The Undeclared Posit of Potential Infinity

1. Introduction

• The safe option.

Since Aristotle, the distinction between actual and potential infinity has structured the debate about the infinite [1]. Actual infinity, the completed totality, has borne the burden of controversy: the set-theoretic antinomies attach to it, the constructivist objections target it, and every foundational crisis of the modern period was litigated on its ground. Potential infinity has enjoyed the opposite reputation. It is the safe option, the concession every party makes: the intuitionist keeps the free choice sequence, the formalist keeps the unbounded supply of strings, and Hilbert’s finitism itself, the most austere of the classical programmes, operates within a potential infinity of numerals while banishing the completed one [2,3]. Even avowed sceptics about the transfinite treat the license to go on, one step at a time, as a neutral workspace rather than a commitment.

This paper audits the safe option. The audit begins where the commitments are made and hidden: at the very first move of the standard foundation, the passage from nothing to one.

• Thesis.

Three claims, in ascending strength.

First, the license is a posit, and it is undeclared. The classical passage from nothing to number is presented as a triviality: ⌀, then $\{⌀\}$, then onward. Examined honestly, the passage carries three separable commitments, none forced by logic: a structured medium in which an absence can be located; a formative operation that is ontologically productive by declaration; and an unbounded license to apply that operation again, at every level, without new commitments. The third commitment is potential infinity, stated exactly. It is not a neutral workspace. An inexhaustible license, asserted as a truth about what is possible, is an infinitary posit at the modal level: the potential infinite is the actual infinite in the subjunctive mood.

Second, the posit is empirically idle. No bounded agent of any capacity can certify that the universe of formations is infinite. Every certificate an agent can possess is a bounded record, and every bounded record is realized in all sufficiently large finite universes. Infinitude is undetectable in principle. Bounded structural facts are not. The asymmetry is a theorem, and it is proved finitarily, with no appeal to compactness.

Third, the posit has no coherent truthmaker except the finite one. If the modal claim “for every formation stage there can be a further stage” is true, something makes it true; the demand is not assumed from truthmaker maximalism but declared and defended for commitments that do foundational work (§5.1). Once it is admitted, exactly three outcomes exist: the assertion is grounded in the completed totality of all possible stages, which abandons potentialism for the actualism it was designed to avoid; or grounded in an irreducible modal realm of possibilities, which is the undeclared structured medium returned at the modal level, and whose rigorous articulation in the literature is given actualist possible-worlds semantics, so that it borrows against the first outcome; or it has no ground and is false, and the actual finite stock of formations explains its appearance. The surviving position is bounded actualism: there is one finite frame; “one can always take another successor” is false absolutely and true only below the horizon.

The constructive completion turns the verdict into an explanation. The potential infinite is the horizon phenomenon of a bounded agent inside a vastly larger finite frame. An agent of capacity K with $s^{K+1}\ll M$ never exhausts the frame’s formations and mistakes inexhaustibility-for-it for inexhaustibility simpliciter. Potentiality is the inside view of finitude, and the appearance is derived rather than explained away (§4).

• Contribution.

We make this precise in four ways.

1.

The audit. The hidden commitments of the classical opening move are separated and stated (§2): the located absence, the productive bracketing, and the unbounded license, with a three-way separability result (Proposition 1). Lemma 1 exonerates induction: Robinson arithmetic realizes iteration without induction and is already forced infinite, while every finite frame satisfies induction as a theorem. The infinitary content of the number construction tracks the iteration license exactly, and induction not at all.

2.

The idleness theorem. Under an explicit bounded-agent certification model, no agent of any capacity can certify infinitude of the universe of formations (Theorem 1), while bounded structural facts remain certifiable (Corollary 1); the symmetry with finitude is itself a theorem (Corollary 2). The proof is finitary and per-instance; compactness is neither used nor needed.

3.

The illusion theorem and the position. For a bounded agent inside a sufficiently large buffered finite frame, every formation act the agent can present succeeds, every enumeration is properly extended, and no available experiment separates the frame from an infinite universe (Proposition 3, with Lemma 2 closing the descriptive route to the frame’s edge: descriptive uniqueness is confined to the act-closure of the agent’s own trace). Bounded actualism is stated as a position (§7), its certificate conservativity over bounded practice is verified (Proposition 4), and effective determinacy, which infinite ontologies must postulate or violate, holds on the finite frame as a theorem (Corollary 3).

4.

The trilemma. The truthmaker demand is declared and defended without maximalism, the neighboring positions are sorted by the same gate, and the trilemma is argued (§5); the modal horn is examined against its articulated, essence-based, and primitive versions, with the fallback for the last stated exactly (§5.3).

• Scope and method.

The position defended descends from Aristotle’s process reading of the infinite [1] and sits within the finite-primacy tradition [3,4,5,6], while diverging from all of these in one respect made exact below: it accepts induction and audits generation. We work in classical logic and ordinary metamathematics, and we argue each elimination inside the eliminated position’s own discourse: the modal horn is examined in the potentialist’s own semantics. This is the proper use of an opponent’s resources in a reductio. No feasibility claim appears anywhere in this paper. The frame bound of bounded actualism is not a human limit, and no sorites is run against large numerals; the position is an ontological thesis, not a scepticism about long computations.

• What is proved and what is diagnosed.

The status of every claim is fixed here once. Accepted unchanged: classical mathematics, including every classical theorem cited. Proved outright: the separability of the three commitments (Proposition 1), the innocence of induction (Lemma 1), the modal ascent (Proposition 2), and effective determinacy as a theorem of the finite frame (Corollary 3). Proved within the declared certification model, unconditionally once the model is admitted: the idleness theorem and its two corollaries (Theorem 1, Corollaries 1 and 2), the uniqueness-certificate lemma (Lemma 2), the illusion theorem (Proposition 3), and certificate conservativity (Proposition 4). Declared: the locality and interface clauses of the certification model (§4) and the restricted truthmaker demand (§5.1); each is defended where declared. Argued, eliminatively: the truthmaker trilemma and the collapse of its modal horn (§5); the verdict is conditional on the declared standards and on declining unanalyzable modal primitives, and §5.1 together with §5.3 states the fallbacks exactly. Diagnosed: that the classical foundation begins from a metaphysical starting position and does not declare it; the diagnosis is the paper’s opening and its close.

2. The Opening Move, Audited

Foundations present their opening moves as too modest to count as assumptions. Consider the standard construction: $0=⌀$, $1=\{⌀\}$, and onward through the von Neumann numerals. Three commitments operate in this passage, and none of them is announced.

The target must be fixed precisely, because axiomatic set theory itself is not it. ZFC declares its completed-infinity commitment maximally explicitly, in the axiom of infinity, and thereby meets the very standard this paper enforces; the von Neumann numerals, unfolded inside ZFC, run on declared axioms throughout. The undeclared commitments live elsewhere: in the pre-axiomatic narrative that presents the construction as conjuring number from nothing and the ellipsis as costless; in the metatheory’s unexamined supply of syntax, where a potential infinity of strings is consumed before any axiom is written; and in every position that declines the axiom of infinity while keeping the license. The audit addresses those, and the declared axiom stands as its exemplar rather than its target.

2.1. The Located Absence

The construction narrative glosses the empty set as the mathematical proxy of nothingness: number conjured from the empty. Set theory’s sober view is different, a set with no members among other sets, and the audit targets the gloss exactly where the gloss does foundational work, in the claim that the construction starts from nothing. It does not. A genuine nothing cannot be named, cannot stand as a term, cannot be distinguished from anything, and cannot be operated on. The moment “⌀” is written, a something is already in place: a medium of terms in which absence can be located and pointed at. Ask where ⌀ is, and the answer is exact: in the universe of discourse, distinguishable from $\{⌀\}$, distinguishable from every other term, with a position and a web of relations. The empty set is a located absence, and a located absence presupposes the structured space that locates it. This is not a criticism of the construction. It is an inventory of what the construction uses: call it the medium.

2.2. The Productive Bracketing

The brace operation is treated as a purely formal move. But formally performing an operation and ontologically producing a being are different acts. $\{⌀\}$ is presented as a new object, available for further constructions, on equal footing with its member. Nothing inside ⌀ produces it; the bracketing does the work. What the bracketing asserts, stripped of familiarity, is a substantive principle: collections are beings; enclosing an object within a formative act creates a new object at a higher level. None of this is forced by logic. All of it is posited: call it productivity.

2.3. The Unbounded License

The third commitment is the least visible and carries the most. The formative act is taken to be applicable again, to its own output, at every level, without new commitments at any level. This is exactly the assumption “the distinction between inside and outside can be iterated indefinitely”: the iteration license. Its unboundedness is nowhere argued. It is absorbed with the notation, since the ellipsis in $⌀,\{⌀\},\left\{\right\{⌀\left\}\right\},\dots$ does not look like an axiom. Potential infinity is this license, stated exactly, and §3 makes its logical form precise.

2.4. Separability

The three commitments are genuinely three: each can be held or declined independently of the others, so none is a consequence of logic together with the rest.

Proposition 1  

(Three separable commitments). The classical construction of the natural numbers employs three commitments: (a) a medium in which terms are located and distinguished; (b) productivity of formation; (c) an unbounded iteration license. None is a logical truth, and (b) and (c) each admit a coherent position that accepts the remaining commitments and declines it. Commitment (a) cannot be declined by any position that speaks at all; it is substantive but unrejectable from within, and the vocabulary should mark the difference: (b ) and (c) are posits, declinable and therefore owing declaration, while (a) is a transcendental presupposition, owed acknowledgement rather than defense. What it shares with the posits is invisibility, and the audit’s claim on it is only that it be seen.

Argument. 

For (c): consider the theory of a finite initial-segment structure of arithmetic, a finite domain with the graphs of successor, addition, and multiplication restricted to it. A finite structure in a finite signature is described up to isomorphism by a single first-order sentence, so its theory is complete by categoricity and decidable by direct evaluation; the structure accepts a medium and finitely many productive formations and declines the unbounded license, coherently. For (b): the formalist reading manipulates the same terms while declining the claim that formation produces beings; the construction then proceeds as bookkeeping, and its coherence is not in dispute. Declining (b) is therefore coherent, at the price of ontological weight, which is the point of the dilemma below. For (a): any assertion, including the formalist’s, is made in a medium of distinguishable terms; the commitment is exercised in the act of declining anything else. It is substantive, since the medium carries structure that a genuine nothing cannot carry, and it is invisible for the same reason it is unrejectable. □

The familiar defense holds that ZFC makes no ontological claims: the symbols are symbols, and reading metaphysics into braces is a category error. The defense is internally sound and cannot be held together with the rest of what modern foundations claim. The completed infinite, the uncountable continuum, and the transfinite hierarchy have their standing through the formalism; they are not results about a prior infinity but what infinity means under these foundations. Either the formalism carries ontological weight, in which case the three commitments are real assumptions and deserve declaration; or it carries none, in which case the infinities it underwrites have no standing either. Both benefits cannot be taken at once. This paper asks only for the declaration, and then audits the declared item that remained hidden longest: the license.

2.5. Induction Is Innocent

One clarification must be made before the license is examined, because a familiar misreading would otherwise attach to everything that follows. The posit under audit is iteration, the act-license. It is not induction, the proof principle. The two are independent, and the infinitary content tracks the first.

Lemma 1  

(Induction innocence). The infinitary content of the classical number construction resides in the unbounded iteration license and not in mathematical induction. Specifically: (i) Robinson arithmetic Q contains no induction and is already forced infinite, since its successor axioms alone, an injective successor missing 0, exclude finite models; (ii) every finite arithmetic frame satisfies induction as a theorem, since the least-number principle holds for arbitrary nonempty subsets of a finite linear order, while the frame lacks the unbounded license entirely.

Proof. (i) is the standard observation about Q [7]: injectivity of successor together with $\mathsf{S}x\ne 0$ forces an infinite successor chain, with no induction axiom present. (ii): in a finite linear order every nonempty subset has a least element, by finite descent; the induction schema follows in its least-number form, for all subsets, not merely definable ones. The frame satisfies induction and contains finitely many objects. □

Iteration without induction forces infinity. Induction without unbounded iteration is finitely realized. The commitments are doubly independent, and the audited posit is the license alone. On the finite ontology defended below, induction is not a casualty; it is a theorem. This paper does not touch mathematical induction, and no reading of it as an attack on induction survives Lemma 1.

3. The License Made Exact

Definition 1  

(Formation acts and configurations). A formation act applies an operation of bounded arity to given objects and yields a new object, distinct from its inputs and available for further acts; the paradigm cases are $x\mapsto \left\{x\right\}$ and the extension of a finite arithmetic frame by one element. A formation configuration is a finite set of objects together with the finitely many acts that produced them from one another; every configuration is a finite acyclic structure.

Definition 2  

(License and supply). A license is a rule: given any already obtained object, one further formation act on it is permitted. A rule is adopted or declined; it licenses acts one at a time and asserts nothing. A supply is a proposition: the totality of objects obtainable by iterated formation exists. A supply is true or false, and if true it has a truthmaker.

The terminology is fixed accordingly for the whole paper. Iteration names repeated application under a license: generation, one act at a time. Induction names the totalizing closure over the acts: the proof principle that quantifies over all results of the iteration at once, and thereby presupposes the iterated supply as one determinate totality. Iteration is the content of the license; induction is the proof-theoretic shadow of the supply. Lemma 1 shows the two are independent and locates the infinitary commitment in the first.

Finitist practice has always used the license in exactly the rule form [2,3]: a permission to write one more numeral, each application a bounded act, checked where it is performed, with no universal closure asserted. As a rule, the license is finitarily unobjectionable. The question this paper answers is what happens when the rule is asserted, as potentialism asserts it: not “one may take a further step” but “for every stage there can be a further stage,” a truth about the space of possibilities.

Proposition 2  

(The modal ascent). Asserting the license as a true proposition is asserting a supply at the modal level. The assertion “for every formation stage there can be a further stage” quantifies universally over stages and existentially over possibilities; if true, it requires a truthmaker that outruns every finite configuration, since every finite configuration is compatible with the license failing beyond it.

Argument. 

A rule, adopted, is not truth-apt and needs no truthmaker. The potentialist assertion is truth-apt by design: it is offered as a fact about mathematical reality that distinguishes potentialism from finitism. Fix any finite configuration C. Every object and act of C is realized in a finite universe extending C by nothing (Definition 1: configurations are finite). The truth of the assertion at C therefore does not supervene on C; it constrains what lies beyond every finite configuration at once. A truthmaker with that reach is a completed range of possibilities or something that serves as one, which is the trilemma of §5. □

This is the precise sense in which the potential infinite is the actual infinite in the subjunctive mood. Taken as a rule, the license is a workspace. Taken as a truth, it is a completed totality with a modal operator in front. The remainder of the paper concerns the second reading only.

4. Empirical Idleness

The epistemology of the question is settled by an explicit model of what bounded agents can possess.

Definition 3  

(Agents and certification). An agent of capacityK is a system whose history is a sequence of records, each a string of at most K symbols over a fixed finite alphabet of size $s\ge 2$; the records are all the agent can store, survey, or verify. A sentence is certified by the agent when some retained record of its history encodes a completed check whose passing entails the sentence; distinct sentences require distinct records, output externalized beyond the record space is not the agent’s, and a schema is certified only instance by instance. Every completed check has a bounded formal trace: it mentions finitely many objects and asserts finitely many local relations among them, all within the record bound. Two stipulations are declared with the model rather than smuggled in it. The first is the locality of traces just stated; the paragraph on the derivational route below defends it against the one serious alternative and prices what it excludes. The second is the interface clause: records are the agent’s sole mode of access to its universe, so an act is the agent’s only when some retained record presents its input, and whatever the agent encounters demonstratively enters as initial data of a re-seeded configuration rather than as an anonymous handle on the frame.

Theorem 1  

(Empirical idleness of infinitude). No agent of any capacity K can certify the sentence “the universe of formations is infinite.” Every record an agent of capacity K can possess is realized in a finite universe of formations; hence no such record entails infinitude.

Proof. 

Let r be a record of length at most K encoding a completed check, and let $t\left(r\right)$ bound the number of objects its trace mentions. The trace asserts local relations among at most $t\left(r\right)$ objects of a formation configuration $C_r$, which is a finite acyclic structure (Definition 1). Let $U^{\prime }$ be any finite universe of formations obtained by performing exactly the acts of $C_r$, in order, inside a fresh finite medium, and closing under nothing further; such a universe exists for every finite configuration, and its size is bounded by $t\left(r\right)$ plus the size of the medium chosen. Every step of the check reads only the mentioned objects and their asserted relations, all of which are present in $U^{\prime }$ with the same local structure; the check therefore passes in $U^{\prime }$ exactly as it passed in the agent’s universe. So $U^{\prime }$ realizes r, and $U^{\prime }$ is finite. A record realized in a finite universe does not entail infinitude. Since r was arbitrary among possessable records, no possessable record entails infinitude, and by the certification clause of Definition 3 the sentence is uncertifiable at every capacity. □

The proof is per-instance and finitary. No compactness theorem is invoked: for each record separately, a finite realizing universe is exhibited by performing the finitely many acts the record mentions. The result is symmetric in one direction only, and the asymmetry is the content.

Corollary 1  

(The asymmetry). Bounded structural facts are certifiable and infinitude is not. For every fact of the form “the objects of configuration C stand in relations R,” with C within the agent’s capacity, there is a certifying record: perform the check and retain it. By Theorem 1 there is no capacity, however large, at which the corresponding record exists for infinitude. The obstruction is not resource shortage but reach: bounded certificates are realized finitely, whatever their length.

Corollary 2  

(Symmetric uncertifiability). Neither the infinitude of the universe of formations, nor its finitude, nor any upper bound on its size is certifiable at any capacity.

Proof. 

Infinitude is Theorem 1. For bounds, pad: let $U^{\prime }$ realize a completed check r as in that proof, and for any B let $U_B^{\prime }$ extend $U^{\prime }$ by more than B fresh formations produced from a disjoint seed. The check reads only its trace, untouched by the padding, so it passes in $U_B^{\prime }$; its passing in $U_B^{\prime }$, where the bound is violated, shows that the passing entails no bound. For finitude simpliciter, the contrast is run inside the classical semantics, since only there does an infinite universe exist to be excluded: the increasing union of the family $U_B^{\prime }$ is an infinite universe realizing the same trace, the check passes in it, and its passing therefore entails no finitude claim. The finite position of §7 needs no certificate of finitude; it declares the frame, and this corollary is its statement of the shared epistemic predicament rather than a report of its own evidence. □

The derivational route.

One apparent counterexample must be met directly, because it is the classical epistemic route to infinitude. A bounded record can encode a formal proof, from the axioms of ZFC, of the sentence “there is an infinite set”; the platonist never claimed to certify infinitude by local inspection, and claims instead to derive it. The reply is that the derivation certifies exactly what its trace contains: the conditional fact that the axioms yield the sentence. Converting conditional warrant into warrant for the conclusion requires warrant for the axioms, and the axiom doing the infinitary work is the axiom of infinity, which is the posit under audit. Offered as a certificate of infinitude, the derivation is a circle: it certifies the posit from the posit. Offered as what it is, a derivability fact, it is a bounded structural fact, certifiable under Definition 3 and realized finitely by Theorem 1 like every other. Formal derivations from an infinity axiom are therefore not counterexamples; they certify conditional derivability from the very posit in question. Warrant for the axiom claimed from extrinsic sources, indispensability or intuition, fares no better: the extrinsic credential is itself either a bounded record, finitely realized like every other, or an undeclared posit, and the audit applies to it directly. This is the content of the locality clause, now defended: what it excludes is precisely warrant whose credential is the audited assumption itself. A reader who holds that axiom-relative warrant is warrant simpliciter has adopted the posit at the ground level, and the audit of §2 applies to that adoption; the theorem’s scope is every route that does not.

Idleness is the exact charge. The potentialist supply, if true, makes no detectable difference to any agent at any capacity: every observation, computation, and proof any agent ever possesses is one a large finite universe supports identically. A posit whose truth is undetectable in principle by every possible bounded agent earns its keep, if at all, elsewhere than in mathematical practice. The next section asks whether it earns its keep in metaphysics, and the answer is that it has nowhere to stand.

Before that, the positive face of the same computation. Two notions must be fixed first, because the result fails if they are left informal. Say that an object of the frame is reachable for the agent when it belongs to the agent’s initial data or results from at most $s^{K+1}$ formation acts performed by the agent on reachable objects. Say that the agent presents an object as the input of an act when some retained record identifies it: either by provenance, as the result of a recorded act sequence, or by description, via a completed check entailing that exactly one object of the frame satisfies the describing condition. Cardinality alone does not make a frame safe: a short definite description such as “the maximal formation” can denote a boundary object of an arbitrarily large frame, and an agent seeded next to the boundary of a long chain reaches it in a few acts. Both gaps close, the first by a lemma, the second by a hypothesis.

Lemma 2  

(No bounded uniqueness certificate). Under Definition 3: no completed check of an agent of any capacity entails, of any describing condition, that exactly one object of the universe satisfies it, except where the entailed satisfier lies in the act-closure of the check’s own trace. Consequently every presentable object is reachable, whether presented by provenance or by description; and in a frame that buffers the agent, in the sense of Proposition 3, no formation failure is ever certified.

Proof. 

Let r encode a completed check alleged to entail unique satisfaction of a condition D. Two cases, by whether D carries parameters.

Case 1: D is parameter-free, hence isomorphism-invariant. Realize the trace of r in a finite universe $U^{\prime }$ as in the proof of Theorem 1, and let $U^{\prime \prime }=U^{\prime }\bigsqcup U^{\prime }$ be the disjoint double. The check reads only its trace, present in the first copy, so it passes in $U^{\prime \prime }$. The swap of the two copies is a fixed-point-free automorphism of $U^{\prime \prime }$, so the satisfiers of any isomorphism-invariant condition form a swap-closed set and come in pairs: zero or at least two, never exactly one. So the check passes in a universe where no invariant condition has a unique satisfier, and its passing entails no uniqueness claim.

Case 2: D carries parameters, which under the interface clause are provenance-presented objects. Take the minimal realization $U^{\prime }$: the universe obtained by performing exactly the trace’s acts and nothing else, so that $U^{\prime }$ contains only the trace’s objects and their act-closure. Any entailment of the check’s passing must hold in $U^{\prime }$, since the check passes there. If the entailed unique satisfier exists in $U^{\prime }$, it lies in the act-closure of the trace, and the frame’s satisfier is obtained from the same provenance-presented parameters by the same bounded chain of formation steps, hence reachable. If instead D points outside the trace’s productions, the entailment fails already in $U^{\prime }$: for a backward condition such as “the object whose formation image is the datum x,” the datum is realized fresh in $U^{\prime }$ with no producing act unless the trace contains one, the condition has zero satisfiers there, and the passing entails nothing. Descriptive uniqueness is therefore confined to trace-derivable objects.

Three consequences deserve stating. Syntactic brevity buys nothing: “the maximal formation” is written in a few symbols, but writing is not presenting, and presentation requires entailed unique denotation, which Case 1 refuses to every parameter-free description; in $U^{\prime \prime }$ there are two maximal formations. Enriching the signature with a global predicate does not help either: the predicate’s extension becomes one more thing the agent’s records must certify, and the double realizes those records with the extension duplicated. And parameters do not open a route to the boundary: Case 2 confines every parameterized identification to the act-closure of what the agent has itself produced. Presentation, by provenance or by description, therefore yields reachable objects only. For the last clause: a certified failure would be a record whose passing entails that a presented object lacks a formation image; its presented object is reachable, and under the buffer hypothesis of Proposition 3 every reachable object has its image in the frame, so no such record’s check can pass. □

Proposition 3  

(The horizon illusion). Let an agent of capacity K inhabit a finite frame containing M formations, with $2s^{K+1}<M$, and let the frame buffer the agent: the formation image of every reachable object exists in the frame. Such frames exist at every sufficiently large size; close a seed configuration, with the agent’s initial data among the seed, under D rounds of formation with $D>s^{K+1}$. Then: (i) every formation act on a presentable object succeeds; (ii) every enumeration the agent runs is properly extended in the frame, since its length is bounded by the record capacity; (iii) no check available to the agent distinguishes the frame from an infinite universe. The agent’s situation therefore satisfies, in full, the phenomenology expressed by “one can always continue.”

Proof. (i): by Lemma 2 every presentable object is reachable, and the buffer hypothesis supplies its formation image in the frame. (ii): an enumeration is a retained record; its length is at most K symbols’ worth of entries, and the buffer supplies a proper extension of the enumerated configuration. (iii): a distinguishing check would have to entail a sentence whose truth value separates the frame from an infinite universe: infinitude, excluded by Theorem 1; finitude or an upper bound, excluded by Corollary 2; or a witnessed formation failure, excluded by Lemma 2. The trichotomy is exhaustive by the padding construction itself: every completed check passes in realizing universes of every sufficiently large size and in their union, so its passing entails nothing that separates them. □

Potentiality is the inside view of finitude. A bounded agent in a large finite frame must experience exactly what the potentialist reports: every attempted step succeeds, no last step is ever met, and nothing available ever witnesses an end. The report is accurate as phenomenology and unwarranted as ontology; it mistakes inexhaustibility relative to the agent for inexhaustibility simpliciter. The illusion is not an error the agent commits; it is the necessary appearance of a horizon from within. What Proposition 3 adds to the familiar thought is derivation: the appearance is a theorem of the finite position, not an embarrassment it explains away.

5. The Truthmaker Trilemma

Suppose the potentialist assertion is true: for every formation stage there can be a further stage. By Proposition 2 the assertion, if it has a truthmaker at all, requires one beyond every finite configuration. The demand for a truthmaker is itself a commitment, and it is declared and defended first; once it is admitted, the assertion faces three exhaustive outcomes under the declared carrier demand: grounded in a completed range, grounded in a modal carrier, or ungrounded and false, with the finite stock explaining its appearance. The third is not a truthmaker for the assertion; it is the outcome in which the assertion has none, and this is exactly why it survives the audit that eliminates the first two.

5.1. The Demand

The paper does not assume truthmaker maximalism, the thesis that every truth has a truthmaker [8]; modal truths are the canonical contested class in that literature, and resting the argument on the general thesis would beg the question against the non-maximalist. The demand made here is narrower and independently motivated: a foundational commitment that does ontological work owes an account of what carries it. The potentialist assertion does such work by design. It is what separates potentialism from finitism; it is what underwrites the modal reconstruction of classical mathematics; the position’s entire content over and above the rule is this assertion. A commitment can decline the demand only by declining the work.

The deflationist reply, that “for every stage there can be a further stage” is true without anything making it true, is therefore answered on the paper’s declared standards rather than refuted from maximalism. A truth with no truthmaker, no operational content, and no detectable difference at any capacity (Theorem 1) has, for every agent and every practice, the exact profile of no truth at all; what survives the deflation intact is the rule, which this paper endorses and does not audit. A reader who nonetheless holds the deflated assertion holds something the paper cannot price further, and the layering applies: the audit, the idleness theorem, and the illusion theorem stand; only the verdict is declined. The demand, so restricted and so defended, is the declared gate of the trilemma.

The gate also sorts the neighboring positions, and the sorting is an argument, not an assertion. Fictionalism, inferentialism, and proof-theoretic constructivism read the license as a rule of a practice, useful, meaning-constituting, or proof-generating, and assert no modal fact; the audit does not target them, because on their reading nothing has been posited, and the paper’s only demand of them is that the reading be kept honestly, without borrowing the assertion when foundational weight is needed. Modal structuralism [9] and indefinite extensibility, on its modal reading [10], assert the modal fact and articulate it; they enter the trilemma at horn (ii). What no position on the map does is a fourth thing: assert the fact, accept the demand, and produce a carrier that is neither a completed range nor a modal realm; and none makes the assertion true with no carrier at all, since that is the deflation already dispatched.

(i)

The completed totality of all possible stages. The possibilities are made true by the actual existence of their completed range: an actually infinite structure of stages.

(ii)

An irreducible modal realm. The possibilities are primitive: a standing space of unactualized formation, not reducible to any actual structure.

(iii)

The actual finite stock. What exists is a finite frame of formations; the license, as a rule, is exercised within it; and the universal modal assertion is false absolutely, true only below the horizon.

5.2. Horn (i): Not a Defense but a Surrender

Horn (i) grounds the modal supply in a completed, actually infinite range of stages. Whatever its merits, it is not a defense of potential infinity; it is the abandonment of potentialism for the actualism the view was designed to avoid. The founding motivation of the potentialist tradition, from Aristotle’s rejection of the completed infinite to the modern revival, is precisely that the totality of stages does not exist as a finished object [1,11]. A potentialist who takes horn (i) has exited the debate, and inherits the classical liabilities of completed infinity wholesale: the antinomies attach to absolute totalities, and the completed range of all possible stages is as absolute as totalities come; a reductio analysis of those liabilities is developed in [6]. Two further points close the horn on its own terms. Theorem 1 applies verbatim: a completed infinite range of stages is exactly as uncertifiable by any bounded agent as the potential supply was, so the ontological escalation purchases no epistemic ground. And the audit of §2 applies verbatim as well: a completed totality of possible stages is a medium of located, distinguished, unactualized objects, posited whole, and its declaration is precisely what the classical foundation never made.

5.3. Horn (ii): The Modal Realm

Horn (ii) is the live one, and it divides into two positions by how much articulation the modal realm receives.

The articulated version.

The rigorous modern articulation of potentialism interprets the modal operators over a space of worlds or stages, with the potentialist hierarchy unfolded along an accessibility relation [11,12]. The semantics that makes the view precise quantifies over the completed collection of worlds one level up: the modal facts are made true by an actualist structure of possibilities. This is not an incidental feature of presentation. It is how the view earns its theorems, since the mirroring results that transfer classical mathematics into the modal idiom are proved in the actualist metatheory. Articulated potentialism therefore borrows its truthmaker from horn (i): the completed totality returns, one level up, wearing a modal operator. The audit of §2 applies verbatim: a standing space of possibilities in which unactualized stages are located and distinguished is the medium of §2.1 at the modal level, and it is undeclared in exactly the same way. Nor is the semantics harmless scaffolding: if the worlds structure is a mere instrument, the mirroring theorems’ truth needs another ground, and the trilemma reapplies to that ground one level up; §6 (O6) develops both this and the schematic counter-move.

The essence version.

Between the articulated and the primitive positions sits a third: the possibility is grounded in the nature of the formation operation, an essence-based truthmaker in the style of [13], a finite carrier with infinitary reach. The fold is explicit: this is horn (ii) under another name. “The nature of formation is inexhaustible” is the license-as-truth again, its ground now labelled essence, and the audit’s question recurs unchanged, namely what, in a finite operation’s nature, carries reach beyond every finite configuration. Either the essence is articulated, as a range of possible manifestations, and the completed range returns as in the articulated version; or it is left primitive, and the position joins the primitive version below, three-step reply and all. The essence label adds a name, not a carrier.

The primitive version.

The remaining escape refuses semantics altogether: the modality is primitive; it needs no possible-worlds articulation and borrows nothing [14]. Three replies, in ascending strength.

First, the declaration standard. The audit’s demand is not zero commitment; it is declaration. A primitive inexhaustible modality is a structured medium of possibility, posited whole, and declined analysis. It consumes the license while refusing to state what carries it. Between a declared finite posit and an undeclared inexhaustible one, the methodological comparison is not close, and it is the foundational literature itself that insists starting points be declared.

Second, the operational-content standard. Existence claims carry content where decisions and realizations are producible; a possibility that no construction ever realizes, whose truthmaker is refused, and whose obtaining is undetectable by every bounded agent (Theorem 1) is, for every agent and every practice, indistinguishable from no possibility at all. To found mathematics on that indistinguishability is a choice, and it should be defended as one. The primitive escape concedes the operational question entirely and retreats to bare assertion.

Third, the cost comparison. Bounded actualism posits one finite frame and declares it. Primitive potentialism posits an unanalyzable modal fact of infinite reach and does not declare its carrier. The declared posit additionally derives the opponent’s phenomenology: Proposition 3 yields “one can always continue” as a theorem about bounded agents in large finite frames. The undeclared posit explains nothing the declared one does not, is undetectable where the declared one is exhibitable, and costs an infinitary commitment where the declared one costs a finite fact. If the standards themselves are contested, the situation reduces to a priced underdetermination, and the paper’s verdict becomes conditional on declining unanalyzable primitives; the audit, the idleness theorem, and the illusion theorem are untouched in every scenario. That is the fallback, stated once and not repeated.

5.4. Horn (iii): The Survivor

What remains is the third outcome. The actual stock of formations is finite; the license is real as a rule and exercised within the stock; the universal modal assertion has no truthmaker and is false. What is true in its vicinity is exactly what Proposition 3 delivers: below the horizon of any bounded agent, every step succeeds, and the appearance of inexhaustibility is guaranteed. Horn (iii) is not a residue; it is a position with a name, and §7 states it.

6. Objections and Replies

• (O1) “You posit the frame; the audit applies to you.”

It does, and the posit is declared; that is the entire asymmetry. The audit’s standard is declaration rather than emptiness: a starting point that is substantial may still be the only way to get started, and what cannot be claimed is that there is no starting point. The choice on the table is one declared finite posit against an undeclared inexhaustible modal supply, and the declared posit derives the appearance of the alternative (Proposition 3).

• (O2) “Mathematics needs unbounded induction.”

Two layers. First, Lemma 1: this paper does not touch induction. On finite frames induction is a theorem, and what the finite position declines is not induction but the universal closure over a completed supply that full induction licenses; each instance survives, finitely verifiable (Proposition 4). Second, the positioning against Nelson is exact and worth stating: predicative arithmetic doubts induction as impredicative while accepting generation [15]; the present position accepts induction and audits generation. The two programmes are inverses, and the present one leaves standard proof practice untouched.

• (O3) “The frame bound is unknowable; this is mysticism.”

The unknowability is a pigeonhole fact, not a mystery. A part with fewer distinguishable states than the whole cannot represent the whole injectively; an agent of capacity K commands fewer than $s^{K+1}$ records, so no agent inside a frame that dwarfs its record space can certify the frame’s cardinality. Of course the bound is not certifiable from within; if it were, the part would mirror the whole. The bound is real and determinate; its value is not an object of any part’s knowledge; and this is a predicted signature of the position, not a defect discovered in it.

• (O4) “This is ultrafinitism, and it inherits the vagueness objection.”

It is not, and it does not. No feasibility claim is made; the frame bound is not a human limit; no sorites is run against exponentiation [16,17,18]. The position is bounded actualism about ontology. Arbitrarily large finite structures below the bound are embraced without discomfort, and the bound itself enters no arithmetic argument, precisely because (O3) places it beyond every internal computation.

• (O5) “The illusion story is unfalsifiable.”

The symmetry is a theorem, not a posture: neither infinitude, nor finitude, nor any bound on the universe’s size is certifiable at any capacity (Corollary 2), so neither party wins by observation. The paper claims no empirical support for finitude; it proves empirical idleness of the opponent’s posit, which is a different and stronger thing. After idleness, the selection is made where foundational selections are made: at the tribunal of declared posits and costs (§5.3).

• (O6) “Modal potentialism is being attacked through its semantics rather than its content.”

The articulated version is engaged exactly where its content lives: the mirroring theorems that give the view its mathematical interest are proved in the actualist metatheory, so the semantics is not packaging but load-bearing [11,12]. Two escapes must be closed. The scaffolding escape holds that the worlds structure is a model-theoretic instrument carrying no commitment; but an instrument proves theorems whose truth then needs a ground the instrument no longer supplies, and the trilemma reapplies to that ground one level up, with nothing gained by the ascent. The schematic escape runs this paper’s own move in reverse: since Proposition 4 lets bounded actualism read universal closures as schemas, the potentialist proposes to read the actualist metatheory schematically, instance by instance, kicking away the worlds ladder. The disanalogy is exact and decisive. Bounded actualism’s schema instances are literally finite checks, each realized inside the frame; the potentialist’s schematic instances of the mirroring theorems each quantify over unbounded stage-structures, so every single instance still carries the infinitary commitment, and schematizing has relocated the supply into each instance rather than discharging it. The primitive version is engaged separately and on its own terms (§5.3), with the fallback stated for a reader who accepts unanalyzable primitives.

• (O7) “Any particular M is an unmotivated free parameter; unboundedness is parameter-free.”

This is the classic objection to every finite ontology, and it deserves the direct reply. First, unboundedness is not an exemption from the question; it is one admissible answer to the same question, its admissibility is exactly what is under audit, and invoking its parameter-freeness against finitism therefore assumes the point. Second, parsimony counts the strength of brute commitments, not only their number: a brute finite contingency is a weaker posit than a brute infinite one, and determinateness without a selection principle is not a defect but the general condition of contingent fact; nothing selects the actual number of anything, and only in foundations has this ever been thought to need apology. Third, Theorem 1 converts the comparison into an asymmetry: the allegedly parameter-free setting is the one setting no agent can ever certify, while every finite setting is certifiable in the bounded facts it supports. A free parameter whose every finite value is exhibitable competes well against a fixed value that is undetectable in principle.

7. Bounded Actualism

The position that survives is stated in one place.

• The position.

There is one finite frame of formations. Its cardinality is determinate and not certifiable by its parts. The formation license is real as a rule and is exercised within the frame; each application is a bounded act. The universal potentialist assertion is false; its phenomenology is derived (Proposition 3). Induction holds on the frame as a theorem (Lemma 1). Mathematics below the horizon proceeds verbatim; unbounded closures are read as schemas, each instance finitely verifiable.

Proposition 4  

(Certificate conservativity of bounded practice). Under bounded actualism, every actually possessed finite certificate, computation, derivation trace, and applied instance is preserved. (i) Every statement about a fixed finite configuration holds or fails exactly as classically, since the frame contains the configuration and the check is the same bounded computation. (ii) Every universally quantified arithmetic generalization is retained as a schema: each instance with parameters below the frame bound is finitely verifiable, and every verification any agent has ever performed is such an instance (Corollary 1).(iii) Induction is available as a theorem (Lemma 1) , and proof practice that uses it instance-wise transfers verbatim. (iv) Derivations that pass through infinitary principles, compactness, completions, nonconstructive existence, are preserved as what their traces contain, derivability facts, and their applied consequences fall under (i) – (ii) ; the infinitary principles themselves are reinterpreted, not preserved verbatim, and this is a cost the position declares rather than hides. What is relinquished is the universal closure read as a single truth about a completed supply; and by Theorem 1 no agent ever possessed a certificate for that reading.

Proof 

(Verification). (i) is immediate: the same finite check, run in the same way. (ii): an actually performed verification is a bounded record; its trace involves finitely many objects, all realized in the frame when the frame exceeds the trace bound. (iii) is Lemma 1(ii) together with the observation that an instance-wise use of induction is a bounded derivation over the objects it mentions. □

Corollary 3  

(Effective determinacy as a theorem). Say that a totality is effectively determinate when it decides every well-posed question about itself, its subsets and relations included, and each decision is effectively producible. Infinite ontologies must either postulate effective determinacy of their totalities or violate it, since the full second-order theory of an infinite structure, with quantification over all subsets and relations, admits no effective presentation; first-order fragments can be decidable, and the claim is not about them. On the finite frame, effective determinacy holds as a theorem: the subsets and relations of a finite structure are finitely many and listable, and its full theory is decided by exhaustive evaluation. Two registers must be kept apart here, and one sentence keeps them apart: determinacy is an ontological property of the structure, and the evaluation is effective relative to the structure presented as a finite object; no claim is made that an agent inside the frame performs it, and (O3) prices exactly that impossibility. The finite ontology is the one ontology whose determinacy costs nothing.

Proof. 

The finite direction is direct: every second-order quantifier over a finite structure is a finite conjunction or disjunction over the listable subsets and relations, so evaluation terminates and is uniform in the structure. For the infinite case, work over the classical background in which every infinite domain contains a Dedekind-infinite substructure; such a substructure supplies a successor structure, and full second-order quantification defines it and pins a categorical copy of the natural numbers by Dedekind’s argument [19], so the full second-order theory interprets true second-order arithmetic and is not recursively enumerable; no effective presentation exists. □

• Costs, stated without decoration.

Three, all inherited knowingly. The universal reading of arithmetic generalizations is relinquished in favour of schema form; Proposition 4 prices this exactly, and practice below the horizon does not feel it. The frame bound is unknowable from within; (O3) converts this into a predicted signature. And the position’s verdict layer rests on declining unanalyzable modal primitives; the fallback of §5.3 keeps the unconditional layer standing for a reader who will not decline them.

8. Conclusions

The safe option was never safe; it was unexamined. The license to iterate formation without bound is a posit, hidden in the notation of the very first move from nothing to one (§2); asserted as a truth, it ascends to a modal supply (Proposition 2); as a supply it is empirically idle, uncertifiable by any bounded agent at any capacity (Theorem 1); and its truthmaker options close one by one, leaving the finite stock (§5). Induction is untouched throughout; the finite ontology proves it rather than mourning it (Lemma 1). The potentialist’s experience is honoured in full and explained: one can always continue, below the horizon, inside a frame one can never exhaust or measure (Proposition 3).

What remains standing is one declared posit: a finite frame, determinate, complete as a finite object, larger than any of its parts can certify. Every certificate mathematics has ever possessed survives on it unchanged (Proposition 4), determinacy holds on it for free (Corollary 3), and it changes only what mathematics claims to be about. The foundational question that opened the audit, what else might be posited in place of the classical starting points, receives its answer: posit the frame, declare it, and derive the appearance of infinity as the view from inside.