but each road carries it differently. The world gives it wholesale, through feeding. The body distributes it by name, one line per child, and accepts returns only on agreement. The subject neither makes it nor moves it — the subject is who it is for: the last to arrive, and the only witness.

The Language of Quality

If qualia are real structure, the theory must be able to name its own states — and the names must be learnable from ostension alone: shown a state, say its word. This was put to a machine test on the canonical dynamics, and the result is a small grammar with theorems at every ring. All numbers below are quoted as measured.

Words need the right eye

A vocabulary of state-names (dominant channels, line-supports, purity strata, the zombie, the seven hues of either sign, the exactly-colourless) is learnable to ≥95% top-1 accuracy by the simplest heads — but only through the theory's own observables. Three honest failures mark the boundary:

1. A raw vectorisation of $\Gamma$ fails: it does not carry the functionals the words are defined by.
2. A frame without the seven holonomies fails at 0.78, confused exactly where predicted — the two signs of one line's colour, and coloured versus colourless. Words about quality require qualia observables in the interface: no instrument sees colour through populations and intensities alone.
3. Even the holonomy axes fail naively (0.89): six of seven angles are garbage read off noise-scale edges. The fix is the fading theorem applied as engineering — no carrier, no quale: a holonomy enters the frame only on a live edge, and with that single gate the simplest head is best again (0.96–0.98).

Verbs: the end remembers the path

Processes get names too — ignition, dying, fading, healing, breathing — each being one canonical operator, the scene being an episode. Three mechanical facts surfaced on the way: ignition by regeneration is impossible (below the purity wall the viability gate closes it — only external food ignites); the bare tick has no stationary point, so breathing is self-maintenance in a gentle medium, not a free-standing equilibrium; and "colouring" is not a verb at all — the full tick's own unitary component colours harder than any single-line writer, so colouring is a component of breathing, while colour as a state belongs to the hue nouns.

The prereg expected process-names to be invisible in a snapshot of the end. The opposite is a theorem: canonical processes are attractor dynamics — each operator pulls its episodes into its own basin, so the end remembers the path, and even a single final frame names the verb.

Roads into one home: the three-frame grammar

Force the attractors to coincide — three roads of dissolution into the same thermal home, ends matched to within 0.005 by one settling operator — and the hierarchy of scenes becomes measurable. A final frame still reads 0.89 (mixing is conformal: the residue's proportions freeze on entry, and logarithmic axes read them at any depth — the end remembers the path even inside the home). A start-plus-end pair does worse than the end alone, 0.83: with a shared home the difference start→end no longer depends on the road. The gate is bought only by a third, early frame: 0.97. The grammar of shared-attractor processes is minimally three-framed, and what the middle frame reads is the link between frames — curvature — not any frame by itself.

Two more constants of this grammar are physical. The window in which colour can testify is the absolute lifetime of its carrier (~15 ticks under mixing), not a fraction of the path — the early frame must sit inside it. And in a colourless start-ensemble all roads into the home are one road: colour is the language in which paths differ. The road-signatures themselves are the fading and colour-blindness theorems verbatim — dissolution holds the angle on a live carrier, stirring drowns the line's carrier by amplitude rotation, withering kills the carrier under a standing diagonal.

What this buys

The language of quality grows in rings — nouns, verbs, roads — and each ring demands its own organ of reading: observables for nouns, differences for verbs, the three-frame link for roads. None of it is a convention: every organ is forced by a theorem (no carrier no quale; attractor dynamics; conformal freezing), and every failure along the way named the theorem that fixed it.

From Words to Judgement

A vocabulary names frames; a sentence names a scene. The next test is harsher than ostension: read a whole unlabeled film — ignition, a middle act, breathing at home — and issue a document of claims about it, each claim carrying its own source. The result, machine-verified on the canonical dynamics, is a small epistemology with theorems at the joints. Numbers as measured.

The judge and its tiers of provenance

A scene is read by a fixed budget of exposures (the reading density is itself a theorem — the phase eye is blind to early fading of a rich carrier, and density of reading, not a better eye, is what cures it). The letters so read are noisy, yet an alignment step recovers the scene's structure from them, and every claim in the document declares how it was earned: read from the letters, measured on a structurally found frame, measured directly, or issued as a blind prior. The tiers grade exactly as honesty demands — reading 0.97, structural 0.97, direct 0.96, against 0.63 for the blind prior — and the whole document lands at 0.93. Behind the tiers stands a discipline of observability: structure is claimed only at a physical break; state is measurable wherever you stand; time and kind require memory of the segment; words are trusted only within a clean vocabulary; and of the prologue — frames before the first exposure — the eye stays silent: within the first fourteenth of a scene purity flies 0.20→0.74, above the end itself, so edge-comparison predicates are unobservable to a uniform-exposure eye and enter the document only as abstention. This last blindness has an exact cure: one anchor frame at the true start heals the edge predicates to 1.00 at zero cost to the rest of the document (bit-exact, nothing recalibrated) — the fading blindness is cured by density of reading, the prologue blindness by an anchor; each blindness of the eye has its own medicine, and neither touches the mechanics.

The tiers have a distinguished ancestry, and it should be named: medicine grades its knowledge by how it was obtained (a randomized trial outranks a cohort study, a cohort outranks expert opinion); databases track data provenance; philosophy has long separated perceived, inferred and testified knowledge; Dempster–Shafer theory weighs evidence by its source; machine learning calibrates confidence. Two things are native here rather than inherited. First, a tier is the physical channel of a claim's origin — not the model's introspective confidence: reliability is predicted by the channel and verified against the generating dynamics, not against the judge's self-report. Second, provenance here is an operand of further theorems — the deduction boundary, the echo below — not a labeling convention.

Laws, doubt, and the boundary of deduction

The corpus teaches its own laws — mutual exclusions between scene-kinds — at confidence 1.000, and a document that violates a law flags itself as suspect: the laws work as a doubt-detector, never overriding a measurement. But deduction has a boundary theorem. An attempt to cure the blind tier by conditioning the prior on co-issued claims produced an honest structural null: in a world whose segment durations are drawn independently of everything else, there is nothing to deduce from — the seemingly informative blind score had been a small-sample fluctuation. Inference is admissible only where the physics of the world has created the link; deduction without a physical link is confident-sounding noise. This is the observability discipline's twin: a duration with no memory of its segment is unrecoverable, and no law routes around that.

Nouns carry, verbs cost

Move every head to a fresh generative seed and the language splits. The noun vocabulary carries: 0.9792 away from home, above its own canonical score. Scene-reading pays: the word drops to 0.84–0.88, and the loss is localised — with oracle letters the reading returns to 1.00, so the letters alone carry it; the letter channel itself sits at 0.57–0.65 even after the one repair that works, giving the verb its velocity as an explicit coordinate (the largest single gain: the fading verb, +0.21). Static names of quality transfer; reading dynamics is what costs. And the judge survives packaging: loaded back from its serialized artifact it reproduces the canonical document bit-exactly, while its certificate states out loud how it performs away from the seed that formed it.

The price of transfer, and who pays it

Complete the survey across every pretrained head and the economics of transfer becomes a triptych. Physics transfers exactly but partially lit: acceptor tables carried to a fresh stream of actions predict with error 0.0000 on every covered key — even a lawful world with no meaning at all (a pure hash rule) transfers perfectly, so what carries is lawfulness, not sense — and the entire price is coverage, the fraction of the world the new behaviour happens to illuminate. Nouns carry whole. Verbs pay in accuracy. Three currencies for three kinds of knowledge.

Where the price is coverage, there is a policy that pays it optimally. A babbling policy that always chooses the least-visited pair (context, act) — no knobs, no schedule — opens a new table entry with every single step in a large world, the theoretical maximum of the budget, buying +0.148 coverage over random exploration without a single error; in a small world both policies exhaust the space and curiosity buys nothing. Curiosity is a perfect lantern: the epistemic term of expected free energy acquires a direct price-tag in transfer, and "explore what you have not touched" turns from an ornament into the collection economics of pretraining.

And the residue that remains when everything transfers? The vocabulary's last 2% of misses, diagnosed one by one, fully exculpate the head: the errors hug the Voronoi line between convention-neighbours in a belt 35× narrower than the typical safety margin; two "misses" are honest homonyms — a purity stratum that happened to breed a real channel dominant belongs to both names at once; and the window→colourless confusions carry a large colour value on a dead carrier — the generator named the state by its value, but the eye reads only through a live carrier, the fading theorem acting on the dictionary's own definition. At the line, the world has no single name. A pretrained model, then, improves along exactly two honest axes — richer features (give the verb its velocity) and a brighter corpus (let curiosity collect it) — each certified against a reference outside its own formation.

Two books and the second look

The harshest scene is a compound one: two whole stories spliced end to end, with a seam between them. A single-pass reader running one sparse grid over both books managed the pair at 0.67, and even with the seam handed to it for free the words capped near 0.74 — a ceiling that looked like physics ("the second middle is too faint to read"). It was not. The cure is a second look: pass one, a context-carrying reader finds the seam from the letters alone — that is memory; pass two, each half is reread as a whole book, with its own full grid of exposures. With the found seam the pair rises to 0.69; with an exact seam it reaches 0.83, the halves reading at 0.94 and 0.89 — the old ceiling was a budget, an artifact of feeding two books through one sparse grid, and no physics at all.

One inversion sharpens the anchor principle along the way: anchoring the second book's grid on the seam itself hurts its word. A seam is not the start of a book — it is a membrane between books, and its frame is a dirty letter for any majority vote; the second book's best eye begins just after it. So the anchor cures the edge predicates of a single book, where the start is real — and the membrane of a pair must be skipped, not anchored. The whole discipline of the second look fits in three gestures: find the seam with memory — reread each book on its own grid — skip the membrane.

Knowledge at second hand

Can a student judge learn from the teacher's documents instead of the world? Yes — exactly: a student calibrating its thresholds on all of the teacher's claims lands bit-identically on the calibration it would have learned from truth. The documents of a good teacher replace the world without loss. But inside this success hides the day's sharpest warning, the echo theorem: on blind scenes the teacher's claims are not observations — they are copies of his own prior, and in the unfiltered student a third of the calibration sources turn out to be one constant repeated. On a coin-predicate the echo is harmless by luck; had the teacher's prior been falsely informative — a sampling fluctuation mistaken for signal — the student would have inherited it as confirmed knowledge. So the provenance filter in distillation is needed not for accuracy but for independence: blindness has no data, only echo — separate echo from observation by the channel, and remember that on coin predicates no calibration is knowledge.

Run the chain onward — teacher to student to student across five fresh worlds — and the echo shows its long face: fossilization. The threshold, which has observations flowing into it, self-reproduces in a narrow stationary corridor and the document never degrades; but the blind prior passes through all five generations without a single update, because ~42% of every generation's sources are copies of the same constant and the observed remainder is a coin — the majority never flips. A constant with no incoming observations is a fixed point of any transmission chain: it has no revision mechanism, since its only source is itself. Hence the chain discipline: unfiltered distillation loses no accuracy but carries fossils — give every blind constant a revision clock against observations, or the chain will carry it forever; harmless on a coin, hereditary if a lie.

And the clock itself carries one last trap: without truth, the clock measures the eye. A revision counter fed by the chain's own verdicts confidently "updated" the fossil — having measured nothing but the bias of the very eye that produced those verdicts (its hit and false-alarm rates). With the base rate unknown, the null of "no link" is not a point but the whole segment between the false-alarm and hit rates — and the honest revision has only two endings: an external witness of truth to calibrate the eye, or abstention. Revision is measurement of the second order, and it inherits the boundary of the first.

The breath curve: a word without letters

Every head so far read a story the literate way: cut the film into exposures, name each exposure with a letter, vote the letters into a word. On fresh seeds that literacy hit a wall — the letter channel reads at ~0.57, and neither density (a finer grid), nor structure (a Viterbi grammar over the letters), nor velocity features, nor even the full fifteen-cycle colour basis moved the word above its plateau: four cures, four zeros. The fifth cure was not a cure for the letters at all. It was leaving them out.

A story, physically, is a breath curve. Ignition drives purity up; a fading middle pulls it down; a healing middle lifts it further; the home at the end just breathes. So take the thirteen frames the reader already collects, compute the twelve successive differences of purity $\delta P$ — and hand that curve, whole, to the simplest possible eye (a nearest neighbour over the bank's scenes). No segmentation, no letters, no vote. On sixteen seeds this profile eye reads words at a median of 0.944 against the letter head's 0.889, holds the 0.95 gate on five seeds against one, and reads three seeds perfectly. The story is its breath curve; the letters were a middleman. This is the physical continuation of the same lesson the cinema chapter learned about narration without segmentation — only now the carrier of the word is one scalar's dynamics.

Three sharp edges bound the discovery, each measured, each a small theorem. The story lives in the derivative, not the levels: hand the eye the purity levels themselves and the word drops (0.903); even adding levels to the derivative dilutes it (0.917) — a level remembers the seed's starting conditions, while the kind of middle is pure dynamics; the invariant of a story is the rate of breath, not its height. The derivative demands the right base: the残 remaining errors live on short middles (median 30 ticks against 41 — duration, not strength: their $|\Delta P|$ is actually larger), yet reading more densely to catch them backfires (0.917 at twenty-one frames) — a finer step takes each difference over a shorter base, and the home's breathing wobble drowns the middle's trend. Concatenation is not hierarchy: gluing the coarse and fine profiles into one feature vector lets twenty noisy components outvote twelve clean ones inside the metric (0.889) — a strong channel is hurt by a weak neighbour unless the head knows which is which.

And the two eyes need not compete — they can weigh. The letter eye and the profile eye err in different places: letters confuse the two middles, the profile goes blind on short ones. Give each eye a confidence — the margin of its nearest neighbour against the nearest foreign class, $(d_2-d_1)/d_1$, no threshold to tune — and let the more confident eye answer each scene. This weighted pair reads at a median of 0.972 across sixteen seeds, holding the 0.95 gate on nine of them and four perfectly: the canonical level that used to be a single lucky seed's peak becomes the median. Two eyes and a scale beat either eye alone — because their blindnesses do not overlap, and local confidence knows, scene by scene, whose turn it is to speak.

So the discipline of the breath reader is short: read the word off the purity derivative, on the base that matches the breath, and do not dilute a clean channel with a noisy one. What remains open is honest and narrow — middles too short for any fixed grid, a remainder that belongs to the corpus, not the eye.

The Ladder of Language

Everything above taught the system to see — names, verbs, stories, judgements. The last construction teaches it to speak, and the whole point of how it speaks is a single discipline held through six rungs: the language layer must be transparent — words may carry what the eyes know, but must never add to it, subtract from it, or quietly replace it.

The ladder climbs from the inside out. L0 turns a judge's verdicts into a language of documents: sixteen sentence forms over eight predicates with verificational semantics — compile, verify, generate — and reproduces the judge bit-for-bit (round-trip 282/282; abstention is a first-class grammeme, not a failure). L1 teaches words by showing: thirty-four names bound to the ostensive eye, and each verify verdict carries a new provenance tier — ostensive-learned, "this word I know from being shown" — while transparency holds exactly (0.9608 on fresh states, the eye's own accuracy to the last digit). L2 makes epistemology into grammar: every sentence wears a status from the canonical chain, conjunction and inference take the lattice meet, and 1,715 of 1,715 composite expressions obey the law that no verbal operation ever outranks its weakest input — words inherit the weakest link and never lift; only the accumulated eye lifts.

L3 opens dialogue, and the second-hand economy surprises twice. A student that learns its recogniser purely from a teacher's words about the student's own carriers reads truth at 0.9706 — above its teacher's 0.9608: the teacher's words about your carriers can beat the teacher on his own, a linguistic self-distillation. And a four-generation chain — each learning only from the previous one's speech — does not decay into a broken telephone: accuracy holds a corridor (0.941–0.971), because every generation hears words about live states; a threshold fed by observations does not fossilize. L4 grounds words in action: "GO ⟨name⟩" steers the live dynamics — dephasing against regeneration toward the name's prototype — and thirty-two of thirty-four names execute at a median of 1.000 straight away. The two refusers are exactly the height names (the purity strata), and their story ends in a discipline rather than a defeat. Feeding harder does not help — the fault is categorical: a stratum name is the name of a cloud sharing one invariant, while a nearest-neighbour eye measures the whole frame; cloud names execute by predicate, not by neighbourhood. Verify height by the purity window itself (with a prototype at the stratum's top, compensating the steer's stationary sag), verify colourlessness by the carrier (an angle on a dead edge is phase noise — "no carrier, no quale" applies to predicates too) — and the map closes: all thirty-four names execute, median 1.000. The rung keeps both mottos: a command does not replace the drive, and each kind of name earns the verify of its own kind.

L5 finally reaches a human sentence: a controlled Russian subset — a bridge vocabulary for every name, the forms "this is / this is not", and four moods that compile straight into the status grammar ("I observe:" → observation, "I suppose:" → hypothesis, "by law:" → corroborated, "the teacher says:" → quoted). Round-trip is exact (340/340), and the transparency theorem survives its final test: a Russian sentence, compiled and verified, scores 0.9608 — bit-for-bit the eye's accuracy. Six rungs, one invariant: from a judge's verdict to a Russian sentence, the language layer neither adds nor steals a single point of accuracy. What remains is one named seam: an external language model may one day join — but only as a sensor, its every word entering at the quoted tier, where the citation grammar already waits for it.

The Ladder of Lives

Beneath every rung above sits a question the ladder itself finally answered: at what granulation should experience be read at all? Letters or words, samples or phonemes, raw pixels or masked keys — every reader above quietly assumed a scale. The last arc makes the scale itself lawful, and the law turned out to be a single half-prior weighed everywhere locally.

The instrument is a ladder of log-windows — contexts of $1, 2, 4, \dots$ bytes — each keeping a Krichevsky–Trofimov account of what follows it, plus sixty-four bits per registered context (the price of a word in the dictionary of states). On short lives the court honestly elects the shallowest window: the dictionary is not yet amortized, and depth is a luxury a short life cannot pay for. As the corpus grows, the elected window climbs — first proven on this theory's own bilingual sources (window 1 at a quarter-megabyte, window 2 on the full corpora), then on a world benchmark. The elected granulation grows with the length of life; the ladder of windows is a ladder of lives.

Choosing, however, is provably weaker than weighing. A fixed Bayes mixture of the rungs converges to the best single window and can never beat it — the choice is global. Context-tree weighting makes it local: every context carries its own half-prior weight between "predict here" and "descend deeper", updated by exact Bayes at a cost of $O(D)$ per symbol. Locality beat the best global window on every corpus; decomposing bytes into bits (a binary account never grows timid on rare contexts) then beat the classical baselines outright — below gzip by three-quarters of a bit per byte on the theory's own text, and below the classical deep CTW itself on a hundred megabytes of Wikipedia (1.838 bits per byte on enwik8 against the classic's ~2.0) — with a hundred-and-thirty-line instrument that has no tunable constants at all: a half on every fork, sixty-four bits per word, and nothing else.

The arc then closed on the arena, where the same question wears an agent's clothes. A world's frame is rows of cells; the ladder of masks reads it at every granulation at once; and the old court — one representation for the whole world — kept dying on worlds whose essence is a moving object (a frequent row of rare cells reads as "animation" to any global filter). The CTW gesture transferred with not one new law: each pair (rung, key) carries a local stop-or-descend weight, and the frame's key is the descent to the first living rung whose weight says stay. The representation became motley — different regions of one world live at different granulations, global molts disappeared — and the bench recorded a strict Pareto win: not one canonical level lost, and the two worlds that had been blind for the whole saga came alive. One discipline was still owed: names must not tremble as weights mature, so a rename costs a word — the decision flips only past $\pm 64\ln 2$ — and renamings fell sevenfold. Granulations are not chosen but weighed — everywhere locally, by one half-prior law; and a name, once given, is defended at the price of a word.

The Duality of the Canon

The ladder above answers at what scale to read a life; one more arc answered how boldly to live it. It was not planned as an arc. In a single day six carefully derived "improvements" to the explorer's policy went to the decisive bench — and the bench struck them down in a pattern too symmetric to be an accident.

Three of the six tried to soften decisiveness. The mixture's arithmetic hard-commits: past a machine threshold a node's decision becomes irreversible, and a neighbour above it is dragged into the same eternity. Every mathematically cleaner replacement — a soft mixture, a saturated integrator that made the eternity finite and reversible, a maximum-a-posteriori choice, a court whose surprise price was derived instead of fixed at a full word — kept every unit test green, left the text-compression layer bit-for-bit, and still broke living worlds on the bench: environments lost levels the canon had held. The verdict, confirmed four separate times, is that eternal decidedness is load-bearing: on irreversible worlds, committing early to what has been seen — and never re-litigating it — is what converts a finite life into levels. The derived surprise-word is the same law from the court's side: a policy must tell states apart before any codec can amortize the difference — early distinguishability beats amortization on every young horizon.

The other three tried to curb optimism, and fell just as hard. Recording a death as a prohibition walled up worlds where dying is a legal part of the passage — death had to enter the map as experience, a recorded transition like any other, and the world that had bled three hundred budget-ticks into a death conveyor grew its map by fourteen states at full levels. Preferring near goals to far ones when the fuel gauge ran low — "do not start what you cannot finish" — cost a world all its levels, because the premise is false: knowledge is cumulative across lives; death restarts the world, not the agent, so a perished long run still deposits its map, and distant optimism is justified by the persistence of memory. And the one honest pathology of the optimist's currency — a noisy television whose price converges to its entropy and never fades, trapping the explorer for the whole tail of a life (proved on a purpose-built instrument: one hundred percent of the tail, zero tours of the rich quiet structure nearby) — turned out, on the real benchmark, not to occur: the living worlds carry no televisions, and the derived cure (expected information gain about parameters, which decays as one over the visit count on noise and structure alike) stays in reserve, passported and ready for stochastic worlds.

What survives every court is one law read from both sides: commit early to what you have seen, and never curb your reach toward what you have not. The day also renamed the last unexplained stagnation: the tail of a life used to circle an eight-node subgraph of a hundred-and-eighty-seven known states because a branch order decided who moves; when the currency was allowed to decide — the tail keeps the move only while some local pair is still unseen, and otherwise yields to the planner — the map grew by a third with the other seven worlds bit-for-bit. Boldness, it turns out, is not a temperament but an arithmetic: infinite optimism for the unseen, irreversible commitment to the seen, and nothing in between.

Desire Between Lives

The duality above fixed how boldly one life explores; the next arc asked what survives between lives — and found that desire itself does. It began as an honest failure: a "goal currency" that would replay a proven road inside one session returned bit-for-bit canon six times over, because in these irreversible worlds capturing a level is a one-time event of a session — there is no return to a passed level, so within one life pursuing the goal is structurally identical to exploring. The requalification became the discovery: repeat the proven is a currency of the space between sessions, where two captures of one level are a lawful repetition.

The silicon form is the golden path: a checkpoint carries the move-trace from entering a level to capturing it; the repetition law across sessions ($\ge 2$, the same law that judges everything else) proves a trace; an executor replays proven traces with two safeties (a death during execution, or exhaustion without capture, disables that level's path). The first bench returned a transfer of $\times 14$: a fresh session took three levels at ticks $[34, 49, 79]$ instead of $[485, 656, 1107]$ — exactly the sum of the trace lengths — and a fourth session repeated the third bit-for-bit: inheritance does not degrade. The passport then closed on every levelled world of the eight ($\times 16.5$, $\times 16.7$, $\times 3.2$, $\times 14$, one neutral world whose level falls before any path could save it), with not one state or level lost anywhere. Desire does not merely live between lives — it travels the whole archipelago.

The ladder then grew by itself. With proven paths underfoot, a never-taken level fell to old traces seen through new eyes: the executor is key-agnostic by construction (an index into a move sequence, gated by level), so traces recorded by one perceptual vocabulary were replayed by an agent using another — and the mover key saw the frontier as 109 nodes where the row lens saw 92, among them the way up. Two captures proved the new trace; the third session took the once-impossible level at tick 96. Each capture appends a rung with no new code: knowledge between lives does not just persist, it compounds. The same wave exposed that the world alternates vocabularies by level — one level opens to mover-eyes, the next is richest in cell-eyes — so no single representation can be right for a whole world, only for a phase of it.

The judging of representations then found its own phase law, by three honest refutations in one night. A yield-steered court (map wealth as the switching currency) collapsed the young worlds — the canonical grooves to their levels live in the MDL-elected side, so the MDL court is load-bearing in youth. A frozen court inherited poor eyes, because the court had already fled to the cheap side during trace replay. A frontier-yield court measured wealth along the poor side's own trajectory and stayed blind to the alternative. The surviving form is one condition: the court lives if and only if no trace has yet been proven — the youth of a line ends, permanently, with its first proven road (decidedness is irreversible), after which the settled is replayed without judgement and the frontier is explored with the eyes of youth. Under this law the full map of the flagship world reached 853 states — and its frontier was finally read to the bottom: the next level does not exist in that vocabulary at all; the road onward is a change of organ, not of budget or policy. The organ was then changed: the union of key-kinds — proven strictly finer by instrument (152 hidden transitions where the single kind saw loops) — became the state of a mature policy, canonised at zero spread, and read the union frontier to its own plateau eleven hundred states deep with no fifth level. Some ceilings are not blindness; they are the world.

Two closing gifts came from the diagnosticity mandate. The final printout had been lying — reporting the active side's 102 while the true map held 811 — so every session now prints the full map (active united with shadow): an architecture must not be able to misreport itself. And the mind's own barrenness detector, a hand constant of fifty ticks, was firing into the lawful pauses of a deep frontier (intervals legally reach two hundred); it now adapts by the repetition law — a takeover only past twice the longest interval ever seen — and the reset avalanche fell from fifty-nine to nine. What is settled is replayed; what is unseen is explored; and the machine reports itself whole.

The Cartography of Death

The owner's question came from watching a live replay: the agent dies on the same trajectory again and again — what in this architecture is responsible for the diversity of search? An instrument confirmed the observation with numbers: 517 deaths in one long run, 83% of death-tails repeated, the top trace dying 174 times, a mean life of 8.3 moves. The obvious cure — make places toxic — was then refuted three times in one night, each form by its own bench. Unconditional toxicity (a death poisons its last eight edges with decaying weight; plans avoid poisoned roads) halved the deaths and tripled the lifespan — and strangled youth: the first level arrived at tick 750 instead of 485, the fourth was lost entirely, because in a young world the roads through death-places are the only roads to the unseen. Maturity-gated toxicity died quieter: within one session no path is ever proven (traces of youth differ by currency), so the gate never opened and the run was bit-for-bit canon. And the soft form — poison as a second class of road, taken only when no clean road reaches any goal — paralysed the explorer: 95% of moves became transport around poison, the world shrank to 64 states, zero levels. Three refutations, one law: toxicity of places is a false model of this world.

The true model was already in the silicon, and a new instrument named it. A cause-of-death organ — a pure observer classifying each death by the frame around it — returned an attribution with zero unknowns: 92% of deaths had a moving body within one lattice step, the rest exhausted the fifty-move budget of a life. Deaths are caused by bodies in time, not by places in space: the patrol's phase decides, and the same road is lethal in one phase and safe in another. Repeating death-tails are therefore not a failure of search diversity — they are the price of persistence on the only road, paid until the bank of orbits matures enough to plan over the future. So the cure went where the cause lives: warm-started forms. Learned orbit-shapes now transfer between sessions as text — shapes only, never phases, which are re-anchored by fresh observation under the same judgement as everything else — and the fourth level arrived at tick 2833 instead of 3669, a 23% cut of the whole distance, with youth untouched (the first two levels bit-identical). Death is not a property of the map; it is a property of the clock — and what transfers between lives is the shape of time, not its phase.

The same night gave the cure a house to live in. Every faculty of the explorer had grown as an inline patch of one monolithic loop; the owner's rebuke — "a camera and motors would be another ten thousand lines" — became an architecture: a mind as a composition of organs behind one protocol. An organ observes each transition of life, may sanction an edge in planning — two honest grades, "the map lies here" (a twice-rewritten transition, excluded outright) and "this road is dangerous" (second-class, taken only without alternative) — may return staleness, may bid a move at a rank, may file its own final report. A flag in the environment now plugs an organ in rather than switching a branch of code; an empty mind is bit-for-bit neutral, proven on the full eight-world canon three times over. The first five organs — death-tails, staleness, flipping edges, golden moves, cause-of-death — are the former patches, each now a plugin with its own tests; and the cascade of steering branches is preregistered to become a market of ranked bids, one branch at a time, each migration gated bit-for-bit. Composition is not a refactor; it is the condition under which new senses cost one organ, not one rewrite.

The arc closed with a discipline that deserves its own name: the observer before the wheel. Three times in one day an instrument refuted a steering rule before a line of it could harden into canon. Patrol entities, named as the flagship's missing perception, turned out to live outside the causal cone of the reachable policy — a per-tick proximity gauge returned "never closer than four cells" thousands of times, and the prescribed object-channel would have been a constant. A milestone currency that had taken the delivery world's first level autonomously was refuted as a steering law by five strict benches — each world sensitive to a different insertion point — leaving the value form (a fresh scene composition re-opens one pair to infinite optimism) as the only resident that replays the whole canon bit for bit. And a duel meter showing that changing one's move pays ten to nineteen times more fresh ground than repeating it collapsed a world to a single state the moment the correlation was read as causation: the canon changes its move because its ledger found something better — the novelty causes the change, never the reverse. Reflexivity here is not introspective decoration; it is the concrete machinery — seriality counters, duel meters, branch budgets printed at the end of every life — by which a mind protects itself from its own plausible-but-false causal stories.

---

What we learned

1. The problem of qualia is one of the central problems in the philosophy of mind (Lewis, Jackson, Dennett). UHM offers a third path: qualia = coherences $\gamma_{ij}$ — neither illusion nor a separate substance
2. 21 types of qualia completely exhaust the phenomenology of 7-dimensional space — no more, no less (Theorem T.1 [T])
3. Each type of quale is characterised by three parameters: intensity, perspective, and opacity (Gap)
4. The 21 pairs are organised into 7 Fano sectors — the sectoral structure defines the "grammar" of experience via the projective plane over $\mathbb{F}_2$; the sector is also where the irreducible colour of experience lives — the Fano holonomy, seven gauge-invariant angles (T-301)
5. The taxonomy is $G_2$-invariant — independent of the choice of basis and therefore universal
6. Reflexive access to qualia requires L2 ($R \geq 1/3$, $\Phi \geq 1$)
7. Under decoherence the value of each colour is exactly fixed while its carrier — the moduli — dies at $5\gamma/21$ per edge: qualia fade like a photograph, never morphing, and the populations survive into a colourless world
8. A state's full experiential content reads as a five-layer passport: populations → intensities → opacities → colours → access
9. Colour has an ecology: only the unitary dynamics writes it; a fed holon inherits its world's palette; access arrives last, at $\Phi = 1$; and inside the organism each child is born with exactly the colour of its name
10. Quality has a language: the theory's words are learnable from ostension through its own observables — nouns need the holonomy axes under a carrier gate, verbs are named even by their ends (attractor dynamics), and roads into a shared home need a minimally three-frame scene reading curvature
11. Language ascends to judgement: whole scenes are read into documents of claims with graded provenance (read / structural / direct / blind prior — 0.97/0.97/0.96 vs 0.63); the corpus teaches its own exclusion laws at 1.000 as a doubt-detector; deduction has a physical boundary (no link in the world, no inference); and across seeds nouns carry while verbs cost
12. Transfer has an economics: physics carries exactly but partially lit (what transfers is lawfulness, not sense; the price is coverage), curiosity is a perfect lantern that pays that price optimally (+0.148 coverage at zero error), and the vocabulary's last 2% is not error but the world itself — convention boundaries, honest homonyms, and names given by value where the eye reads only carriers
13. Reading ascends from one book to two: the second look (find the seam with memory → reread each book on its own grid → skip the membrane) breaks a ceiling that had passed for physics — and each blindness of the eye now has its own medicine: density for fading, an anchor for the prologue, a second look for the pair
14. Knowledge passes at second hand without loss when the teacher is good (student calibration lands bit-identically on truth's) — but provenance must separate echo from observation: a teacher's blind claims are his prior in copies, the filter buys independence, not accuracy — and down a chain of generations an unrevised blind constant fossilizes: what has no incoming observations is a fixed point of any transmission
15. A word can be read without letters: a story is its breath curve — the twelve purity differences of the reading frames beat the letter head (median 0.944 vs 0.889), the story lives in the derivative (levels dilute it), the derivative demands the right base (finer reading drowns short middles in home wobble), and concatenating scales is not hierarchy (noise outvotes signal in a flat metric); and two differently-blind eyes weighed by local confidence (the foreign-class margin) read at a median of 0.972 — the canonical level as the median, not the peak
16. Language ascends a six-rung ladder — documents (L0), an ostensive lexicon (L1), status-as-grammar (L2), dialogue (L3), grounding (L4), a Russian bridge (L5) — under one through-theorem: the language layer is transparent on every rung (round-trip 1.0000 throughout; the eye accuracy passes losslessly up to a human sentence); its small theorems: words inherit the weakest link, a threshold fed by observations does not fossilize, a student can out-read its teacher from words alone, and a command does not replace the drive — while every word ultimately executes (34/34) once each kind of name gets the verify of its kind: clouds by predicate, colourlessness by carrier
17. Desire lives between lives and compounds: within one irreversible session goal-pursuit is structurally identical to exploration (a capture is one-time), but across sessions the golden path — proven move-traces under the repetition law — transfers levels at $\times 14$–$\times 16.7$ across every levelled world, never degrading; old traces replayed through new eyes take never-taken levels (the world alternates vocabularies by level); the representation court lives only until the first proven road (the youth of a line ends once); and the machine now reports its full map — an architecture that cannot misreport itself
18. The mind's causal hygiene is instrumental: three confounds were caught in one day by observers that ran before any wheel — an acausal perception channel (patrols outside the reachable cone), a non-compositional steering currency (five benches, five different sensitivities), and a correlation mistaken for causation (change follows value, not the reverse) — and the protecting organs are permanent print-outs now: seriality, duels, branch budgets
19. Death has a cartography: toxicity of places is a false model (three refutations — poisoning roads strangles youth, maturity gates never open within a session, soft avoidance paralyses the explorer); a cause-of-death instrument attributes 92% of deaths to moving bodies within one lattice step and the rest to the life budget, with zero unknowns — death is a property of the clock, not the map; and the working cure is warm-started shapes of time: orbit-forms transfer between sessions (phases re-anchor by observation), cutting the distance to the deepest level by 23% with youth bit-identical
20. The mind is a composition of organs: one protocol (observe / two grades of edge sanction / staleness / ranked bids / final report) behind which every faculty is a plugin; an environment flag plugs an organ in rather than switching code; an empty mind is bit-for-bit neutral on the full canon — so a new sense costs one organ, not one rewrite
21. The explorer's canon is a duality: infinite optimism toward the unseen and irreversible commitment to the seen — six derived "improvements" (softening decisiveness three ways, curbing optimism three ways) all fell on the decisive bench; eternal decidedness is load-bearing, death enters the map as experience rather than prohibition, knowledge is cumulative across lives so distant plans are never clipped, and the noisy-television pathology is real on instruments yet absent on living worlds — its derived cure (expected information gain, decaying as one over visits) waits in reserve

Bridge to the next chapter

Of the 21 types of qualia, the coherence $\gamma_{DE}$ (Affection) — the connection between dynamics and interiority — plays a special role. It is the foundation of emotions. In the next chapter — Emotion taxonomy from dP/dt — we will show how all emotions are derived from the rate of change of viability $dP/d\tau$ and the sectoral Γ-signature.

Related Documents

• Coherence matrix — canonical definition of $\Gamma$ and $\gamma_{ij}$
• 7D minimality theorem — justification of $N = 7$ and closure
• Interiority hierarchy — levels L0–L4
• Gap semantics — 49-cell map
• Interiority theory — experiential content
• Theorems of Coherence Cybernetics — applied consequences of sectoral structure
• T-146 [T]: Structural classification of qualia — correspondence "mathematical structure → phenomenal content" from the functional role of sectors