Gap

Formal definition

A Gap is a triple $(a, \sigma_t(a), \Delta_t(a))$:

$$\mathrm{Gap}_t(a) = (a,\; \sigma_t(a),\; \Delta_t(a)) \quad a \in H_t$$

where:

• $a \in H_t$ is the element — the fiber element being assessed
• $\sigma_t(a) \in \mathrm{Fix}(\sigma_t)$ is the saturation-closure — the smallest $\sigma_t$-fixed element $\geq a$; how far $a$ must move to be meaning-settled
• $\Delta_t(a) \in \mathrm{Fix}(\Delta_t)$ is the transfer-closure — the smallest $\Delta_t$-fixed element $\geq a$; how far $a$ must move to be execution-settled

Zero-gap condition. $a \in H^*_t$ iff $\sigma_t(a) = a$ and $\Delta_t(a) = a$ — both gaps are zero. The element is at its own closures; neither nucleus demands more. Full settlement is the coincidence of both gaps being zero simultaneously.

Gap as a quality. $\mathrm{Gap}_t : H_t \to H_t \times H_t$ is a Quality on fiber elements: the quality-map assigns to each $a$ the pair of its closures. Because both nuclei are natural transformations, the gap assignment is compatible with restriction maps — it is a genuine quality in the presheaf sense.

Partial gaps. The saturation gap $\sigma_t(a) / a$ (the closure modulo the element itself, in the Heyting algebra sense) measures how much saturation is owed. The transfer gap $\Delta_t(a) / a$ measures how much transfer is owed. An element can have zero saturation gap but positive transfer gap ($a \in \mathrm{Fix}(\sigma_t) \setminus \mathrm{Fix}(\Delta_t)$): meaning-settled but not execution-settled — endorsed but not enacted. This is the formal structure of akrasia.

What this is

The Gap is the operational diagnostic of incompleteness. In the running system, every element in $H_t$ has a gap. The gap tells you:

1. What the element owes to $\sigma_t$: the difference between what it claims and what meaning-settlement demands
2. What the element owes to $\Delta_t$: the difference between what it claims and what execution-settlement demands
3. Whether the element is at fixed point: both gaps zero iff $a \in H^*_t$

The gap is not a failure — it is a position. Every non-settled element has a gap; that gap is the productive tension (the Tension) that drives the element toward $H^*_t$. Without gap, there is nothing to settle.

Gap and telos

The Teleology of an element $a$ is its telos $\pi_t(a) = \min\{b \in H^*_t : a \leq b\}$ — the smallest fixed-fiber element above $a$. The telos IS the zero-gap point toward which $a$ is headed.

By the commutation axiom (an axiom of the relational history fiber), $\sigma_t \circ \Delta_t = \Delta_t \circ \sigma_t$. Therefore the combined closure is computable along either path:

$$\pi_t(a) \;=\; \sigma_t(\Delta_t(a)) \;=\; \Delta_t(\sigma_t(a))$$

This equality means two things. First: both computation paths agree — saturating then transferring gives the same result as transferring then saturating. Second: the intermediate elements $\Delta_t(a)$ and $\sigma_t(a)$ are both one nucleus away from $\pi_t(a)$. For elements with only one nontrivial gap, the commutation axiom forces the projection gap to collapse to a single interval:

• If $a \in \mathrm{SatShadow}_t$ (saturation gap is trivial: $\sigma_t(a) = a$), then $\pi_t(a) = \sigma_t(\Delta_t(a)) = \Delta_t(a)$, because $\sigma_t(\Delta_t(a)) = \Delta_t(\sigma_t(a)) = \Delta_t(a)$. The projection gap $[a, \pi_t(a)]$ equals the transfer gap $[a, \Delta_t(a)]$.
• If $a \in \mathrm{TrShadow}_t$ (transfer gap is trivial: $\Delta_t(a) = a$), then $\pi_t(a) = \Delta_t(\sigma_t(a)) = \sigma_t(a)$ by the same argument. The projection gap equals the saturation gap.
• If $a \in \mathrm{FreeShadow}_t$ (both gaps nontrivial), then both intermediate elements $\Delta_t(a)$ and $\sigma_t(a)$ are strictly between $a$ and $\pi_t(a)$, and neither equals the other in general. The projection gap is genuinely two-dimensional.

Gap and friction

Friction in the history fiber is the failure of restriction maps to preserve settled elements: an element $I \in H^*_{t_0}$ may restrict to $\rho^*(I) \notin H^*_{t_n}$ — the gap at $t_n$ is nonzero even though the gap at $t_0$ was zero. Friction is the introduction of gap by restriction. A system with no friction would have gap-preserving restriction maps: elements that are at zero gap at the origin remain at zero gap throughout the restriction.

Gap and discharge

Discharge of a duty $d$ is the event of reaching zero gap: $\sigma_t(a) = a$ and $\Delta_t(a) = a$ for the performance-act $a$ associated with $d$. Before discharge, the performance has positive gap — it owes more to one or both nuclei. The discharge event is the transition $\mathrm{Gap}_t(a) > (a, a) \to \mathrm{Gap}_{t'}(a) = (a, a)$ — the gap collapses to zero at some history $t' \geq t$.

Gap as measure of institutional compliance

In a NormativeSystem, the compliance of an agent’s output is the gap of that output relative to $H^*_t$. A fully compliant output has zero gap. An obligation $O(\varphi)$ is discharged iff the agent’s performance of $\varphi$ has zero gap at the required history. The norm system is satisfied when all obligated outputs are at zero gap.

The gap as spectral eigenvalue

The two gap components — saturation gap and transfer gap — tell you which nuclei don’t yet fix $a$. But neither by itself tells you how far $a$ is from the fixed fiber. The answer to “how far” is a single number: the projection gap size

$$\lambda_a \;:=\; \mathrm{RelationalHistoryProjectionGapSize}(a) \;=\; |[a,\, \pi_t(a)]| - 1 \;\geq\; 0$$

(RelationalHistoryProjectionGapSize), counting all strict intermediates between $a$ and its joint nuclear image $\pi_t(a)$. This is a whole number: it is the number of distinct propositions in $H_t$ that strictly separate $a$ from its telos.

The projection gap size is the eigenvalue of the settlement defect operator at $e_a$. The defect operator $\mathrm{RelationalHistoryFiberSettlementDefectOperator}_t$ is a $K$-linear map on the free module $\widetilde{V}_t = K^{H_t}$, diagonal on the basis $\{e_a : a \in H_t\}$:

$$\mathrm{RelationalHistoryFiberSettlementDefectOperator}_t(e_a) \;=\; \lambda_a \cdot e_a.$$

The zero eigenspace $E_{t,0} = K^{H^*_t}$ consists exactly of the settled elements. Everything outside the fixed fiber contributes a strictly positive eigenvalue.

Shadow class type and gap size are orthogonal. Two elements can have the same $\lambda_a$ while sitting in different shadow classes — equal in distance from $H^*_t$, but reaching it by different paths. And two elements in the same shadow class can have different $\lambda_a$ values — same qualitative structure, different quantitative distance. The shadow class type says which nucleus is blocking; the gap size says how many steps of blocking remain. These are independent coordinates on the settlement deficit.

Kind-1 and Kind-2: gap structure across tower levels

The fiber’s growth structure across tower levels is governed by a $2 \times 2$matrix that encodes how gap types propagate. Every element in$H^{*,\leq n}_t$(the fixed fiber through depth$n$) belongs to exactly one kind:

• Kind-1: elements carried forward from the previous depth — already in $H^{*,\leq n-1}_t$, inherited without change. At the next level, each Kind-1 element remains Kind-1. Self-replication coefficient: $M_{11} = 1$.
• Kind-2: elements newly generated at depth $n$ via the composite nucleus application $\mathrm{RelationalHistoryFiberTransferringNucleus}_t \circ \mathrm{RelationalHistoryFiberSaturatingNucleus}_t$ — they were not in the fixed fiber at depth $n-1$ but reach it at depth $n$ by the two-step closure. At the next level, each Kind-2 element becomes Kind-1 (its gap is now zero, so it carries forward). Non-self-replication coefficient: $M_{22} = 0$.

The count $\binom{d_1(n)}{d_2(n)}$ of Kind-1 and Kind-2 elements at depth $n$ obeys the matrix recurrence

$$\begin{pmatrix}d_1(n+1) \\ d_2(n+1)\end{pmatrix} = M \begin{pmatrix}d_1(n) \\ d_2(n)\end{pmatrix}, \qquad M = \begin{pmatrix}1 & 1 \\ 1 & 0\end{pmatrix}.$$

The matrix entries (RelationalHistoryFiberNuclearTransferMatrixDeterminant) are:

Entry	Value	Meaning
$M_{11} = 1$	Kind-1 → Kind-1	Every carried-forward element stays carried-forward
$M_{22} = 0$	Kind-2 not → Kind-2	No newly-generated element generates a new successor at the same kind
$M_{12} = 1$	Kind-2 → Kind-1	Every newly-generated element graduates to carried-forward at the next level
$M_{21} = 1$	Kind-1 → Kind-2	Each carried-forward element produces exactly one new Kind-2 element

Determinant = $M_{11}M_{22} - M_{12}M_{21} = 0 - 1 = -1$. This is a theorem, not a parameter. Its sign is forced by $M_{22} = 0$ (Kind-2 non-self-replication): since the diagonal term vanishes, the determinant equals $-M_{12}M_{21} = -1$. The negative determinant forces one eigenvalue to be negative — this is the RelationalHyperverseGoldenConjugate = $-1/\varphi \approx -0.618$.

Eigenvalues: the characteristic polynomial of $M$ is $\lambda^2 - \lambda - 1 = 0$, with roots the RelationalHyperverseGoldenRatio $\varphi = (1+\sqrt{5})/2 \approx 1.618$ and RelationalHyperverseGoldenConjugate $= -1/\varphi \approx -0.618$.

Extremal slowness: the minimum gap-closure rate

The count of zero-gap elements at depth $n$ — the size of the fixed fiber at depth $n$, $|H^{*,\leq n}_t|$ — is the RelationalHistoryFiberNuclearQuartetExtremallySlowGrowthSequence:

$$F(n) := |H^{*,\leq n}_t| = 1, 1, 2, 3, 5, 8, 13, 21, \ldots$$

under the recurrence $F(n+1) = F(n) + F(n-1)$ (open at fiber level, established at tower level). Its $n$-th term is given by the Binet formula:

$$F(n) = \frac{\varphi^{n+1} - \mathrm{RelationalHyperverseGoldenConjugate}^{n+1}}{2\varphi - 1}$$

where $2\varphi - 1 = \sqrt{5}$is the eigenvalue separation of$M$.

The name “extremally slow” is structural: by RelationalHistoryFixedFiberGrowthExtremalSlowness, $\varphi$ is the minimum Perron–Frobenius eigenvalue of any primitive $2 \times 2$ nonneg integer matrix. No primitive two-generator substitution system can resolve its gaps more slowly than this rate. The gap-closure process, driven by the Kind-1/Kind-2 decomposition, is as slow as it can possibly be while remaining structurally non-trivial.

What this means for the gap: an element with a large gap eigenvalue $\lambda_a$ takes many tower levels to settle. The rate at which total settled content grows across levels is $\varphi$ per level — the slowest growth that a non-trivial nuclear pair can produce. This is not a limitation but a structural fact: the fiber’s two-dimensional gap structure (non-adjoint nuclei, Kind-1/Kind-2 decomposition) produces exactly the extremally-slow settlement sequence.

The growth rate is model-independent (RelationalHistoryFiberGrowthRateModelIndependence): $\varphi$ does not depend on the specific history $t$, the model (syntactic or other), or the initial conditions $F(0), F(1)$. It is a ring-theoretic invariant of $\mathbb{Z}[\varphi]$: any system governed by the same fusion ring has growth rate $\varphi$. The “constant” $\sqrt{5}$ appearing throughout this theory is $2\varphi - 1$, the eigenvalue separation of$M$, derived from the discriminant formula$\mathrm{tr}(M)^2 - 4\det(M) = 1 + 4 = 5$.

Nuclear reading

Definitions. The saturation gap of $a \in H_t$ is the interval $[a, \sigma_t(a)] = \{b \in H_t : a \leq b \leq \sigma_t(a)\}$. The transfer gap is $[a, \Delta_t(a)]$. The projection gap is $[a, \pi_t(a)]$ where $\pi_t(a) = \sigma_t(\Delta_t(a)) \in H^*_t$. The gap is trivial (a single element) iff the nucleus fixes $a$.

Proposition 1 (The gap types partition $H_t$ into the four shadow class types). The four combinations of trivial/nontrivial saturation and transfer gaps exhaust $H_t$:

$$H_t \;=\; H^*_t \;\sqcup\; \mathrm{SatShadow}_t \;\sqcup\; \mathrm{TrShadow}_t \;\sqcup\; \mathrm{FreeShadow}_t$$

where:

• $H^*_t$: both gaps trivial ($\sigma_t(a) = a$ and $\Delta_t(a) = a$)
• $\mathrm{SatShadow}_t$: saturation gap trivial, transfer gap nontrivial ($\sigma_t(a) = a$, $\Delta_t(a) > a$)
• $\mathrm{TrShadow}_t$: transfer gap trivial, saturation gap nontrivial ($\Delta_t(a) = a$, $\sigma_t(a) > a$)
• $\mathrm{FreeShadow}_t$: both gaps nontrivial ($\sigma_t(a) > a$ and $\Delta_t(a) > a$)

Proof. Each nucleus either fixes $a$ or does not; these are decidable equalities. The four resulting combinations are pairwise incompatible and exhaustive. $\square$

The content. The shadow class type is the qualitative gap dimension: it records which nucleus is missing, not how far away the fixed point is. How far is measured by the gap size $|[a, \pi_t(a)]|$ — a separate quantity tracked in relational-history-fiber-nuclear-quartet-extremally-slow-growth-sequence.md.

Proposition 2 (Single-gap elements have a one-step path to zero gap). For $a \in \mathrm{TrShadow}_t$: applying $\sigma_t$ reaches $H^*_t$ directly — $\sigma_t(a) \in H^*_t$. For $a \in \mathrm{SatShadow}_t$: applying $\Delta_t$ reaches $H^*_t$ directly — $\Delta_t(a) \in H^*_t$.

Proof. Let $a \in \mathrm{TrShadow}_t$, so $\Delta_t(a) = a$. By the commutation axiom: $\Delta_t(\sigma_t(a)) = \sigma_t(\Delta_t(a)) = \sigma_t(a)$. Therefore $\sigma_t(a) \in \mathrm{Fix}(\Delta_t)$. By idempotence, $\sigma_t(\sigma_t(a)) = \sigma_t(a)$, so $\sigma_t(a) \in \mathrm{Fix}(\sigma_t)$. Therefore $\sigma_t(a) \in H^*_t$. The same argument with $\sigma_t$ and $\Delta_t$ swapped handles the SatShadow case. $\square$

The content. Single-gap elements are one nucleus application away from full settlement. The commutation axiom is doing real work here: without it, $\sigma_t(a)$ would land in $\mathrm{Fix}(\sigma_t)$ but might fail $\mathrm{Fix}(\Delta_t)$. Commutation guarantees that resolving one gap automatically resolves the other for single-gap elements. FreeShadow elements require both — and in either order, since both paths reach the same telos.

Proposition 3 (Nucleus application reduces gap type). Applying $\sigma_t$ to an element does not increase the shadow class dimension: it maps:

Input shadow class type	Output shadow class type after applying $\sigma_t$
$H^*_t$	$H^*_t$
$\mathrm{SatShadow}_t$	$\mathrm{SatShadow}_t$
$\mathrm{TrShadow}_t$	$H^*_t$
$\mathrm{FreeShadow}_t$	$\mathrm{SatShadow}_t \cup H^*_t$

Symmetrically, $\Delta_t$ maps TrShadow to TrShadow, SatShadow to $H^*_t$, and FreeShadow to $\mathrm{TrShadow}_t \cup H^*_t$.

Proof. $\sigma_t(a) \in \mathrm{Fix}(\sigma_t)$ by idempotence (the σ-gap of $\sigma_t(a)$ is always zero). The Δ-gap of $\sigma_t(a)$: by commutation, $\Delta_t(\sigma_t(a)) = \sigma_t(\Delta_t(a))$. If $a \in \mathrm{Fix}(\sigma_t)$: $\sigma_t(a) = a$, so the output type is the same as the input type (the $\sigma_t$ application is trivial). If $a \in \mathrm{TrShadow}_t$ ($\Delta_t(a) = a$): $\Delta_t(\sigma_t(a)) = \sigma_t(\Delta_t(a)) = \sigma_t(a)$, so $\sigma_t(a) \in \mathrm{Fix}(\Delta_t)$, giving $\sigma_t(a) \in H^*_t$. If $a \in \mathrm{FreeShadow}_t$: $\sigma_t(a) \in \mathrm{Fix}(\sigma_t)$, and whether $\sigma_t(a) \in \mathrm{Fix}(\Delta_t)$ depends on whether $\sigma_t(\Delta_t(a)) = \sigma_t(a)$, which is not forced by the axioms alone. $\square$

Proposition 4 (Non-adjointness of the nuclear pair = irreducibility of the two-dimensional gap). The saturation and transferring nuclei are not in Galois connection in either direction. If they were — if $\sigma_t \dashv \Delta_t$ — then $\mathrm{Fix}(\Delta_t) \subseteq \mathrm{Fix}(\sigma_t)$, making $\mathrm{TrShadow}_t$ empty. But $\mathrm{TrShadow}_t$ is non-empty (it contributes to the Fibonacci shadow count). Therefore $\sigma_t \not\dashv \Delta_t$. Symmetrically $\Delta_t \not\dashv \sigma_t$.

Content. Non-adjointness is the structural precondition for the gap being genuinely two-dimensional. If either adjunction held, one of the two shadow classes would be empty — every element’s gap could be reduced to a single nucleus — and the Fibonacci growth count

$$d_t(n+1) \;=\; d_t(n) \;+\; d_t(n-1)$$

would degenerate to $d_t(n+1) = d_t(n)$ (linear, not exponential growth). The Fibonacci structure of the shadow resolution count across the tower is the long-range consequence of the gap being irreducibly two-dimensional at each history.

Proposition 5 (Gap type under nucleus-intertwining morphisms). Let $f : H_t \to H_s$ be a Heyting algebra morphism. The shadow class type of $f(a)$ in $H_s$ is determined by which nuclear intertwining conditions $f$ satisfies:

$\sigma_s \circ f = f \circ \sigma_t$	$\Delta_s \circ f = f \circ \Delta_t$	Fate of shadow class type under $f$
Yes	Yes	$f$ preserves shadow class type: $\sigma$-fixed maps to $\sigma$-fixed, etc.
Yes	No	$f$ preserves $\mathrm{Fix}(\sigma)$-membership; Δ-gap of $f(a)$ may differ
No	Yes	$f$ preserves $\mathrm{Fix}(\Delta)$-membership; σ-gap of $f(a)$ may differ
No	No	No shadow class type information is preserved

Proof. If $\sigma_s \circ f = f \circ \sigma_t$ and $a \in \mathrm{Fix}(\sigma_t)$, then $\sigma_s(f(a)) = f(\sigma_t(a)) = f(a)$, so $f(a) \in \mathrm{Fix}(\sigma_s)$. If the condition fails, $\sigma_s(f(a)) = f(\sigma_t(a)) \neq f(a)$, so $f(a) \notin \mathrm{Fix}(\sigma_s)$. The Δ case is identical. $\square$

The delegation connection. Proposition 5 is the structural content behind the shadow class stratification of delegated authority: the restriction map $\rho_{t \to s}$ from principal’s fiber to beneficiary’s fiber may or may not intertwine each nucleus, and the four cases of the intertwining table determine what class the beneficiary’s received authority lands in.

Proposition 6 (The gap size is the defect operator eigenvalue; the spectral radius of gap-closure is the RelationalHyperverseGoldenRatio). The projection gap size $\lambda_a = |[a, \pi_t(a)]| - 1$ is the eigenvalue of the defect operator $\mathrm{RelationalHistoryFiberSettlementDefectOperator}_t$ at basis vector $e_a$. The spectral radius of the defect operator — the largest eigenvalue over all $a \in H_t \setminus H^*_t$ — is bounded by $F(n) - 1$ where $F(n)$ is the $n$-th term of the extremally slow growth sequence. As $n \to \infty$, the maximum possible eigenvalue grows like $\varphi^n$.

Proof. By RelationalHistoryProjectionGapSize: the eigenvalue is the cardinality of the interval $[a, \pi_t(a)]$ minus 1. The defect operator is diagonal; its spectral radius is the maximum eigenvalue. The maximum size of any interval $[a, \pi_t(a)]$ at depth $n$ is bounded by the total fiber size at depth $n$, which grows like $\varphi^n$ by the Binet formula. $\square$

Content. Gap size and shadow class type are genuinely orthogonal spectral data: two elements can have the same shadow class type but different eigenvalues (same qualitative position, different quantitative distance), or the same eigenvalue but different shadow class types (same distance, different directional defect). The defect operator captures the quantitative dimension only; the shadow class captures the qualitative.

Proposition 7 (The Fibonacci recurrence is the eigenvalue equation for the gap operator on settled-element counts; the golden ratio is the spectral radius). The total number of zero-gap elements at depth $n$ satisfies $F(n+1) = F(n) + F(n-1)$, which is the characteristic equation $\lambda^2 - \lambda - 1 = 0$ evaluated at the transfer matrix $M = [[1,1],[1,0]]$. The positive eigenvalue $\varphi = (1+\sqrt{5})/2$ is the spectral radius of $M$ — the dominant growth rate.

Proof. By RelationalHistoryFiberNuclearQuartetExtremallySlowGrowthSequence: the recurrence follows from the Kind-1/Kind-2 decomposition, which itself follows from the matrix entries of $M$. The characteristic polynomial of $M$ is $\lambda^2 - \mathrm{tr}(M)\lambda + \det(M) = \lambda^2 - \lambda - 1$, whose positive root is $\varphi$. $\square$

Content. The open question in the earlier version of this spec — “whether the Fibonacci recurrence is the eigenvalue equation for the gap operator on the shadow count, and whether the golden ratio φ is the spectral radius of the combined gap closure” — is answered affirmatively by RelationalHistoryFiberNuclearTransferMatrixDeterminant and RelationalHistoryFiberGrowthRateModelIndependence. The growth rate is not an approximation but an exact ring-theoretic invariant.

Proposition 8 (The determinant of the gap-transition matrix is −1; this is forced by Kind-2 non-self-replication and is a theorem not a parameter). $\det(M) = M_{11}M_{22} - M_{12}M_{21} = 1 \cdot 0 - 1 \cdot 1 = -1$.

Proof. By RelationalHistoryFiberNuclearTransferMatrixDeterminant: $M_{22} = 0$ because no Kind-2 element generates a Kind-2 successor; $M_{12} = M_{21} = 1$ by primitivity. The determinant vanishes in the self-replication term and equals $-1$ in the cross-replication term. $\square$

Content. The negative determinant is not an accident. It is the algebraic encoding of the fact that the two kinds of settled elements — carried-forward and newly-generated — cannot mutually generate the same kind: Kind-2 produces Kind-1 in the next generation, and Kind-1 produces exactly one Kind-2, but Kind-2 does not produce Kind-2. The $-1$ forces the golden conjugate to be negative, which is what makes the Binet formula’s correction term $\mathrm{GoldenConjugate}^{n+1}$ alternate in sign. For large $n$, the correction term decays to zero, and $F(n) \approx \varphi^{n+1}/(2\varphi-1)$.

External connection: balayage in potential theory. The saturation nucleus $\sigma_t$ on $H_t$ is the formal analogue of the balayage (sweeping) operator in classical potential theory. In potential theory, the balayage of a measure $\mu$ with respect to a compact set $K$ is the smallest superharmonic measure $\geq \mu$ that is harmonic outside $K$. The operation $\mu \mapsto \mathrm{bal}(\mu, K)$ is:

1. Extensive: $\mathrm{bal}(\mu, K) \geq \mu$
2. Idempotent: $\mathrm{bal}(\mathrm{bal}(\mu, K), K) = \mathrm{bal}(\mu, K)$
3. Meet-preserving: $\mathrm{bal}(\mu \wedge \nu, K) = \mathrm{bal}(\mu, K) \wedge \mathrm{bal}(\nu, K)$

— precisely the three nucleus axioms. The “gap” $[\mu, \mathrm{bal}(\mu, K)]$ in potential theory is the energy swept in the balayage: the measure of how far $\mu$ must be lifted to become harmonic on the complement of $K$. The saturation gap in this system is the analogous sweep: how far $a \in H_t$ must be lifted to be recognized as backward-settled (harmonic with respect to the restriction history).

The analogy is not merely formal: in both cases, the nucleus/(balayage) is determined by a boundary condition (past restriction maps / the compact set $K$), and the gap measures failure to satisfy that condition. The zero-gap elements are the “harmonic” propositions — those fully consistent with their own history.

Open questions

• Whether the two-dimensional projection gap $(|[a, \sigma_t(a)]| - 1,\; |[a, \Delta_t(a)]| - 1)$ defines a well-formed distance on $H_t$ satisfying a triangle inequality — whether $\lambda_{a \wedge b} \leq \lambda_a + \lambda_b$ holds under the Derivation Condition, or whether the gap eigenvalue can be superadditive under conjunction.
• Whether Zeno-settling elements — elements that never reach $H^*_t$ at any finite tower level but whose projection gap size $\lambda_a$ decreases at rate $1/\varphi per level — converge to gap-zero at the tower colimit, and whether the colimit fiber H^*_\infty$ contains elements whose gap was strictly positive at every finite stage (corresponding to the Zeno settling cohomology zeno-settling-k1-cohomology.md).
• Whether the gap size $\lambda_a$ is monotone-decreasing under nucleus-intertwining morphisms: whether $\lambda_{f(a)} \leq \lambda_a$ for all Heyting algebra morphisms $f : H_t \to H_s$ satisfying $\sigma_s \circ f = f \circ \sigma_t$ and $\Delta_s \circ f = f \circ \Delta_t$, giving a functor from the history category to the poset of gap sizes.
• Whether the extremal slowness of the gap-resolution sequence (minimum Perron–Frobenius eigenvalue over all primitive $2 \times 2$ matrices) has a physical interpretation — whether the minimum possible rate of gap-closure corresponds to a maximum entropy principle for the settlement process, and whether the departure from the extremally slow rate in more complex nuclear structures is measurable.
• Whether the Kind-1/Kind-2 decomposition — and hence the transfer matrix $M = [[1,1],[1,0]]$ and the $\det(M) = -1$ theorem — can be given a fiber-level proof independent of the Fibonacci recurrence conjecture, or whether the two are exactly equivalent: $M = [[1,1],[1,0]]$ forces the Fibonacci recurrence, and the Fibonacci recurrence forces the transfer matrix. If they are equivalent, the open conjecture and the matrix structure are the same open question asked in two different languages.