theory.md

A Mathematical Theory of Fixed Gaps

Companion to essay.md. We formalize "fixed gaps" — relations between entities that are invariant over time — and prove the essay's claims.

The theory has three pillars:

1. Gap structures (§1–§4): quantitative gaps as torsor-like structures, where only differences are defined. Main results: absolute values exist only up to gauge (Prop 1.3), gaps are fixed iff everything drifts together (Thm 2.3), the gap table determines the world up to translation (Thm 3.1), and fixed additive gaps fade in relative terms (Thm 4.1).
2. Gaps as group invariants (§5): the Klein-program view. Which gaps are fixed depends on which transformations you admit; relativity is a change of group, not a refutation.
3. Append-only worlds (§6): qualitative gaps like kinship and causal order. Main result: positive existential facts are exactly the kind that, once true, stay true (Prop 6.2).

Throughout, $A$ is a nonempty set of objects and $(G, +)$ an abelian group of gap values (typically $\mathbb{R}$; §8 drops commutativity).

---

1. Gap structures

Definition 1.1 (gap structure). A $G$-gap structure on $A$ is a map $\delta : A \times A \to G$ satisfying the cocycle identity

$$\delta(a,c) = \delta(a,b) + \delta(b,c) \qquad \text{for all } a,b,c \in A.$$

It is separated if $\delta(a,b) = 0 \implies a = b$.

The cocycle identity is the essence of gap-ness: gaps compose along chains. The age gap from you to your grandmother is your gap to your mother plus hers to her mother.

Lemma 1.2. $\delta(a,a) = 0$ and $\delta(b,a) = -\delta(a,b)$.

Proof. Setting $a=b=c$ in the cocycle identity gives $\delta(a,a) = 2,\delta(a,a)$, so $\delta(a,a)=0$. Then $\delta(a,b) + \delta(b,a) = \delta(a,a) = 0$. ∎

Proposition 1.3 (gauge freedom — "the zero point is a convention"). Fix any $a_0 \in A$ and define $f(x) := \delta(a_0, x)$. Then

$$\delta(a,b) = f(b) - f(a) \quad \text{for all } a, b.$$

Moreover, if $f'$ is any other function with $\delta(a,b) = f'(b) - f'(a)$, then $f' - f$ is constant. Conversely, every $f : A \to G$ defines a gap structure by $\delta(a,b) := f(b) - f(a)$.

Proof. $f(b) - f(a) = \delta(a_0,b) - \delta(a_0,a) = \delta(a,a_0) + \delta(a_0,b) = \delta(a,b)$ by the cocycle identity and Lemma 1.2. If $f'(b)-f'(a) = f(b)-f(a)$ for all $a,b$, then $(f'-f)(b) = (f'-f)(a)$, i.e. $f'-f$ is constant. The converse is a direct check of the cocycle identity. ∎

Interpretation. Gap structures on $A$ are exactly potentials modulo constants. You can always assign absolute values (a birthdate coordinate, a voltage, a height above sea level) consistent with the gaps — but never canonically: the assignment is unique only up to a global constant. The "absolute value" is real as a gauge choice and fictional as a fact. This is precisely why the zero of potential energy, of voltage, and of longitude are conventions, while their differences are physics.

Note that $\delta$ is separated iff $f$ is injective, so a separated gap structure embeds $A$ into $G$; if $f$ is also surjective, $A$ is a $G$-torsor — a copy of the group that has forgotten its identity element. Time itself is the paradigm: instants form an $\mathbb{R}$-torsor; there is no "instant zero," only durations.

---

2. Dynamics: when is a gap fixed?

Definition 2.1. An evolution of $A$ is a family of maps $\varphi_t : A \to A$, $t \in T$ (time; any monoid acting on $A$). The gap structure $\delta$ is fixed under $\varphi$ if

$$\delta(\varphi_t a, \varphi_t b) = \delta(a, b) \qquad \text{for all } a, b \in A,\ t \in T.$$

Proposition 2.2 (common drift $\Rightarrow$ fixed). Suppose the evolution is a uniform drift: there exist $g_t \in G$ with $f(\varphi_t a) = f(a) + g_t$ for all $a$ (every object's potential advances by the same amount). Then $\delta$ is fixed under $\varphi$.

Proof. $\delta(\varphi_t a, \varphi_t b) = f(\varphi_t b) - f(\varphi_t a) = (f(b) + g_t) - (f(a) + g_t) = \delta(a,b)$. ∎

Ages are the motivating case: with birthdates $b_i$ and $f = $ age, $f_t(i) = t - b_i$ advances at rate $1$ for everyone, so all age gaps are conserved.

The converse is the substantive theorem: uniform drift is not just a way to keep gaps fixed — it is the only way.

Theorem 2.3 (rigidity). Let $\delta$ be a separated gap structure and $\varphi : A \to A$ any map with $\delta(\varphi a, \varphi b) = \delta(a,b)$ for all $a,b$. Then $h(a) := \delta(a, \varphi a)$ is constant in $a$ — i.e. $\varphi$ is a uniform drift.

Proof. For any $a, b$:

$$\delta(a, \varphi a) = \delta(a,b) + \delta(b, \varphi b) + \delta(\varphi b, \varphi a) = \delta(a,b) + \delta(b,\varphi b) - \delta(\varphi a, \varphi b) = \delta(b, \varphi b),$$

using the cocycle identity, Lemma 1.2, and gap-preservation. ∎

Corollary 2.4. The symmetry group of a separated gap structure — all invertible gap-preserving self-maps — is exactly the group of uniform drifts, a subgroup of $G$ acting by translation. For a torsor it is all of $G$.

Interpretation. Fixed gaps and lockstep motion are equivalent. Age gaps are frozen because everyone ages at the same rate; Theorem 2.3 says any frozen quantitative gap forces an "equal aging rate" for the quantity behind it. Conversely, gaps break exactly when drift stops being uniform — which is precisely what special relativity discovered about elapsed time (§5). This is a small Noether principle: the conserved quantity (all pairwise differences) corresponds exactly to the symmetry (global translation).

---

3. Reconstruction: structuralism as a theorem

Theorem 3.1. Let $\delta$ be a separated $G$-gap structure on $A$. Then the potential $f : A \hookrightarrow G$ of Prop 1.3 is determined by the gap data uniquely up to a single additive constant. Consequently:

1. The full gap table ${\delta(a,b)}_{a,b \in A}$ determines the configuration of all objects up to one global translation of $G$.
2. Any two gap structures with the same gap table are isomorphic via a unique gap-preserving bijection matching corresponding objects.

Proof. Immediate from Prop 1.3 (existence and uniqueness-up-to-constant of potentials) and Theorem 2.3 (gap-preserving maps are translations). ∎

Interpretation. The relations are the reality: the complete table of gaps pins down everything about the configuration except an absolute origin — which is exactly the datum that was never observable in the first place. This is the essay's structuralist claim (Leibniz contra Newton) in theorem form. Nothing about the objects' "positions" exceeds the gap data plus one unphysical choice.

---

4. The convergence theorem: why fixed gaps fade

Additive invariance and multiplicative invariance behave oppositely, and the human phenomenology of age gaps falls out of the additive case.

Theorem 4.1 (relative fading). Let $x(t), y(t) \to \infty$ with $x(t) - y(t)$ bounded. Then $x(t)/y(t) \to 1$.

Proof. $x/y = 1 + (x - y)/y$, and the numerator is bounded while $y \to \infty$. ∎

Corollary 4.2 (ages). For ages $a_i(t) = t - b_i$, every pairwise gap $a_i - a_j = b_j - b_i$ is exactly conserved, while every ratio $a_i / a_j \to 1$. All beings become contemporaries in the limit.

Contrast 4.3 (multiplicative gaps diverge). If instead the ratio is invariant, $x(t) = \lambda, y(t)$ with $\lambda > 1$ and $y \to \infty$, then $x - y = (\lambda - 1), y \to \infty$: fixed relative gap, exploding absolute gap. Compound growth (wealth, populations, citation counts) lives in this regime.

Interpretation. "Which kind of gap does the domain conserve?" is the first question to ask of any inequality. Domains with additive conservation (age, seniority) equalize in proportion over time; domains with multiplicative conservation (compounding capital) diverge in absolute terms while looking statically "fair" in ratio. The lived softening of the parent–child hierarchy and the widening of compounding wealth are the two branches of the same dichotomy.

---

5. Gaps as invariants of a group: the Klein view

Sections 1–4 assumed the gap values were differences in an abelian group. The general notion drops that: a gap is anything about a pair that the admissible transformations cannot touch.

Definition 5.1. Let a group $\Gamma$ act on a state space $S$. A $\Gamma$-gap with values in a set $V$ is a map $g : S \times S \to V$ invariant under the diagonal action:

$$g(\gamma x, \gamma y) = g(x, y) \qquad \text{for all } \gamma \in \Gamma,\ x, y \in S.$$

Write $\mathrm{Gaps}(\Gamma)$ for the set of such invariants.

Proposition 5.2 (more symmetry, fewer gaps). If $\Gamma \subseteq \Gamma'$ then $\mathrm{Gaps}(\Gamma') \subseteq \mathrm{Gaps}(\Gamma)$.

Proof. Invariance under every element of $\Gamma'$ includes invariance under every element of $\Gamma$. ∎

Examples.

Admissible group $\Gamma$	Surviving gaps	Casualties
time translations	age differences, order of birth	absolute dates
Euclidean isometries	distances, angles	absolute positions, orientations
Galilean group	relative velocities; distances at a common instant	absolute velocity ("the ether")
Poincaré group	spacetime interval, proper time, causal order	simultaneity, elapsed coordinate time, spatial distance

Interpretation. "Is this gap fixed?" is not absolute — it is relative to the group of transformations the world (or the observer) is allowed to apply. Special relativity did not refute the fixed-gap thesis; it enlarged $\Gamma$, which by Prop 5.2 shrank the inventory of invariant gaps, leaving the deeper ones (the interval, the causal order). The research program of physics can be read as: find nature's true $\Gamma$, then compute $\mathrm{Gaps}(\Gamma)$ — and Theorem 2.3 is the special case saying that for one-dimensional additive quantities, conserving all pairwise gaps is translation symmetry.

---

6. Append-only worlds: kinship, causality, ledgers

Kinship gaps ("she is my sister") are not numbers, and their fixedness has a different mechanism: not lockstep drift, but the append-only character of the past. We model it logically.

Definition 6.1. Fix a relational signature $\sigma$ (e.g. binary $\mathrm{parent}(x,y)$). A history is a chain of $\sigma$-structures

$$M_0 \subseteq M_1 \subseteq M_2 \subseteq \cdots$$

where each inclusion is append-only: the universe may gain elements and relations may gain tuples, but no element or atomic fact is ever removed. A property $P(\bar a)$ is stable (a fixed gap) if $M_s \models P(\bar a)$ implies $M_t \models P(\bar a)$ for all $t \ge s$.

Proposition 6.2. Every existential-positive formula — built from atomic formulas using only $\wedge$, $\vee$, and $\exists$ — is stable. Formulas involving negation or universal quantification need not be.

Proof. Induction on formula structure. Atomic facts persist by the append-only condition. Conjunction and disjunction preserve persistence. For $\exists x, \psi$: a witness in $M_s$ still exists and still satisfies $\psi$ in $M_t$ by the inductive hypothesis. For the negative claim, two counterexamples: $\neg\exists y, \mathrm{parent}(x,y)$ ("$x$ is childless") is destroyed by appending a child; $\forall y, (\mathrm{parent}(x,y) \to \mathrm{parent}(z,y))$ is destroyed by appending a child of $x$ alone. ∎

Application (kinship). Parenthood edges are appended at birth and never deleted. Hence sibling ($\exists z,\mathrm{parent}(z,x) \wedge \mathrm{parent}(z,y)$), grandparent, cousin, and ancestor (a monotone transitive closure, an increasing union of existential-positive formulas) are all stable — fixed forever once true. The asymmetry of family life falls out of the syntax: positive kinship facts are fixed; negative ones ("childless," "unmarried," "only child") are not. Blood relations are fixed gaps not because of biology per se, but because the world never un-happens an event.

Application (causality and ledgers). Causal precedence is a partial order to which history only ever adds comparable pairs; an existing pair $a \prec b$ is never reversed — order facts are atomic, hence stable. A blockchain is this proposition turned into engineering: an append-only structure whose entire security model is making retro-deletion of facts computationally infeasible. It is a machine for manufacturing fixed gaps (confirmed precedence of transactions) in an adversarial environment.

Remark. Prop 6.2 is the finger-exercise version of preservation theorems in model theory (Łoś–Tarski; the homomorphism-preservation theorem of Rossman), which characterize stable properties as exactly the existential-positive ones under suitable hypotheses. In slogan form: the fixed gaps of an append-only world are precisely its positive existential facts.

---

7. The meaning layer: readings over a frozen substrate

The essay claims: "the gap is fixed, but its meaning is not." Formally:

Definition 7.1. Given a gap structure with potential $f_t$ evolving by uniform drift (so all gaps are fixed), a reading is any function $m(a, b, t)$ of the pair's full state — not required to be invariant.

Readings vary even though gaps do not, because they can depend on the absolute potentials, not just their difference. Three canonical readings of an age pair with fixed gap $d = f(b) - f(a) > 0$:

• Ratio reading $m = f_t(b)/f_t(a)$: decays monotonically to $1$ (Thm 4.1). Authority, era-membership, "different generations." Fades on schedule.
• Threshold reading $m = \mathbf{1}[f_t(a) \ge \theta]$: flips once, permanently, at a computable time. Adulthood, retirement, outliving.
• Positional reading $m = $ the pair's location in a fixed order lattice: never changes, but each object's other relations accumulate around it (§6 appends). Becoming a parent while remaining a child.

Claim (informal). The dynamics of long-term relationships is exhausted by the motion of readings over the frozen gap substrate. Nothing structural moves; everything interpretive does — and the three reading types above (decaying ratios, one-way thresholds, accumulating positions) are the three characteristic time-signatures of that motion.

---

8. Extensions and open problems

1. Non-abelian gaps. Replace $(G,+)$ with any group and the cocycle identity with $\delta(a,c) = \delta(a,b)\cdot\delta(b,c)$. Everything in §1–§3 survives with care about sides. Example: relative orientation between rigid bodies ($G = SO(3)$) — a fixed gap when bodies rotate in lockstep.
2. Categorical form. A gap structure is exactly a functor from the pair groupoid of $A$ (objects $A$, one isomorphism between each pair) to $G$ viewed as a one-object groupoid; the cocycle identity is functoriality. Fixedness of $\delta$ under an evolution is naturality. What do gap structures valued in a general groupoid buy?
3. Noisy gaps. Gaps fixed only in expectation, or as martingales: $\mathbb{E}[\delta_{t+1} \mid \mathcal{F}_t] = \delta_t$. Which of Theorems 2.3 and 3.1 survive in distribution? (Candidate domain: relative skill ratings, exchange rates.)
4. Almost-fixed gaps. Slowly varying invariants ($|\dot\delta| \le \epsilon$, adiabatic invariants): quantify how much of the reconstruction theorem (Thm 3.1) degrades with $\epsilon$ — a stability-of-structuralism question.
5. $n$-ary gaps. Invariants of triples and $n$-tuples under the diagonal action (Def 5.1 with $S^n$): birth-order permutations, angles, cross-ratios. The cross-ratio — the fixed gap of projective geometry — is a 4-ary example, suggesting a hierarchy: which geometries have no fixed binary gaps at all, only higher-arity ones? Resolved in frontier.md §3: the hierarchy is the sharp-transitivity ladder — a sharply $k$-transitive group has its first invariant at arity $k+1$ (gap, ratio, cross-ratio at $k = 1, 2, 3$) — and by Tits's theorem on sharply 4-transitive groups the ladder terminates at the cross-ratio, leaving order type as the sole invariant of maximal symmetry.
6. Classifying readings. Given a fixed-gap substrate, classify all readings by asymptotic behavior. Conjecture (informal): under mild regularity, every reading of an additively drifting pair decomposes into the three time-signatures of §7 — decaying, one-way switching, and accumulating components. Resolved in readings.md: false as a trichotomy — there is an unavoidable fourth signature, recurrent readings (anniversary cycles) — but true as an amended tetrachotomy. The three conjectured types are recovered as the building blocks of all locally-BV readings, the signature is computable from the reading's symmetry (with a quantitative hyperbolic fading law: gap memory decays as $d/t$), and the recurrent class factors through a new fixed gap valued in a quotient group (the phase gap), strengthening the structuralist thesis.