Abstract
Contemporary foundations of mathematics is a closed shop. ZFC, HoTT, univalent foundations, category-theoretic foundations — the programs argue about which mathematical primitives are most elegant while sharing a meta-commitment none of them examine: that mathematics is the floor. This paper shows that the meta-commitment is false, that a working alternative now exists, and that the alternative changes what the foundational conversation is about. The Unified Conceptual Framework / Grand Unified Tensor Theory (UCF/GUTT) derives the number tower ℕ → ℤ → ℚ → ℝ from a relational substrate that is prior to mathematics, under a machine-verified zero-axiom discipline in the Rocq (Coq) proof assistant. The Layer Distinction — the public statement of this foundational posture — is not another proposal for the pluralist ecology. It is the claim that the pluralist ecology has been arguing one level too high. I distinguish substrate-level from content-level foundations, identify three criteria the substrate-level program satisfies, and dismantle the three objections the field will raise against it. The philosophical lineage of substrate-level thinking — Whitehead, Peirce, Cassirer, ontic structural realism — has been right and disarmed for a century. The disarmament is over. Machine verification is the operational turn that converts substrate-level claims from metaphysical speculation into foundational engineering, and the field's response to this turn will be the measure of whether foundations is a serious inquiry or a closed rhetorical game.
1. The Closed Shop
Foundations of mathematics in the contemporary period looks like a live debate. It is not. It is an intramural squabble among partisans who share a premise none of them examines. Set theorists argue with type theorists about whether sets or types are the better primitive. Categorists argue with both about whether categorical or elementary-topos-theoretic foundations are more natural. Univalent foundations proposes that homotopy types are the right primitive and that univalence is an axiom worth accepting or, in cubical variants, proving. Structural realists in mathematics argue about ante rem, in re, eliminative, and modal versions of structuralism. Reverse mathematics measures exactly which subsystems of second-order arithmetic are required for classical theorems.
All of this is sophisticated. Almost none of it is foundational in the sense the name implies. The shared premise across every program listed is that mathematics grounds itself — that the foundational task is to find the most parsimonious or theoretically attractive mathematical primitives and show the rest of mathematics emerges from them. This premise is held so fixed that it rarely appears as a thesis. It is the water the fish don't see.
The premise is false. Or more precisely: a working alternative now exists, and its existence demonstrates that the premise was an assumption rather than a necessity. The Unified Conceptual Framework / Grand Unified Tensor Theory (UCF/GUTT) derives the number tower from a relational substrate that is not itself a mathematical object. The derivation is machine-verified in Rocq under a zero-axiom discipline, enforced mechanically by a library-wide axiom audit that reports "Closed under the global context" for every declared theorem in scope. The tower is not constructed within mathematics. It falls out of what the substrate does when you take seriality seriously.
I will call a foundational program content-level if its primitives are mathematical objects and its target is a mathematical corpus. Set theory, type theory, category theory, and structural foundations are all content-level in this sense. I will call a program substrate-level if its primitives are claimed to be prior to mathematics — not mathematical objects but relational structure from which mathematical objects emerge as stabilizations — and its target includes both the corpus and the mathematical scaffolding that corpus requires. UCF/GUTT is the first substrate-level program to be machine-verified under zero-axiom discipline, and its public statement of this posture — the Layer Distinction essay recently published on the framework's site — is the event that makes the substrate/content distinction live.
The field will not welcome this. It does not have to. The files are private, the audit is mechanical, and the argument is stated. The slow update will happen because the alternative — continuing to argue about which mathematical primitives are most elegant while a working substrate-level alternative sits on the table — is not defensible. The rest of this paper is the argument the field will have to engage when it stops pretending it doesn't have to.
2. Substrate-Level and Content-Level Are Not The Same Kind Of Thing
The temptation, for anyone encountering the substrate/content distinction for the first time, is to file it as another option in the pluralist ecology. Another program for the list; ZFC, HoTT, topos-theoretic foundations, substrate-level foundations. Four options instead of three. The taxonomy has been updated; carry on.
This is the wrong response, and it is wrong in a way worth stating directly.
Content-level programs derive mathematics within mathematics. Their primitives are mathematical objects. Their derivation medium is mathematics. Their target is mathematics. The foundational move, in every content-level program, consists in finding a subset of mathematics from which the rest of mathematics can be recovered. The meta-level posture is mathematics-as-self-grounding, and the programs differ only in which subset they pick.
Substrate-level programs derive mathematics from outside mathematics. The primitives are not mathematical objects. The derivation target includes the mathematical scaffolding on which the corpus rides. The foundational move is not to find a parsimonious mathematical primitive but to identify what mathematics itself is an expression of. The meta-level posture is the inverse of content-level work: mathematics is not the ground floor but the first upper floor, and the ground floor is something else.
These two kinds of program are not more-and-less ambitious versions of the same enterprise. They are different enterprises. A content-level program at its most ambitious achieves maximal parsimony within the content level; it does not, because it cannot, answer what mathematics itself is an expression of. A substrate-level program at its most modest still operates at a different layer. The two cannot be compared by theorem count or proof-theoretic strength, and anyone who tries is making a category error.
The correct way to hold the distinction is this.. in my view... Content-level foundations answers: given that we are doing mathematics, what are the most parsimonious mathematical primitives from which the rest emerges? Substrate-level foundations answers a different question: why is there mathematics at all, and what is the structure of the reality that produces it? The first question is technical. The second question is ontological. Both are legitimate. But the second question cannot be answered from within the first, because the first presupposes the existence of mathematical practice as its medium of operation.
Wigner's 1960 question — why is mathematics unreasonably effective in the natural sciences — is a second-question question. It cannot be answered at the content level, which is why content-level foundations has never answered it. The question's force comes from asking why the medium of mathematical practice should apply to a physical world that did not have to be mathematical, and any answer that stays inside the medium presupposes what the question is asking about. A century of mathematicians and philosophers have shrugged at Wigner's question because the tools to answer it did not exist. The tools now exist. The answer is not mysterious. Mathematics is effective in describing the physical world because both mathematics and the physical world are stabilizations of the same relational substrate, at different structural depths. The Layer Distinction is the shape this answer takes when it is stated in its own proper register.
3. The Philosophical Lineage Was Right and Disarmed
The idea that relation is ontologically prior to object is not new, and none of what follows should be taken as claiming it is. What is new is the operational form, and the difference between "new idea" and "newly operationalized idea" is the difference between a philosophical position and foundational engineering. Both matter. The second is what the field has to engage with now; the first is why the second could be attempted at all.
Whitehead got this right in 1929. Process and Reality is explicit that process and relation are prior to substance, that objects are "societies" of actual occasions exhibiting stable relational patterns, and that mathematical structure is a specific kind of stable pattern — eternal objects that actual occasions may instantiate. The view is systematic, and it is correct as far as it goes.
Where it does not go is to a derivation of arithmetic, analysis, or anything else from the process-relational substrate. Whitehead's account is philosophical; the derivations are gestural. This was not a failure of rigor. It was a limit of the tools available in 1929. A philosopher writing before the existence of proof assistants could argue that relation is prior without being able to show, mechanically, that the priority is consistent with the machinery it is supposed to ground.
Peirce got this right earlier, in his own vocabulary. Firstness, Secondness, and Thirdness are categories of relational structure; Thirdness in particular is where mathematical content is supposed to emerge. The semiotic account of continuity gestures at exactly the kind of stabilization a substrate-level program needs. The gesture is what it is: a gesture. No derivation is provided, because no mechanism for rigorous derivation from the triadic primitive was available to Peirce.
Cassirer got this right at the level of scientific methodology in Substanzbegriff und Funktionsbegriff (1910). The argument is that modern science, from the calculus onward, has already adopted a functional (relational) primitive over a substantial one, and that philosophy has not caught up. Cassirer was arguing with the philosophy of his time; he was not arguing with contemporary foundations, because contemporary foundations in its current form did not yet exist.
Ontic structural realism — Ladyman and Ross (2007), French (2014) — makes the strongest recent case that the ontological primitives of physics are relational. It is a live position in philosophy of science. It has not led to a substrate-level foundational program because its adherents have not had, or have not pursued, the tools to derive the mathematical scaffolding of physics from the structural primitives they propose. The work has remained philosophical; the scaffolding has remained imported.
This is the story of substrate-level thinking across the twentieth century and into the twenty-first. The position is correct; the tools for discharging it rigorously did not exist. Philosophers knew they were disarmed. The field that would have had to engage with them — contemporary mathematical foundations — did not engage, partly because the philosophers were disarmed and partly because the field had reason to prefer content-level work, which was tractable. This was a reasonable equilibrium under the constraints that obtained.
Proof assistants break the equilibrium. A substrate-level claim that can be expressed in Rocq or Lean or Agda, and whose derivation compiles under a zero-axiom audit, is not a philosophical claim that the field can file under "metaphysics" and ignore. It is a proof that the derivation exists.
Whether the primitives are genuinely pre-mathematical is now the sharp question, not the vague one. "Is relation prior to object?" has always been a philosophical question. "Does this specific library, with these specific primitive declarations, derive ℕ → ℤ → ℚ → ℝ under the constraints its axiom audit enforces?" is a mechanical question with a mechanical answer. The audit either passes or fails. The primitives either survive inspection as relational or they do not. The philosophical tradition's century of arguing for relational priority was the necessary preparation for a programmatic turn it could not itself execute. The turn has now been executed. The arguing is over. The inspecting begins.
4. The Layer Distinction Is The Instance
UCF/GUTT's Layer Distinction essay is not a case study of substrate-level foundations. It is substrate-level foundations, publicly stated. Treating it as a "case study" would imply other instances to study and a category to calibrate; there are no other instances. What the essay marks is the moment the category becomes populated.
The essay's central move is to distinguish substrate-level from content-level in terms that survive adversarial reading. It names the content-level programs from which it distinguishes itself: ZFC, dependent type theory, univalent foundations, topos-theoretic foundations. It names the philosophical precursors whose work it operationalizes. It states the asymmetry directly — "other programs derive the tower inside mathematics; UCF derives both the tower and the mathematics from something prior" — and it identifies where the claim can be pressed: the question is whether the substrate is genuinely pre-mathematical.
The essay's most important methodological feature is the caveat structure. It distinguishes eligibility claims (the substrate removes the importation problem in principle) from completion claims (specific derivations are complete, and here are the file references). It states that "in principle reachable" is not the same as "completed," that specific construction work is still required, and that the Marcus chain of proof files is what end-to-end derivation actually looks like when the construction work is done.
This caveat structure is not a concession. It is a strategic feature. A triumphalist statement would hand skeptics an obvious target: the gap between the framework's large claims and its current completion. The essay pre-closes that target. A reader who wants to press the framework is not given a strawman to dismantle; they are given an already-accurate statement of where the framework is and is not yet complete, and invited to press on the specific derivations rather than on the general framing. The effect is to shift the conversation from "is this framework overclaiming?" to "are the specific derivations it points at sound?" — which is the conversation the framework can win, because the derivations are what they are.
Two further features deserve note. First, the essay does its philosophical work explicitly rather than implicitly. The load-bearing claim is not buried in the middle of a technical exposition; it is the thesis statement, placed where a reader cannot miss it. Second, the essay's relationship to the technical corpus is cleanly structured: the essay states the methodological consequences of what the proof library contains, and links to the library for readers who want to verify. This separation means the essay can be read by philosophers of mathematics who do not read Rocq, and the library can be read by verification experts who do not need philosophical framing. Each layer is responsible for what it can actually deliver.
What the essay demonstrates, taken as a whole, is that substrate-level foundations can be presented in a form the field can engage with. It is not an esoteric manifesto. It is an academic-register statement of a technical-philosophical position, with its pressure points explicitly marked and its caveats explicitly stated. The field has been given no excuse to dismiss the program as poorly articulated. The only remaining reasons to dismiss it are reasons at the content level: either the primitives fail the pre-mathematical test, or the derivations fail the audit. Both are questions with answers, and both answers are publicly checkable.
5. Three Criteria The Framework Satisfies
Substrate-level foundations require evaluative criteria specific to them. Content-level criteria — parsimony of primitives, proof-theoretic strength, faithfulness of reconstruction — are the wrong tests, because they apply at a level the substrate-level program has already passed through. What follows are three criteria appropriate to substrate-level work, and for each I will note where the UCF/GUTT library already satisfies it.
5.1 Importation Removal
The first criterion is whether the program actually removes the importation problem it claims to remove. A substrate-level program that imports mathematical structure at the primitive level is a content-level program with a relational gloss; it has not done what it claims to do. The test is mechanical: the axiom audit must report zero axioms beyond the kernel's type theory for all theorems in scope, and the primitive declarations must survive inspection as pre-mathematical.
UCF/GUTT's AxiomAudit.v discharges the first half of this test. The file imports every module in the library, runs Print Assumptions on over a hundred declared theorems, and compiles if and only if each report is "Closed under the global context." The audit is not a stylistic preference. It is the compile-time discharge condition of a specific file, and the claim of zero-axiom closure stands or falls at the moment that file is type-checked.
The second half of the test — whether the primitives are pre-mathematical — is philosophical rather than mechanical, and it is the place the framework genuinely invites adversarial engagement. The RelationalGrounding.NaturalsInUniverse section takes U : Type and R : U -> U -> Prop as variables. A skeptic can argue that U : Type is already mathematical structure. The response is: the encoding is the medium, not the premise.
Philosophical arguments in English do not reduce to claims about English grammar; foundational derivations in Rocq do not reduce to claims about Rocq's type theory. The substrate-level commitment is about what the primitives are intended to be and can be instantiated as: any carrier, any binary relation, with no further mathematical structure presupposed. The medium is not required to vanish; the medium is required not to smuggle in content. Whether the medium smuggles content is answered by examining the primitive declarations and the axiom audit, both of which are public.
A sophisticated skeptic can press further: the very notion of "any carrier" and "any relation" presupposes the notion of universal quantification over types, which is mathematical. At some sufficiently recursive depth, the objection becomes the claim that no formal system can ever be genuinely pre-mathematical, because formality itself is mathematical. This objection, if pressed all the way down, is not an objection to substrate-level foundations specifically; it is a general skepticism about the possibility of rigorous ontology. It is a defensible position, but it has the property of being equally effective against every formal foundational proposal, which means it does not distinguish substrate-level from content-level work. It proves too much to prove what it needs to prove.
5.2 Derivation Depth
The second criterion is whether the program reaches deeper targets from a fixed set of primitives, or whether it proliferates primitives as new domains are attempted. A genuine substrate-level program should accumulate theorems without accumulating axioms. If each new domain requires new foundational commitments, the program is not substrate-level; it is domain-by-domain modeling with a substrate-level marketing line.
UCF/GUTT satisfies this criterion concretely. The same relational substrate that grounds N_rel_serial also grounds the Bishop-style Cauchy reals, the constructive R_sqrt2 witness, the three-state division-with-tagging of RelationalState, the QM/GR recovery as T¹ and T³ diagonal restrictions, the SU(3) × SU(2) × U(1) gauge-group uniqueness, the Planck-constant emergence from discrete relational structure, the Yang-Mills mass gap with microcausality via the tensor commutativity theorem, the CPT theorem from relational Lorentz invariance, and the Marcus electron transfer chain from harmonic characterization through MarcusRate. Each is reached under the same axiom audit. Each extends derivation depth without enlarging the primitive base. The library has more than 130,000 lines of proof and continues to add domain reach without adding axioms.
The corollary matters. Unfinished branches should be advanced by the same invariant logic as finished branches. The GR_Necessity_Gap6_Closures file closes specific admits that the earlier GR_Necessity_Theorem and GR_Necessity_3plus1D files had left open. UCF_GUTT_Completed_QR_GR_Proofs.v strengthens technical closure around extension, bounded evolution, singularity impossibility, and Laplacian uniqueness. The library is moving from scaffold to tightened scaffold under the same discipline that closed the lower branches. This is what a live substrate-level program looks like when it is working: not an empire of completed claims but an ongoing derivational build whose invariant logic holds as the build extends.
5.3 Eligibility-versus-Completion Honesty
The third criterion is reflexive. Substrate-level programs by their nature make large in-principle claims — "mathematics is downstream of the substrate," "the formal apparatus of physics is reachable without importation" — that outrun their completed derivations. A well-disciplined program makes the in-principle claims confidently and the completion claims file-by-file, and it does not let the two collapse into each other in public presentation.
The Layer Distinction essay does this explicitly. The eligibility claim (reachability in principle) is stated directly and defended. The completion claim (Marcus chain, Planck emergence, gauge-group uniqueness, GR necessity closures) is stated specifically with references. The caveat section names the distinction in terms a reader cannot miss. The presentation does not let a casual reader walk away thinking the framework has completed what it has only established in principle, and it does not let a casual reader walk away thinking the framework has only in-principle claims when specific derivations are already machine-verified.
This is not false modesty. It is the form rigor takes for a substrate-level program, because the program's claims are large enough that a lack of this discipline would make the whole presentation indefensible. The caveat section is not a concession to skeptics; it is the feature that makes skeptical engagement productive rather than at cross-purposes.
Three criteria. Three satisfactions. The framework satisfies them not as a matter of interpretive charity but as a matter of what its files contain and its public essay states. A reader who wants to check this can open the repository and the site. The check is not asked for on faith.
6. Three Objections And Why They Fail
The field will not update quietly. Three objections will be raised, and each is worth naming and rebutting directly.
6.1 "This Is Just Ontic Structural Realism With A Coq File"
The objection is that UCF/GUTT is a technical dressing of a philosophical position already developed in the literature — ontic structural realism, Whiteheadian process ontology, Cassirer's functional primitive — and that the novelty is cosmetic rather than substantive.
The objection fails because it misidentifies what the operational turn contributes. Ontic structural realism is a philosophical position about how to interpret physical theory. It has not, in its extant formulations, derived the number tower from its structural primitives, because its proponents have not had, or have not pursued, the tools. A philosophical position that relations are prior is compatible with, and underdetermined by, many technical implementations — some of which would smuggle mathematical content at the primitive level and some of which would not. The mechanical derivation is what disambiguates.
Saying "UCF/GUTT is just ontic structural realism with a Coq file" is like saying general relativity is just Mach's principle with tensors. The "with tensors" is doing all the work. The difference between a philosophical position and a mechanically derivable derivation is the difference between a claim about what is possible in principle and a demonstration of what is actually achievable in practice. Philosophers of science have spent a hundred years arguing for the first. UCF/GUTT provides the second. The two are not the same activity, and treating them as the same is a category error.
6.2 "Relations Are Mathematical Objects; Your Primitives Are Crypto-Mathematical"
The objection is that a Coq-declared relation is formally U -> U -> Prop, which is a function-valued predicate in type theory, which is mathematical structure. The claim of pre-mathematical primitives is therefore false; the framework merely rebrands mathematical primitives as relational.
The objection has a sophisticated form and a lazy form. The lazy form treats the type-theoretic encoding as decisive: because the formal declaration uses mathematical notation, the primitives must be mathematical. This is the objection of someone who has not distinguished encoding from intention. Natural language arguments are expressed in linguistic structure; nobody treats them as claims about grammar. Formal derivations are expressed in type-theoretic structure; the encoding does not reduce the substantive claim to a claim about type theory. The medium is not the premise.
The sophisticated form of the objection presses harder. It accepts that the encoding is a medium and asks whether the intended interpretation of the primitive — relational structure grounded in a pre-mathematical ontology — is coherent. This is a legitimate philosophical question, and the framework invites it. The response is that the intended interpretation is specified at the grounding section of the library: RelationalGrounding.NaturalsInUniverse takes U : Type and R : U -> U -> Prop as arbitrary, quantifying over any relational universe. The primitives are not instantiated to any specific mathematical domain at the grounding level. The claim is that whatever carrier and whatever binary relation you instantiate, the serial-pointed structure emerges, and from that structure the minimal inductive shape consistent with seriality is what we call ℕ. The pre-mathematical commitment is that "any carrier, any binary relation" is an ontological primitive that does not presuppose mathematical content — it is as permissive as ontology gets without becoming vacuous.
A skeptic can push one level further and claim that universal quantification over types is already mathematical. This is the recursive objection noted earlier. It proves too much: it disqualifies all formal foundational work, not just substrate-level work, and therefore does not distinguish the two. A skeptic who accepts the objection must accept that neither content-level nor substrate-level foundations can achieve what they claim, and at that point the skeptic is not arguing against UCF/GUTT but against the possibility of rigorous ontology in general. That argument can be made, but it is a different argument, and making it is a concession that the substrate/content distinction is not the issue.
6.3 "Until Every Constant Is Derived, You Haven't Really Done Anything"
The objection is that UCF/GUTT has not derived the fine-structure constant, the fermion masses, the CKM matrix, or the full content of quantum field theory from the substrate, and that until it does, its claims are promissory rather than substantive.
This is the moving-goalposts objection, and it is not a serious objection. Derivation programs do not complete in one paper or even one decade. They complete over research generations, and the question of whether a program is working is not "is it complete?" but "is it advancing by closure rather than by proliferation?" UCF/GUTT is advancing by closure. The GR necessity closures discharged admits in earlier files; the UCF_GUTT_Completed_QR_GR_Proofs work strengthens structural results around the unified system; the Marcus chain went from harmonic activation through Complete to NRT to MarcusRate, each file building on the last. The pattern is the pattern of a working program.
The sharper form of the objection notes that the framework uses Parameters (not Axioms) for certain abstract physical constants in Layer 11 files, and that this is where the 19 free parameters of the Standard Model still live. The response is: this is honest scope discipline operationalized in the type system. A Parameter is a declared symbol with no defining equation; it is the framework's explicit acknowledgement that the constant has not yet been reduced to a structural consequence of more basic primitives. The fact that these are Parameters rather than Axioms means they are not committed premises of the library; they are marked as open derivation targets. The framework's discipline is to convert Parameters into theorems as the structural reductions become available — as has already happened for the Lorentzian signature, the gauge group, the Planck constant, and the CPT theorem. The remaining Parameters are the research agenda, publicly marked as such. This is the opposite of overclaiming; it is the most honest form of incompleteness available.
A program that promises to derive everything and has not derived anything is a fraud. A program that promises to derive everything and has in fact derived a substantial and growing portion under machine verification, while marking the rest as open research, is a working program. The first should be dismissed; the second should be engaged with. Confusing the two is a failure of calibration the field should not make.
7. The Slow Update
The publication of the Layer Distinction essay is not the end of a debate. It is the beginning of one the field has been avoiding for a century. The substrate/content distinction has been available in philosophical form since at least Whitehead; it has been unavailable in operational form until the combination of proof assistants, zero-axiom discipline, and a research program with the ambition and discipline to execute both. The combination now exists. The essay's publication makes the combination publicly citable. The next move belongs to the field.
The most likely near-term response is inertia. Foundations will continue to hold workshops on ZFC vs. HoTT, on set-theoretic multiverses, on the axiom of choice, on category-theoretic vs. type-theoretic foundations, and the Layer Distinction will be absent from the discussion. This is fine. The essay is on the open web. It will find readers. The discipline that lets a reader press on the primitives and the audit is the discipline of a technical subfield; it does not require the permission of the philosophical establishment to proceed.
The medium-term response will be engagement, and it will take one of two forms. The first form is serious engagement, in which philosophers of mathematics read the essay, check the audit, press on the primitives, and either accept the substrate-level posture as a genuine new category or articulate a specific and defensible reason for rejecting it. This is the response the framework invites and will benefit from, regardless of outcome. If the primitives survive, the framework is vindicated. If they don't, the framework is refined. Either result advances the inquiry.
The second form is lazy engagement, in which the essay is dismissed with one of the three objections rebutted above, without serious examination of the primitives or the audit. This response will not change what the framework has done; it will only change the reputational standing of the philosophers who resort to it. The longer the framework's derivations stand and the longer its audit continues to pass, the more costly the lazy dismissal becomes. Inertia has a shelf life.
The long-term response, if the framework continues to advance by closure rather than by proliferation, is the slow update. The slow update is not dramatic. It consists of philosophers of mathematics beginning to distinguish substrate-level from content-level work in their vocabulary, beginning to cite the Layer Distinction when framing the question of mathematical foundations, beginning to treat Wigner's question as a question that admits of structural rather than philosophical answers. The slow update will happen because the alternative — pretending that a working substrate-level program does not exist while a working substrate-level program does exist — is not defensible over the long run. It will happen on the timescale that academic updates happen, which is not the timescale anyone finds satisfying.
What can be said now, with confidence the published materials support:
The substrate-level posture is no longer hypothetical. The operational turn has been executed. The Layer Distinction is live. The field will have to update. The question is not whether, but on what timeline.
The arguing is over. The inspecting begins.
References
Awodey, S. (2004). An answer to Hellman's question: Does category theory provide a framework for mathematical structuralism? Philosophia Mathematica, 12(1), 54–64.
Cassirer, E. (1910). Substanzbegriff und Funktionsbegriff: Untersuchungen über die Grundfragen der Erkenntniskritik. Berlin: Bruno Cassirer.
French, S. (2014). The Structure of the World: Metaphysics and Representation. Oxford: Oxford University Press.
Ladyman, J., & Ross, D. (2007). Every Thing Must Go: Metaphysics Naturalized. Oxford: Oxford University Press.
Maddy, P. (2017). What Do We Want a Foundation to Do? Comparing Set-Theoretic, Category-Theoretic, and Univalent Approaches. In S. Centrone, D. Kant, & D. Sarikaya (Eds.), Reflections on the Foundations of Mathematics (pp. 293–311). Cham: Springer.
Merleau-Ponty, M. (1945). Phénoménologie de la perception. Paris: Gallimard.
Peirce, C. S. (1931–1958). Collected Papers of Charles Sanders Peirce (Vols. 1–8). C. Hartshorne, P. Weiss, & A. W. Burks (Eds.). Cambridge, MA: Harvard University Press.
Shapiro, S. (1997). Philosophy of Mathematics: Structure and Ontology. Oxford: Oxford University Press.
Univalent Foundations Program. (2013). Homotopy Type Theory: Univalent Foundations of Mathematics. Institute for Advanced Study.
Whitehead, A. N. (1929). Process and Reality: An Essay in Cosmology. New York: Macmillan.
Wigner, E. P. (1960). The unreasonable effectiveness of mathematics in the natural sciences. Communications on Pure and Applied Mathematics, 13(1), 1–14.