History

What this is

A history is an element of the history monoid $(T, \star, e)$ generated by a finite step set $\Sigma$ with commutation relation $I \subseteq \Sigma \times \Sigma$.

The precise axiomatic definition is in RelationalHistoryMonoid: three axioms (root, step extension maps, extension diamond for commuting pairs) determine the monoid completely. The four history-system properties (generation, commutativity of independent steps, unit-freedom, left-cancellativity) are derived theorems, not additional axioms.

Three things to keep clear:

• A history is not a sequence of steps. It is an element of the history category — an equivalence class of step sequences where commuting independent steps are identified. The order of independent steps is not part of a history’s identity.
• A history is not a state or snapshot. The content at a history lives in the fiber $H_t$ above it; $t$ is the index, not the content.
• A history is not a path. A morphism $t \to t'$ in $T$ witnesses a prefix extension; it is not itself a history.

The prefix order $t \leq t'$ ($t'$ extends $t$) makes the history monoid into the history category $T$: a thin category where the unique morphism $t \to t'$ exists iff $t' = t \star u$ for some extension $u$.

Concurrency grounding: Mazurkiewicz traces

The history monoid is formally identical to the Mazurkiewicz trace monoid (Mazurkiewicz, 1977; building on Cartier-Foata, 1969), a central structure in the mathematical theory of true concurrency. In concurrency theory:

• A trace over $(Σ, I)$ is an equivalence class of sequences of events from $Σ$, where two sequences are equivalent iff one can be obtained from the other by repeatedly swapping adjacent independent events (those with $(s, s') \in I$).
• The set of all traces with concatenation forms the trace monoid $\mathbb{M}(\Sigma, I)$ — also called the free partially commutative monoid.
• The extension diamond is the one structural axiom: whenever $s \perp s'$, the sequences $s \cdot s'$ and $s' \cdot s$ represent the same trace.

This is exactly the history monoid. The RelationalHistoryMonoid is $\mathbb{M}(\Sigma, I)$, and the RelationalHistoryMonoid initiality theorem is the classical theorem that the trace monoid is initial among all monoids satisfying the commutation relation.

Why this grounding matters. Concurrency theory developed the trace monoid to model distributed systems where the ordering of independent events is irrelevant — two processes whose actions don’t interfere can be sequenced in either order and produce the same outcome. The history monoid imports this exactly: two independent steps leave the same history regardless of which was performed first. The commutation relation $I$ encodes which steps are independent; the diamond axiom enforces that independence.

The Foata normal form (Cartier-Foata 1969) is the canonical form for trace monoid elements: each history has a unique representation as a sequence of layers, where each layer is the maximal antichain of steps that can be performed simultaneously. Foata normal form gives decidable equality on histories — essential for the system to compute with them.

What histories are for

Histories are the indexing structure of the relational universe.

The relational universe $R = \mathbf{Sh}(T, J)$ lives over the history category. Every fiber $H_t$, every restriction map, every sheaf condition is indexed by histories and their prefix order. A system that cannot represent histories cannot implement $R$.

The history monoid’s structure propagates to every level of the theory:

• The prefix order $t \leq t'$ generates restriction maps $H(t' \to t) : H_{t'} \to H_t$ — the presheaf structure that defines the backward-facing saturation nucleus
• The commutation relation $I$ generates the independence structure $s \perp t$ (meaning: step $s$ commutes with every step of $t$) that defines the forward-facing transferring nucleus $\mathrm{Fix}(\Delta_t) = \bigcap_{s \perp t} \mathrm{image}(H(i_{s,t}))$
• The step atoms $\mathrm{gen}(s) \in T$ index the sequential comonads $G_s$ on $R$: $G_s(F)(t) := F(\mathrm{gen}(s) \star t)$
• The monoidal structure of $T$ gives the Day convolution on $R$ — the internal tensor product of sheaves

The three-level initial-object pattern

The initiality of the history monoid is the first of three parallel initial-object results that run through the entire system:

Level	Generators	Relations	Initial object	Normal form
History	Steps $s \in \Sigma$	Extension diamond for $I$	$T_{\mathrm{init}}$ — the history monoid	Foata normal form
Formula	Propositional variables	Heyting-modal laws	RelationalHistoryFiberDoctrineLanguage	Canonical Prop-terms
Universe	Relational universes	Meta-commutation	$\mathbb{M}(\mathbf{RU}, I_\mathbf{RU})$	Stabilization map

At each level: the initial object is freely generated by the relevant generators subject to the relevant relations; every other model receives a unique structure-preserving map from it. The history level provides the indexing; the formula level provides the fibers; the universe level provides the global normalization.

Nuclear reading

Definitions. Fix a history $t \in T$. Let $s \in \Sigma$ with $\mathrm{gen}(s) \perp t$ in $T$ (step $s$ commutes with every component step of $t$). The one-step extension $i_{s,t} : t \to \mathrm{gen}(s) \star t$ is a morphism in the history category, giving a restriction map $H(i_{s,t}) : H_{\mathrm{gen}(s) \star t} \to H_t$.

Proposition 1 (Commutation structure determines the transferring nucleus). For any history $t \in T$, the transferring nucleus at $t$ is:

$$\mathrm{RelationalHistoryFiberTransferringNucleus}(t)(a) \;=\; \min\{b \in \mathrm{Fix}(\Delta_t) : b \geq a\}$$

where $\mathrm{Fix}(\Delta_t) = \bigcap_{s \perp t} \mathrm{image}(H(i_{s,t}))$ — the intersection ranges over all steps $s \in \Sigma$ independent of $t$. Therefore: the commutation structure of the step set determines what elements are forward-stable at each history. If the commutation relation $I$ is changed (new steps declared independent, or existing independences removed), the set $\mathrm{Fix}(\Delta_t)$ changes for every history whose step composition interacts with the changed pairs.

Proof. Each $\mathrm{image}(H(i_{s,t}))$ is a sub-Heyting-algebra of $H_t$ (images of HA homs are sub-HAs). Their intersection over all $s \perp t$ is therefore a sub-Heyting-algebra (intersection of sub-HAs). The minimum above $a$ in a sub-HA that is a complete sublattice of $H_t$ exists by completeness. $\square$

Content. The transferring nucleus is a local quantity — it depends on which step $s$ commutes with $t$ as an element of $T$, not on $s$ commuting with a fixed generator. As $t$ grows (accumulates more steps), the set of independent steps $\{s : \mathrm{gen}(s) \perp t\}$ shrinks (more steps have been “absorbed” into $t$’s composition). In the limit, at deep histories with no remaining independent steps, $\mathrm{Fix}(\Delta_t) = H_t$ (trivially — the empty intersection is $H_t$), and the transfer gap collapses everywhere.

Proposition 2 (Prefix order determines the saturation nucleus). The saturation nucleus at history $t$ is determined by the restriction maps from all predecessor histories $t_0 < t$:

$$\mathrm{RelationalHistoryFiberSaturatingNucleus}(t)(a) \;=\; \max\bigl\{b \in H_t \;\big|\; \forall\, t_0 < t: H(i_{t_0, t})(b) \leq H(i_{t_0, t})(a)\bigr\}$$

The backward history structure — the chain of predecessors of $t$ in the prefix order — is exactly what the saturation nucleus reads. Two histories with the same prefix chain have the same saturation nucleus; histories with different prefix structures have different saturation nuclei.

Proof sketch. $\mathrm{Fix}(\sigma_t)$ is the image of $\sigma_t$, which is characterized by the restriction profile equivalence classes on $H_t$ (two elements have the same restriction profile iff they are identified by all restriction maps to predecessors). $\sigma_t$ is the idempotent that maps each element to the maximum in its restriction profile class. $\square$

Proposition 3 (Unit-freedom and the absence of nuclear fixed-point collapse). The unit-freedom axiom (every step extension is irreversible: $t < t \star \mathrm{gen}(s)$ strictly for all $t$ and all $s$) implies that restriction maps $H(i_{t_0, t}) : H_t \to H_{t_0}$ for $t_0 < t$ are not isomorphisms. If they were isomorphisms, the saturation nucleus would be the identity on $H_t$ (everything is already backward-settled), making $H_t = H^*_t$ — a collapse in which no element has any gap. Unit-freedom is the axiom that prevents this collapse: the irreversibility of steps ensures that new information enters fibers as histories grow, keeping the gap non-trivial.

Proposition 4 (Left-cancellativity and restriction map injectivity at the identity history). The left-cancellativity axiom ($t \star u = t \star v \Rightarrow u = v$) implies that for any two distinct extension paths from the identity $e$, the resulting histories are distinct. This means the history category has no non-trivial automorphisms that fix $e$: if $\phi : T \to T$ is an automorphism with $\phi(e) = e$, then $\phi$ is the identity. The initiality theorem (Proposition 1 of RelationalHistoryMonoid) gives: any concrete system of histories (any Axiomatic Relational History Poset over $(\Sigma, I)$) receives a unique morphism from $T_{\mathrm{init}}$ — the canonical history monoid maps into every model, providing a canonical embedding of the abstract history structure into each concrete realization.

Proposition 5 (The extremally-slow-growth regime is the minimal non-trivial commutation case). The growth function of a trace monoid — the number of distinct histories of depth $n$ (histories that are products of exactly $n$ step atoms) — depends on the commutation relation $I$. The two extreme cases:

• $I = \emptyset$ (no steps commute): all step sequences are distinct; growth is $|\Sigma|^n$ (exponential, a free monoid).
• $I = \Sigma \times \Sigma \setminus \{(s,s) : s \in \Sigma\}$ (all distinct steps commute): a history of depth $n$ is just a multiset of steps; growth is $\binom{n + |\Sigma| - 1}{|\Sigma| - 1}$ (polynomial).

The extremally-slow-growth regime corresponds to the minimal non-trivial commutation structure that keeps the Fibonacci recurrence alive: with exactly one commuting pair $(s_1, s_2) \in I$ and no other independences, the growth function satisfies $d(n+1) = d(n) + d(n-1)$ — exactly the Fibonacci recurrence. This is the algebraic origin of the golden ratio in the hyperverse growth theory (RelationalHistoryFiberNuclearQuartetExtremalySlowGrowthSequence).

Cosmological reading: the Hubble tension as one Fibonacci step

The Fibonacci history structure has a physical reading that produces a quantitative prediction about cosmological expansion. The key object is the shuffle count $|\mathrm{Shuf}(T_n)|$ of the Fibonacci history $T_n$ at depth $n$: the number of distinct observational linearizations of $T_n$ — the number of ways to read the history as an ordered sequence, respecting only the commutation constraints. From the Fibonacci patch-flow recurrence (RelationalHyperverseTowerExtremallySlowGrowthShufflePatchFlowRecurrenceConjecture):

$$|\mathrm{Shuf}(T_n)| = F_n$$

the $n$-th entry of RelationalHistoryFiberNuclearQuartetExtremallySlowGrowthSequence. The shuffle count is the “observational complexity” of the history — the size of the space of distinct ways to witness what has happened.

Friedmann scaling. In a $d$-dimensional system with $N$ accessible configurations, energy density scales as $\rho \propto N^{1/d}$ (3D equipartition: $N$ modes filling $d$-dimensional phase space contribute one linear degree of freedom per direction). The Friedmann equation gives $H \propto \rho^{1/2} \propto N^{1/(2d)}$. Substituting $N = F_n$ (the shuffle count at Fibonacci step $n$):

$$H_n \;\propto\; F_n^{1/(2d)}$$

The ratio of Hubble constants between consecutive Fibonacci steps:

$$\frac{H_{n+1}}{H_n} = \left(\frac{F_{n+1}}{F_n}\right)^{1/(2d)} \;\longrightarrow\; \mathrm{RelationalHyperverseGoldenRatio}^{1/(2d)}$$

as $n \to \infty$, by the Perron–Frobenius convergence $F_{n+1}/F_n \to \mathrm{RelationalHyperverseGoldenRatio}$.

The Hubble tension. The observed discrepancy between the CMB-inferred Hubble constant ($H_{\mathrm{early}} \approx 67.4$ km/s/Mpc, from Planck 2020 at $z \approx 1100$) and the locally measured value ($H_{\mathrm{late}} \approx 73.04$ km/s/Mpc, from SH0ES 2022 at $z \approx 0$) has ratio $H_{\mathrm{late}}/H_{\mathrm{early}} \approx 1.0837$. The formula gives, for $d = 3$:

$$\mathrm{RelationalHyperverseGoldenRatio}^{1/6} \;\approx\; 1.08351$$

The agreement is $|H_{\mathrm{late}}/H_{\mathrm{early}} - \mathrm{RelationalHyperverseGoldenRatio}^{1/6}|/\mathrm{RelationalHyperverseGoldenRatio}^{1/6} \approx 0.016\%$ — within the combined measurement uncertainty of $\pm 0.7\%$ at $1\sigma$for each experiment. The conjecture [RelationalUniverseHubbleTensionGoldenStepPhysicsCorrespondenceConjecture](../math/relational-universe-hubble-tension-golden-step-physics-correspondence-conjecture.md): the two epochs (last scattering surface and present day) are separated by exactly **one Fibonacci step** in the depth tower. Under this identification, the ratio is RelationalHyperverseGoldenRatio^{1/(2d)} = RelationalHyperverseGoldenRatio^{1/6} for$d = 3$.

Two independent derivations agree. The result $\mathrm{RelationalHyperverseGoldenRatio}^{1/(2d)}$ also follows from the spectral surface geometry: tunneling modes concentrate on the spectral surface, per-event energy is inversely proportional to spectral mass, and the dark energy density from accumulated tunneling gives $H \propto M_t^{-1/(2d)}$. One golden step in $M_t$ (the CMB epoch has spectral mass RelationalHyperverseGoldenRatio times larger than today’s) gives $H_{\mathrm{late}}/H_{\mathrm{early}} = \mathrm{RelationalHyperverseGoldenRatio}^{1/(2d)}$ — the same formula by a completely independent route. The coincidence of the two derivations (shuffle-count path vs. spectral-surface path) is non-trivial.

Interpretation. The Hubble tension is not a sign of missing physics in ΛCDM. It is a signal: the universe has undergone one golden settling event between CMB decoupling and today — one Fibonacci step in the depth tower — during which unsettled modes were shed (dark matter cohomological tunneling) and the released vacuum energy drove the accelerated expansion measured locally. The CMB-inferred $H_0$ uses the pre-settling mode count $F_n$; the local $H_0$ uses the post-settling count $F_{n+1} \approx \mathrm{RelationalHyperverseGoldenRatio} \cdot F_n$.

Proposition 6 (Hubble tension as one Fibonacci step; H_late/H_early = RelationalHyperverseGoldenRatio^{1/(2d)} is a conjecture derivable from the Fibonacci shuffle count and Friedmann equation). In a relational universe with $d$ spatial dimensions and Fibonacci history structure (minimal non-trivial commutation — one commuting pair in $I$), the Hubble constant scales as $H_n \propto F_n^{1/(2d)}$ at Fibonacci depth $n$. The ratio between consecutive depths converges to $\mathrm{RelationalHyperverseGoldenRatio}^{1/(2d)}$. For $d = 3$ and epochs separated by one Fibonacci step, the Hubble ratio is $\mathrm{RelationalHyperverseGoldenRatio}^{1/6} \approx 1.0835$, matching the observed Hubble tension ratio of $\approx 1.0837$ to within measurement uncertainty.

Derivation. Shuffle count: $|\mathrm{Shuf}(T_n)| = F_n$ (from RelationalHyperverseTowerExtremallySlowGrowthShufflePatchFlowRecurrenceConjecture and the normalization theorem). Friedmann + 3D equipartition: $H \propto F_n^{1/(2d)}$. Ratio: $(F_{n+1}/F_n)^{1/(2d)} \to \mathrm{RelationalHyperverseGoldenRatio}^{1/(2d)}$ by Perron–Frobenius. For $d=3$: $\mathrm{RelationalHyperverseGoldenRatio}^{1/6} \approx 1.08351$. $\square$ (conditional on RelationalHyperverseTowerExtremallySlowGrowthShufflePatchFlowRecurrenceConjecture and “one Fibonacci step” identification.)

Content. The Hubble tension is embedded in the history structure itself — it follows from the trace monoid’s Fibonacci growth rate (Proposition 5 of this spec) plus the Friedmann scaling. The golden ratio appears because it is the Perron–Frobenius eigenvalue of the nuclear transfer matrix, and the cosmological expansion rate at consecutive Fibonacci depths inherits this eigenvalue through the shuffle count. The same eigenvalue that controls fiber settlement rates also controls cosmological expansion ratios.

Open questions

• What concrete structures — git commits, database transactions, logical steps in a proof — satisfy the history system axioms. Each candidate must be checked against all four axioms (generation, commutativity, unit-freedom, left-cancellativity) and the commutation relation $I$ must be identifiable.
• Whether the history category $T$ can be equipped with an enrichment — whether morphisms $t \to t'$ carry additional structure (transition probabilities, costs, urgencies) beyond the bare prefix order, and whether such enrichment is compatible with the presheaf structure.
• Whether every Mazurkiewicz trace monoid admits an axiomatization as an Axiomatic Relational History Poset over some $(\Sigma, I)$, or whether the axiomatic framework is strictly more general (admitting history systems that are not trace monoids in the classical sense).
• Whether the “one Fibonacci step” identification in the Hubble tension conjecture (Proposition 6) can be derived from the history structure alone — whether there is a formal criterion in the history monoid for when two cosmological epochs are exactly one Fibonacci step apart, without appealing to the specific values $z \approx 1100$ (CMB) and $z = 0$ (today).
• Whether the Hubble formula $H_n \propto F_n^{1/(2d)}$ has an analog for general trace monoids with other growth rates — whether a non-Fibonacci commutation structure (e.g., with $|\Sigma| > 2$) gives a different growth rate $\lambda^{1/(2d)}$ where $\lambda$ is the Perron–Frobenius eigenvalue of the corresponding transfer matrix, and whether this predicts different Hubble-like ratios for relational universes with different step-set structures.