The Totality Paradox: Self-Reference, Decoder Selection, and Coherence Defects in Quantum Physics

The Totality Paradox: Self-Reference, Decoder Selection, and Coherence Defects in Quantum Physics Andrei T. Patrascu FAST Foundation, Destin FL, 32541, USA email: andrei.patr[email protected] We argue that a wide class of conceptual puzzles in quantum gravity and quantum information, including the black hole information paradox, arise from an implicit demand for total internal closure: the assumption that there exists a single, context-independent description from which all information about a self-referential quantum system can be decoded. We show that this demand is structurally inconsistent in sufficiently expressive systems, by connecting physical decoding problems to diagonal no-go results of Cantor, Gödel, Turing, and their categorical unification via Lawvere fixed-point arguments. We formulate this obstruction as a principle of anti-totality: while information about a system may exist and be preserved in correlations, no terminal, globally valid internal decoder can exist once the system contains the means to represent and act upon its own descriptions. Nevertheless, we show that this does not preclude universality. Canonical encodings may still exist as initial objects or universal factorization cores, yielding universality without terminality. We apply this logic to holography by recasting bulk–boundary relations as encoding functors and proving a no-total-holography statement: any boundary encoding expressive enough to internalize its own reconstruction cannot serve as a total classifier of bulk predicates. Within this framework, modern holographic mechanisms such as quantum error-correcting code subspaces and entanglement wedge reconstruction emerge as structurally necessary instances of relative universality. We further reinterpret the island and quantum extremal surface prescription as a variational decoder-selection law, which replaces a forbidden fixed global decoder by regime-dependent decoding contexts selected by minimization of generalized entropy. Finally, we propose Higher Categorical Coherence Breakdown (HCCB) as the physical mechanism realizing anti-totality. In this picture, unitary and linear dynamics persist locally within coherent contexts, but fail to glue into a single global semantics due to nontrivial higher-coherence obstructions. Operationally, this yields sectorization, history dependence, and completely positive effective dynamics on accessible algebras. We argue that relaxing demands for totality not only avoids paradox, but generically increases predictive power by preserving structured memory and contextual coherence. Our results suggest that fundamental physical theories may be universal without being terminal, and that coherence closure, rather than global self-closure, is the minimal consistency requirement for self-referential quantum systems. One-Line Thesis Self-referential quantum systems admit universal (canonical) encodings but forbid terminal (total) decoding; diagonal logic explains the impossibility, islands implement a variational decoder-selection law in gravity, and Higher Categorical Coherence Breakdown supplies the physical mechanism realizing non-totality as coherence defects, sectorization, and completely positive effective dynamics. Keywords: diagonal arguments; Lawvere fixed-point theorem; categorical logic; universality without terminality; holography; islands; quantum extremal surfaces; entanglement wedge reconstruction; Higher Categorical Coherence Breakdown; completely positive maps; Type-III von Neumann algebras; edge modes; self-reference. I. INTRODUCTION A. Motivation: why the “information paradox” is misnamed The phrase “black hole information paradox” suggests that black holes either destroy information or violate quantum mechanics. In the way the term is commonly used, “information” implicitly means something like: (i) global purity of the complete quantum state, (ii) the possibility of reconstructing the initial microstate from late-time data, and (iii) the existence of a single, context-independent decoding 2 map that extracts all relevant facts about the system from the Hawking radiation. The present work argues that it is precisely this conflation that makes the “paradox” appear unavoidable. The relevant tension is not that information fails to exist, but that a certain totality demand is ill-posed. Hawking’s result is about semiclassical observables and coarse-grained entropy. Hawking’s semiclassical computation shows that black holes radiate and that, to leading order in a fixed semiclassical background, the outgoing state is approximately thermal in its local properties and in low-point correlators [ 1 ]. This naturally leads to an entropy increase of the radiation subsystem when one traces out degrees of freedom that remain inaccessible to an exterior observer. At this stage there is no logical contradiction: an exterior observer is describing an open effective system (the accessible exterior algebra), and open systems generically exhibit entropy production even when the global dynamics is unitary. The Page curve reframes the question as one of fine-grained consistency. If the complete evaporation process is globally unitary, then the full state of (black hole + radiation) remains pure, and the fine-grained entropy of the radiation must eventually decrease after reaching a maximum around the Page time [ 2 ]. Thus, the sharp question is not “does entropy increase?” (it can, for restricted descriptions), but rather: Can a semiclassical description reproduce a Page-curve-consistent fine-grained entropy while maintaining internal consistency of the observer-dependent descriptions? AMPS diagnoses an inconsistency that arises from demanding too much at once. The firewall argument [ 3 ] sharpened the tension by showing that if one tries to maintain simultaneously: (i) unitarity of evaporation, (ii) semiclassical effective field theory near the horizon, and (iii) a single globally valid factorization/decoding picture that allows the same interior mode to be purified in incompatible ways, then one runs into an apparent contradiction with entanglement monogamy. From the perspective of this paper, AMPS is best read not as evidence that “information is lost,” but as evidence that a certain global totality assumption about decoding and factorization is too strong. Islands and entanglement wedge reconstruction demonstrate contextual decoding, not information destruction. Modern developments show that the semiclassical computation of fine-grained entropy must be supplemented by new saddles (replica wormholes) leading to the island prescription, which reproduces Page-curve behavior in controlled settings [ 5 ]. In parallel, holographic entanglement and reconstruction frameworks (RT/HRT and their quantum refinements) provide a principled understanding of which bulk degrees of freedom are reconstructible from which boundary subsystem [ 4 ]. The essential conceptual point is that reconstruction is relative: it depends on the chosen subsystem, the code subspace, and the semiclassical regime. In this view, islands do not “move information” or “destroy information”; they implement a context-dependent selection of what is reconstructible from the radiation. Replace “information paradox” by “totality paradox.” The phrase “information paradox” is therefore misnamed in two logically important ways. (1) It suggests that information itself is in jeopardy, whereas what is actually strained is the existence of a single globally valid internal decoding that recovers all semantic facts about the system from one fixed notion of “the radiation.” (2) It suggests that the resolution must be a statement about breaking unitarity, whereas the modern picture points to a different possibility: unitarity may hold globally, yet no single fixed decoding map is valid across all regimes and observer contexts. Accordingly, we propose to rename the core tension the totality paradox: The paradox is not that correlations fail to exist, but that one implicitly demands a terminal, context-independent decoding/description of a self-referential system. Clarification: we are not denying correlations or unitary patches.Nothing in this reframing denies that: • correlations exist and can be tracked operationally (e.g. via entropy and relative-entropy control), nor that Hawking radiation contains nontrivial correlations beyond leading thermality [1]; • quantum evolution can be unitary on sufficiently well-defined local or contextual descriptions (code subspaces, effective algebras, semiclassical patches), as assumed in the Page-curve and island analyses [2, 5]. The point is instead that demanding a single global, context-independent, terminal description/decoder is an over-demand in self-referential regimes. The remainder of the paper makes this precise by (i) isolating the diagonal/anti-totality obstruction, and (ii) showing how holography, islands, and Higher Categorical Coherence Breakdown provide a consistent replacement: universality without terminality. 3 B. Main claim in one paragraph The central claim of this paper is a precise separation between correlations and information about the system, together with a corresponding separation between preservation and total decodability. By correlations we mean operationally accessible statistical structure in quantum states—quantified, for example, by entropic and distinguishability measures such as von Neumann entropy, mutual information, and relative entropy—which are well-defined for states on observable algebras and are constrained by dataprocessing inequalities under physical channels [ 6 , 7 ]. By information about the system we mean semantic predicates (truth-apt questions) concerning the system’s states, observables, or histories—for instance, which microstate a black hole is in, which interior sector is realized, or which decoding map is consistent in a given regime. Our claim is that in sufficiently expressive self-referential quantum systems—systems that can internally represent descriptions of their own behavior and act upon those descriptions—it is generically impossible to internalize all such semantic predicates into a single, globally coherent, uniformly valid decoding scheme. Formally, the obstruction is diagonal: the same schema that yields Cantor’s theorem, Gödel incompleteness, and Turing undecidability can be stated categorically (Lawvere fixed-point mechanism) and implies that no total internal decoder exists that decides all predicates about the system from within the system [ 8 – 10 ]. Thus, the correct replacement for the misnamed “information paradox” is the totality paradox: information about the system may exist and be preserved in correlations (and may be recoverable in appropriate contexts), yet no single terminal, context-independent internal decoder can exist once self-reference is present. This distinction is the logical backbone of our interpretation of holography and islands: modern bulk reconstruction mechanisms already implement relative universality (code subspaces, regionand regime-dependent recovery), precisely because totality is forbidden. Finally, we show how Higher Categorical Coherence Breakdown (HCCB) provides a physical mechanism implementing this semantic non-totality as coherence defects, sectorization, and completely positive operational dynamics [11]. C. What is new here The novelty of this paper is not a new semiclassical saddle, a new bulk reconstruction algorithm, or a new quantum-gravitational model. Rather, the novelty is a unified logical spine—a precise reclassification of what is (and is not) being demanded in the standard “information paradox” narrative—together with a systematic explanation of why the modern holography/islands picture has exactly the form it does, and how this form extends beyond gravity via Higher Categorical Coherence Breakdown (HCCB). (i) A unified logical spine: diagonal logic as a constraint on physics questions. A central contribution is to elevate diagonal logic (Cantor/Gödel/Turing, unified in categorical form by Lawvere) from a background philosophical analogy to an explicit constraint on what a self-referential physical theory can demand of itself [ 8 – 10 , 12 ]. While diagonal arguments are standard in logic and computation, their role in black-hole physics is typically discussed only indirectly (e.g. via informal “no global interior operator” or “state dependence” debates). Here we identify a concrete structural target: the (usually tacit) demand for atotal internal decoder that decides all semantic predicates about the system from within the system (Section I B). The diagonal spine explains why this demand is not merely difficult but generically ill-posed: once the system contains internal encodings of its own descriptions and allows self-application, total internal classification produces a diagonal predicate that escapes classification. (ii) From the diagonal spine to holography and islands: universality without totality. A second contribution is to use this spine to clarify the conceptual status of modern holography. The QEC/EWR viewpoint already implies that bulk reconstruction is relative: it is defined on code subspaces and depends on the boundary region and semiclassical regime [ 13 , 14 ]. Islands and QES add a further refinement: the reconstructible region itself is selected by a variational principle and can change discontinuously across regimes, reproducing Page-curve behavior in semiclassical gravity [ 5 , 15 ]. Our new conceptual step is to show that these features are not ad hoc: they are precisely what one should expect once totality is forbidden. In our language, holography supports universality without terminality: universal (canonical) encodings and factorization cores may exist, but no single globally valid decoder exists across all contexts. This reclassification explains why “the interior becomes part of the radiation” in island physics: not by semantic fiat, but because a fixed global partition would amount to a total decoder after the Page time. (iii) HCCB as the physical mechanism realizing non-totality. A third contribution is to integrate the above semantic constraint with a physical mechanism that is not specific to gravity. Higher Categorical Coherence Breakdown (HCCB) proposes that the minimal consistency requirement is not a single global 4 unitary/linear semantics, but coherence closure: local/sectorial descriptions glue consistently, while higher coherence obstructions may prevent global strictification [ 11 ]. Operationally, such coherence obstructions manifest as sectorization, history dependence, and completely positive (CP) effective dynamics on accessible algebras, rather than as a single terminal decoder. In the present narrative, HCCB provides the bridge from diagonal non-totality (a semantic no-go) to concrete physical phenomenology: it explains how the world can preserve local predictability while refusing global totality. (iv) Reframing rather than competing with standard results. It is important to emphasize that our claims do not compete with established results in holography, islands, or quantum information; rather, they reorganize them. In particular: • We do not deny Hawking’s semiclassical result or the operational existence of thermal-like radiation [1]. • We do not deny Page-curve reasoning as a conditional statement under unitarity, nor the controlled semiclassical derivations of the island formula [2, 5]. • We do not propose a replacement for entanglement wedge reconstruction; we explain why its code-subspace/region dependence is structurally necessary once totality is forbidden [14]. • We do not assert fundamental nonunitarity as a dogma; rather, HCCB distinguishes global unitarity as a total semantics from patchwise unitarity within coherent contexts, allowing CP effective dynamics as an operational remnant [11]. Thus the “newness” is a principled separation of levels: the diagonal spine constrains what can be demanded globally; holography/islands show how gravity realizes contextual decoding; and HCCB generalizes the same coherence logic beyond gravity. D. Roadmap and reading guide This paper is intentionally multi-layered. The core message is conceptual and logical, but we also aim for mathematical precision and for explicit contact with modern holography and quantum-information technology. To make the manuscript readable to audiences with different backgrounds, we provide a structured roadmap and a “choose-your-own-depth” guide. The main text is written to be self-contained for a theoretical physicist familiar with the standard black-hole information problem and with modern holography, while the appendices collect technical material on diagonal arguments and categorical formulations that can be read independently. Core narrative of the main text (physics-first, logic-explicit). The main text follows a single spine: 1. Reclassification of the paradox. Subsections I A and I B separate correlations (operational information) from information about the system (semantic predicates) and identify the hidden over-demand: the assumption of a total internal decoder. This motivates the renaming of the “information paradox” as a totality paradox. 2. Diagonal constraint. Section III A develops the diagonal no-go mechanism (Cantor/Gödel/Turing/Lawvere) and extracts a precise semantic statement: in sufficiently expressive self-referential systems, no terminal context-independent internal decoder exists. The reader who wants only the conceptual conclusion can read the statements and proofs sketches; the reader who wants full formal details can consult Appendix A. 3. Categorical holography: universality without terminality. Section V translates the diagonal constraint into the language of encoding functors and predicate classification. The key message is that holography can be universal (canonical encodings exist in context) without being terminal (no total classifier of all bulk predicates). This section is designed to be readable without deep category theory; the categorical background is optional and collected in Appendix B. 4. Modern holography already avoids totality. Section VI reviews the quantum error correction and entanglement wedge reconstruction viewpoint and shows how it naturally implements relative universality: code-subspace and region dependence are structural features, not flaws. 5. Islands as decoder-selection law. Section VII interprets the island/QES prescription as a variational decoder-selection law and explains why it is conceptually inevitable once totality is forbidden. This provides the physical realization of contextual decoding in evaporating black holes. 5 6. HCCB as physical mechanism. Section VIII integrates Higher Categorical Coherence Breakdown (HCCB) as a mechanism that realizes semantic non-totality dynamically via coherence defects, sectorization, and CP effective dynamics on accessible algebras. This section is written so that the conceptual mechanism is clear even if the reader does not adopt the full higher-categorical formalism. 7. Synthesis and generalization. Sections VII and X synthesize the story: black holes are maximally self-referential and therefore force the non-totality boundary to become geometric, while the same structural pattern appears beyond gravity in AQFT, gauge theories, measurement chains, and open quantum systems. Reading guide (three entry points). Depending on the reader’s goals, the paper can be approached in three ways: •Physics route: read Subsections I A–I C, then Sections VI–VII, and return to Sections III A and V as needed. This route emphasizes islands and reconstruction. •Logic route: read Subsection I B and Section III A first, then Section V, and only afterwards the holography sections. This route emphasizes why total decoding is forbidden. •Mechanism route: read Sections III A, V, and VIII, then Section VII. This route emphasizes HCCB as the general physical realization and treats islands as a concrete gravity instantiation. Main text vs appendices: where formalism and undecidability live. The main text keeps technical overhead minimal and focuses on: (i) unambiguous definitions, (ii) logically correct constraint statements, and (iii) explicit physical interpretation in holography and open-system dynamics. Two kinds of material are placed in appendices: (A) Formal diagonal toolkit. Appendix A collects a compact but fully explicit presentation of the diagonal constructions and the Lawvere fixed-point theorem. This appendix is included to make the logical spine transparent to physicists and to prevent misreadings of “diagonal” as metaphor. (B) Categorical formalism and (non-)strictification. Appendix B spells out the categorical structures used in the main text (e.g. fibers/comma categories, initial objects, natural equivalences, and the meaning of “universality without terminality”). It also clarifies in what sense higher coherence obstructions prevent global strictification. (C) (Optional) undecidability of global closure. A final appendix (Appendix D) outlines how, in sufficiently rich families of context systems, the question “does a global strictification exist?” can become algorithmically undecidable, by reductions in the spirit of Turing/Gödel. This material is explicitly marked as a schema rather than a claim about all physically realizable systems; it is included to clarify why “global unitarity as a total semantic property” may be non-certifiable from within the system even if local unitarity holds. How to use this roadmap. A reader interested primarily in holography and islands can read the main text without any appendix. A reader interested in categorical logic can read the appendices independently. The core results of the paper—the reclassification of the paradox as a totality paradox, the diagonal obstruction to total internal decoding, the no-total-holography statement, and the coherence-first physical mechanism via HCCB—do not depend on the undecidability appendix. II. DEFINITIONS: INFORMATION, ENCODING, DECODING, TOTALITY A. Operational information (correlations) Our later claims hinge on a strict distinction between (i) operational information, meaning correlation and distinguishability structure accessible to physical procedures, and (ii) information about the system, meaning semantic predicates about states, observables, or histories. The diagonal/anti-totality obstruction concerns (ii) and the existence of total internal decoding schemes, not the existence of (i). We therefore begin with a precise operational baseline: the information-theoretic quantities that physically quantify correlations and distinguishability. 6 States and operational predictions. Let H be a (separable) Hilbert space. A (normal) quantum state is represented by a density operator ρ≥ 0with Trρ = 1. Operational predictions are expectation values hOiρ = Tr ( ρ O )for bounded observables O∈ B ( H ), or, more generally, for observables in an accessible von Neumann algebra. The key point is that all operational content is encoded in correlation functionals of this form. Entropy. The von Neumann entropy is S(ρ) := −Tr(ρlog ρ).(1) It reduces to the Shannon entropy on classical distributions and quantifies uncertainty in the operational sense [ 6 , 7 ]. In a bipartite setting HAB = HA⊗ HB with state ρAB , the reduced states ρA = TrBρAB and ρB = TrAρAB define the entanglement entropy of a subsystem, which is the canonical operational measure of correlations induced by discarding the complement. Mutual information. The mutual information I(A:B)ρ:= S(ρA) + S(ρB)−S(ρAB)(2) quantifies total correlations (classical and quantum) between A and B . It has an equivalent and conceptually important form in terms of relative entropy: I(A:B)ρ=D(ρAB kρA⊗ρB),(3) where D ( ρkσ )is defined below. Equation (3) is a direct algebraic identity: substitute the definitions, use log ( ρA⊗ρB ) = log ρA⊗1 + 1⊗log ρB , and note that partial traces convert Tr ( ρAB log ρA⊗1 ) into Tr ( ρAlog ρA ), etc. Thus mutual information is precisely a distinguishability between ρAB and the uncorrelated product state. Relative entropy and distinguishability. Quantum relative entropy (Umegaki relative entropy) is defined by D(ρkσ) := Tr ρ(log ρ−log σ),(4) for states ρ, σ with supp ( ρ ) ⊆supp ( σ )(otherwise D ( ρkσ )=+ ∞ ). Operationally, D ( ρkσ )governs asymptotic hypothesis testing and measures how well ρ can be distinguished from σ by any physically allowed measurement strategy. In particular, it is a monotone under coarse-graining, which we now state precisely. Channels and the data processing inequality. A quantum channel is a completely positive tracepreserving (CPTP) map Φ : S(H)−→ S(H0),(5) representing any physically allowed transformation of states (unitary evolution with an environment, measurement-and-forget, restriction to a subalgebra, etc.). The fundamental monotonicity property of operational distinguishability is the data processing inequality: D(ρkσ)≥D(Φ(ρ)kΦ(σ)) for all CPTP Φ.(6) In words: no physical processing can increase the distinguishability of two states. This inequality is the formal backbone of statements such as “coarse-graining loses information” and “entropy can increase under restriction”. It is therefore essential to our narrative: it guarantees that when observers have access only to a restricted algebra (or radiation subsystem), their operational description is generically non-unitary and exhibits entropy production, even if a larger description is unitary. What this subsection does not claim. Equations (1) – (6) formalize operational information: correlations and distinguishability accessible to physical procedures. They do not constitute a claim that all semantic questions about the system are decidable from such data. In particular, the existence and monotonicity of operational information do not imply the existence of any total internal decoder for information about the system. Establishing that impossibility is the role of the diagonal/anti-totality arguments developed later. B. Information about the system Section II A defined operational information in the standard information-theoretic sense: correlations and distinguishability quantified by entropy, mutual information, and relative entropy, and constrained by 7 data processing. In contrast, much of the debate around “the information paradox” implicitly concerns a different notion: information about the system, meaning the space of semantic distinctions one wishes to decide about the physical situation. This subsection makes that notion precise in a way that is both physically meaningful and mathematically checkable. Semantic distinctions as predicates. Fix a physical system described (in some context) by a state ρ on an algebra of observables A (often a von Neumann algebra). A semantic predicate is a yes/no question about the system that can be represented as a proposition in the operational language. In quantum theory, the canonical mathematical representatives of such propositions are projections P∈ A (idempotent self-adjoint operators), a perspective already present in the logical interpretation of quantum mechanics [18]. The predicate “Pholds” has a Born-rule truth value valρ(P) := Tr(ρ P)∈[0,1],(7) which reduces to { 0 , 1 } for dispersion-free (sharp) cases. Thus, even at the most elementary level, “information about the system” is not merely an entropy number; it is the (generally context-dependent) assignment of truth values to a family of predicates. Predicates about observables and about states. Many semantic distinctions are naturally expressed as predicates about observables (events), e.g. P representing “the detector clicks” or “the particle lies in region Ω.” Other distinctions are predicates about states themselves, e.g. “the state belongs to sector α ” or “the state has property P.” Formally, these are higher-level predicates on the state space S(A): Π : S(A)−→ {0,1}or [0,1].(8) In practice, such predicates can encode questions about superselection sectors, boundary conditions, admissible code subspaces, or consistency of reconstruction maps. These are exactly the kinds of “facts” that an alleged total decoder would be expected to decide. Contextuality: not all predicate families admit global truth assignments. A crucial point (and one that will reappear later as “anti-totality”) is that in quantum theory there is no globally consistent assignment of sharp (0 / 1) truth values to all projection predicates that preserves functional relations among commuting sets, except in highly restricted cases. This is the content of the Kochen–Specker theorem [19]. For our purposes, the lesson is structural: The space of semantic distinctions is not merely large; it is constrained by contextuality. Any claim of a single globally valid, context-independent “truth map” is already suspect in quantum theory before gravity enters. Kochen–Specker is not the same as diagonal undecidability, but it is conceptually aligned: both show that certain global totalizations of “truth” are ill-posed once the system is sufficiently rich. Predicates about histories (path dependence). Many physical questions are not instantaneous but history-dependent: they concern sequences of events or path-dependent observables. A natural formalism for such semantic distinctions is the consistent/decoherent histories framework [ 20 , 21 ]. A (coarse-grained) history α may be represented by a time-ordered sequence of projections ( P(1) α1, . . . , P(n) αn ), and the associated class operator can be written (schematically, in the Heisenberg picture) as Cα=P(n) αn(tn)· · · P(1) α1(t1),(9) with the corresponding decoherence functional D(α, β) := Tr Cαρ C† β.(10) A family of histories is said to decohere (and hence admit a consistent probabilistic interpretation) when D(α, β)≈0for α6=β. This illustrates a key point for our broader narrative: •semantic distinctions can live at the level of paths and processes, not just states at an instant; • “what is true” (or even what is probabilistically meaningful) depends on a chosen coarse-graining and on a consistency condition (decoherence). Thus “information about the system” is inherently structured by context and history, which already points away from the possibility of a single terminal, context-independent decoder. 8 Why this distinction matters for later sections. Operational information (correlations) concerns quantities like S ( ρ )and D ( ρkσ )and their monotonicity under channels. Information about the system concerns the predicate structure—instantaneous and historical—that one attempts to decide or reconstruct. Diagonal arguments will later show that in sufficiently expressive self-referential regimes, no total internal decoding scheme can decide all such predicates uniformly. Holography and islands will then be interpreted as mechanisms that preserve robust (universal) encoding in restricted contexts while forbidding terminal closure. C. Encoding and decoding We now formalize encoding and decoding in a way that is simultaneously (i) operationally meaningful in quantum information theory, (ii) compatible with algebraic (QFT/AQFT) formulations, and (iii) flexible enough to capture holographic “bulk–to–boundary” maps and island/QES decoder selection later in the paper. Encoding as an internal representation map. Let Sys denote the full physical system under consideration, and let ASys denote its (accessible or global) observable algebra. An encoding is the choice of an internal representation of (some aspects of) Sys into a code system Code , together with a physically admissible map that produces the corresponding code state. In the Schrödinger picture, the encoding is a completely positive trace-preserving (CPTP) map E:S(HSys)−→ S(HCode),(11) where S(H)denotes the density operators on H. Operationally, Ecan represent: •restriction to a subsystem (partial trace), •restriction to an accessible algebra (coarse-graining), •a measurement-and-record procedure (followed by discarding degrees of freedom), •or a genuine code embedding (as in quantum error correction). Stinespring dilation and physical implementability. A key structural theorem is that every CPTP map admits a Stinespring dilation: there exists an environment Hilbert space HE , a fixed environment state |0ih0|, and a unitary Uon HSys ⊗ HEsuch that E(ρ) = TrEU(ρ⊗ |0ih0|)U†.(12) This theorem is essential conceptually: it makes explicit that “encoding” is not a mysterious act but a perfectly standard physical process—unitary coupling to degrees of freedom that are later ignored [22]. Kraus representation. Equivalently, Eadmits a Kraus representation E(ρ) = X α Kαρ K† α,X α K† αKα=1,(13) which is often the most convenient operational form [ 23 ]. The Kraus operators encode, in a basis-free way, the different “branches” of the physical interaction with the environment. Heisenberg picture and algebraic encoding. In the Heisenberg picture, the adjoint map E∗:B(HCode)−→ B(HSys)(14) is unital and completely positive, and it pushes observables on the code back to observables on the system. In AQFT-like settings, one often works directly with *-homomorphisms between observable algebras or with inclusions of von Neumann algebras; the channel picture is the most general and contains these as special (noise-free) cases. Decoding as recovery map / interpreter. Adecoder (or recovery map) is a physically admissible CPTP map D:S(HCode)−→ S(HSys)(15) intended to reconstruct the relevant information about Sys from the code. In general, D is not an inverse of E on all states; demanding that would be exactly the “totality” over-demand discussed throughout this work. Instead, decoding is defined relative to a domain of interest and/or a chosen algebra of observables. 9 Code subspaces and relative decoding. A standard and physically meaningful notion is decoding on a code subspace. Let Hphys be a physical Hilbert space (e.g. boundary or radiation), and let Hcode ⊆ Hphys be a subspace whose states admit a controlled interpretation (e.g. a semiclassical bulk EFT sector). The encoding may be represented by an isometry V : Hcode → Hphys , and the induced channel is ρ7→ V ρV † . A decoder Dis then required to satisfy a relative recovery condition: D ◦ E = id on Ωcode ⊆ S(HSys),(16) for a specified code domain Ω code (often all states supported on the code subspace). This is the precise sense in which quantum error correction is “decoding”: it is a left-inverse on a restricted domain, not a global inverse [24]. Operator-algebra error correction (the right level for holography). In holography and AQFT, one often aims to reconstruct not the full state but an algebra of bulk observables. Let A⊆B ( HSys )be the algebra of observables whose semantic content we wish to preserve. A decoding scheme for an algebra can be expressed in Heisenberg form by requiring that each A∈ A has a representative e A on the code such that expectation values agree on the code domain. One convenient algebraic statement is that for all A∈ A there exists e A∈ B(HCode)with Tr ρ A= Tr E(ρ)e Afor all ρ∈Ωcode.(17) Equivalently, one can demand A = E∗ ( e A )on the relevant support. This is the correct notion of “encoding the information about the system”: preserving the truth values of a chosen predicate/observable family (Section II B). Canonical recovery and the Petz map. A particularly important point for our later narrative is that in some contexts decoding is not only possible but canonical: it is fixed by a universal property. When relative entropy saturates a data processing inequality for a channel, the Petz recovery map provides a canonical recovery channel that reverses the coarse-graining on the relevant family of states [ 17 ]. This is an example of “universality without terminality” already at the level of quantum information: recovery exists and is canonical relative to a context (a specified channel and reference state), but cannot be promoted to a single global decoder for all possible queries. Why these definitions matter for anti-totality and islands. The rest of this paper will repeatedly use the following principle: Decoding is always relative: it is defined with respect to a chosen code domain and/or a chosen algebra of semantic predicates. Demanding a single context-independent decoder for all predicates is precisely the over-demand forbidden by diagonal logic. In holography, R (a boundary/radiation region) together with a code domain and an algebra of reconstructible bulk observables defines a decoding context. In island physics, the QES/island variational principle selects which decoding context is consistent in a given regime. HCCB then provides the physical mechanism by which global strictification into a single decoder can fail even when patchwise decoding remains coherent. D. Total internal decoding (the forbidden demand) We now formalize the over-demand that drives the “totality paradox” introduced in Section I A and stated succinctly in Section I B. This subsection gives a definition that is (i) mathematically checkable, (ii) physically interpretable, and (iii) directly aligned with diagonal/anti-totality arguments to be developed in Section III A. The key point is that the paradox does not arise from the existence of correlations, but from the implicit requirement that there exist a single internal procedure that decides all semantic predicates about a self-referential system. From operational data to semantic questions. Fix a physical system Sys and an encoding context (Section II C) consisting of: •a system Hilbert space HSys (or algebra ASys), •a code Hilbert space HCode (or accessible algebra), •an encoding channel E:S(HSys)→ S(HCode). 16 Predicates as subsets. Let Xbe a set. A yes/no predicate on Xis equivalently a subset S⊆X: the predicate “ x∈S ” is true exactly on S . Thus the set of all predicates on X is its power set P ( X ). The claim “there exists a total internal encoding of all predicates” would mean: there exists a map f : X→ P ( X )that assigns to each element x∈X the predicate that x represents, and that assignment is surjective, i.e. every predicate arises from some x. This is exactly the hypothesis that Cantor refutes. Theorem III.1 (Cantor) . For any set X , there is no surjection f : X→ P ( X ). Equivalently, |X|< |P(X)|. Proof. Assume, for contradiction, that f:X→ P(X)is surjective. Define the diagonal subset D:= {x∈X:x /∈f(x)}∈P(X).(29) By surjectivity, there exists d∈Xsuch that f(d) = D. Now ask whether d∈D. By definition (29), d∈D⇐⇒ d /∈f(d). But since f(d) = D, this becomes d∈D⇐⇒ d /∈D, a contradiction. Therefore no such surjection exists. Interpretation as “no total predicate encoding.” Cantor’s theorem can be read directly as an anti-totality statement: There is no way to encode all predicates on Xas elements of Xitself. Importantly, this failure does not depend on subtle properties of X ; it holds for every set, even finite ones in the sense that P ( X )always has strictly larger cardinality than X . The failure is therefore structural: the space of predicates is inherently larger than the space of “names” inside the domain. Why this already matters for physics. Cantor’s theorem is sometimes viewed as a purely set-theoretic curiosity. In our context it has a direct methodological role: it shows that total classification is an exceptionally strong demand even before we consider self-referential dynamics. In physical terms, demanding a single internal coding of all semantic distinctions about a system is analogous to demanding a surjection X→ P ( X ): the claim is that the system contains within itself a name/degree of freedom for every possible predicate about itself. Cantor shows that such a demand is generically incompatible with the size of the predicate space. From Cantor to diagonal constraints in self-referential systems. Cantor by itself does not yet involve evaluation or self-reference beyond x being used as an index; nevertheless it exhibits the diagonal template (27) – (28) in its simplest form. In later diagonal no-go results (Gödel/Turing/Lawvere), the role of P ( X ) is played by spaces of semantic predicates about proofs, computations, or decoding procedures, and the diagonal subset D becomes a self-referential “liar” predicate. Thus Cantor provides the correct intuition: Totality fails because the act of trying to list/classify all predicates creates a new predicate that is defined precisely to evade the list. This is why totality must be abandoned even when universality survives (Section II E). D. Gödel: truth 6=provability Gödel’s incompleteness theorem is the paradigmatic example of diagonal logic operating inside a formal system. While Cantor’s theorem (Subsection III C) already forbids total predicate encoding on cardinality grounds, Gödel shows that even when one restricts attention to syntactically well-formed statements and effective proof systems, a total internal notion of truth is impossible. For our purposes, the key lesson is not incompleteness per se, but the precise reason why total internal truth predicates fail once self-reference and evaluation are internalized. 17 Formal systems and internal semantics. Let Tbe a consistent, effectively axiomatized formal theory that is expressive enough to represent elementary arithmetic. Such a theory has: •a formal language Lof sentences, •a notion of provability ProvT(ϕ), meaning that ϕ∈ L has a proof from the axioms of T, •an intended (external) notion of truth in a standard model. The informal hope behind “total internal truth” would be the existence of a predicate True ( x )definable inside Tthat correctly decides truth for all sentences of L. Gödel numbering and representability. The first crucial ingredient is representability. Gödel showed that syntactic objects (formulas, proofs, derivations) can be encoded as natural numbers, and that relations such as “ x is the Gödel number of a proof of y ” are arithmetically definable [ 8 ]. This establishes the representability condition (R) of Subsection III A: sentences of Tcan refer, via their codes, to other sentences of T, including themselves. Evaluation as provability checking. The second ingredient is evaluation. The relation ProvT ( x )is a computably enumerable predicate on codes, meaning that Tcan internally reason about whether a given code represents a provable sentence. This internal evaluation plays the role of Eval in Subsection III A. Closure under reflection. The third ingredient is closure under reflection. Because Tcan express statements about provability of coded formulas, it can form sentences that talk about their own provability. This closure is not an extra assumption; it is forced by expressiveness. Without it, arithmetic could not reason about proofs at all. The diagonal (Gödel) sentence. Gödel’s diagonal lemma shows that for any sufficiently well-behaved predicate P(x)in L, there exists a sentence Gsuch that T`G↔ ¬P(pGq),(30) where pGq denotes the Gödel number of G . Choosing P ( x ) ≡ProvT ( x )yields the standard Gödel sentence: “This sentence is not provable in T.” Why total truth predicates fail. Suppose, for contradiction, that there exists a predicate True ( x ) definable in Tsuch that for every sentence ϕ, T`True(pϕq)⇐⇒ ϕis true.(31) Applying the diagonal lemma to P(x)≡ ¬True(x)produces a sentence Lsatisfying L↔ ¬True(pLq). If L were true, then True ( pLq )would hold, contradicting the equivalence; if L were false, then ¬True ( pLq ) would be true, forcing L to be true. Thus no such total internal truth predicate can exist. This reasoning is made fully precise in Tarski’s undefinability theorem, which shows that truth for arithmetic cannot be defined within arithmetic itself [26]. Truth versus provability. Gödel’s first incompleteness theorem is often summarized by the slogan “there are true but unprovable statements.” For our purposes, a more structurally accurate formulation is: Truth and provability cannot be identified by any single internal predicate without triggering diagonal inconsistency. Provability is an operationally checkable notion internal to the system; truth is a semantic notion that transcends any fixed formalization. The attempt to collapse the two is exactly a terminal closure move, and diagonal logic shows why it fails. Why this matters for physics. The analogy with physics is direct. In later sections, “provability” will be replaced by “decodability from accessible degrees of freedom,” and “truth” by “semantic facts about the system (e.g. interior predicates).” Gödel’s theorem then foreshadows our main claim: There is no total internal decoder that decides all semantic truths about a sufficiently expressive self-referential physical system. Just as arithmetic remains meaningful and predictive without a total truth predicate, physical theories can remain operationally predictive without terminal decoding. Universality survives; terminality does not. 18 E. Turing: behavior 6=decidability Turing’s halting theorem is the computational analogue of Gödel’s incompleteness. Where Gödel shows that no formal theory can internally decide all semantic truths about arithmetic, Turing shows that no algorithm can decide all semantic facts about algorithmic behavior. For our purposes, the halting problem provides the cleanest and most operational picture of “total decoder failure”: There exists no single procedure that, given an arbitrary internal description of a process and an input, always decides the process’s behavior. This is exactly the structure we later transfer to decoding problems in physics, with “program behavior” replaced by “semantic facts about the system.” Programs as internal descriptions (representability). Fix a standard model of computation (Turing machines). Let N denote the set of (finite) descriptions of machines and let X denote the set of inputs. The key representability property is that machines can take descriptions of machines as inputs: N can be embedded in X . This is the computational realization of (R) and (E) from Subsection III A and is exactly what makes diagonalization possible. The halting predicate (semantic information about behavior). Define the semantic predicate H(M, x) := (1,if machine Mhalts on input x, 0,if machine Mdoes not halt on input x.(32) This predicate is a prototypical example of “information about the system”: it is a fact about a process that may be well-defined (the computation either halts or does not) even if no internal procedure can always decide it. Total decider hypothesis. Assume for contradiction that there exists a total halting decider, i.e. a computable function (algorithm) Halt :N × X −→ {0,1}(33) such that Halt ( M, x ) = H ( M, x )for all ( M, x ). This is the computational prototype of the “total internal decoder” demand: uniformity (single decider for all inputs), correctness (always right), internality (it is itself a computable procedure), and closure under reflection (it can be applied to inputs involving itself). Diagonal machine construction. Given Halt , define a new machine D (the diagonal machine) that takes a machine description Mas input and behaves as follows: 1. Compute Halt(M, M). 2. If Halt ( M, M ) = 1 (predicting that M halts on input M ), then D enters an infinite loop (does not halt). 3. If Halt(M, M)=0(predicting that Mdoes not halt on input M), then Dhalts immediately. This definition is effective if Halt is effective, so Dis a valid machine. Contradiction by self-application. Now evaluate D on its own description. Consider D ( D ). There are two cases: •If Halt(D, D)=1, then by definition D(D)does not halt, contradicting the prediction. •If Halt(D, D)=0, then by definition D(D)halts, contradicting the prediction. Thus Halt cannot exist. This is Turing’s original diagonal argument [9]. Theorem III.2 (Turing) . There exists no total computable function Halt ( M, x )deciding halting for all machines M and inputs x . Equivalently, the halting predicate H ( M, x )is not decidable by any single algorithm. Interpretation: behavior 6=decidability. The theorem can be summarized by the slogan: Behavior (what the process does) is not identical to decidability (what a uniform internal procedure can always decide). In our language, the semantic fact “ M halts on x ” exists, but it cannot be totally internalized as the output of a single universal internal decider. 19 Why this matters for physics. The relevance to physics is conceptual but precise. Many physical claims about black holes implicitly demand something halting-decider-like: Given the radiation state (a code) and an intended interior query, there exists a single, uniformly valid internal decoder that always outputs the correct interior answer. Turing’s theorem shows that any time a system can encode and act on descriptions of its own procedures (Subsection III A), the demand for a total decider is unstable under diagonalization. This is why we treat “total internal decoding” as an over-demand rather than as a computationally hard but well-defined problem. Later, when we interpret islands as a variational decoder-selection law, the correct analogy is: physics replaces the impossible global decider by context-dependent decoding rules, just as computation replaces an impossible universal halting decider by restricted decidability within well-defined classes of programs. F. Lawvere fixed-point theorem Cantor, Gödel, and Turing are often taught as separate pillars: set theory, proof theory, and computation. Lawvere’s insight is that their diagonal cores admit a single categorical expression [ 10 ]. This subsection presents the Lawvere fixed-point theorem in a form that makes it a physics tool: it isolates the minimal structural assumptions under which diagonal self-reference produces either (i) forced fixed points or (ii) an inconsistency in the presence of a negation-like operation. In later sections, this categorical skeleton will be used to interpret why total internal decoding is forbidden while universality can survive. Two structural ingredients. Lawvere’s theorem rests on exactly the two diagonal trigger components discussed earlier: •Evaluation: a canonical way to apply a “code” to an input. •Weak surjectivity of coding: the ability to represent every predicate/procedure (in a target class) by some internal name. These are (E) and a strengthened form of (R) from Subsection III A. The closure-under-reflection condition (C) is implicit in the ability to form diagonal maps. Cartesian closed categories and evaluation Let C be a cartesian closed category (CCC). This means that C has finite products and, for any objects A, B , an exponential object BA representing morphisms A→B . The key structure is the evaluation morphism evA,B :BA×A−→ B, (34) which is universal for morphisms out of (−)×A: C(X×A, B)∼ =C(X, BA). Categorically, ev is the abstract form of “running a program” or “evaluating a predicate.” Weak point-surjectivity as categorical representability Fix A, B ∈ C. A morphism φ:A−→ BA(35) is called weakly point-surjective if every morphism g : A→B is realized by evaluation of some “code” in the image of φ at the corresponding input. Formally, for every g : A→B there exists a global element (point) a: 1 →Asuch that g= ev ◦(φ◦a×idA).(36) Intuitively: 20 Every predicate g is represented by some internal name φ ( a ), and evaluating that name on inputs reproduces g. This is the categorical sharpening of “representability”: it is exactly the hypothesis that the system internally names all predicates in the class under consideration. The fixed-point theorem The core statement is: Theorem III.3 (Lawvere fixed-point theorem) . Let C be a CCC, let A, B ∈ C , and assume there exists a weakly point-surjective morphism φ : A→BA . Then every endomorphism h : B→B has a fixed point: there exists a global element b: 1 →Bsuch that h◦b=b. (37) Proof. Fix h:B→B. Define a morphism g:A→Bby diagonalizing evaluation: g:= h◦ev ◦(φ×idA)◦∆,(38) where ∆ : A→A×Ais the diagonal. By weak point-surjectivity, there exists a: 1 →Asuch that g= ev ◦(φ◦a×idA).(39) Define b: 1 →Bby evaluating the code φ(a)on its own point a: b:= ev ◦(φ◦a×a).(40) Now compute: h◦b=h◦ev ◦(φ◦a×a) =h◦ev ◦(φ×idA)◦∆◦a(by definition of ∆) =g◦a(by (38)) =ev ◦(φ◦a×idA)◦a(by (39)) = ev ◦(φ◦a×a) =b(by (40)). Thus h◦b=b. Why this is “diagonal logic” in categorical form. The proof is an abstract version of the liar construction: •φprovides internal names for predicates (representability). •ev provides self-application (evaluation). •∆implements diagonalization. The fixed point b is the categorical analogue of the “self-referential sentence” produced by diagonalization. From fixed points to contradiction with negation Lawvere’s theorem becomes a no-go when B carries an endomorphism with no fixed points. In the boolean case, negation has no fixed point ( ¬ 0=1, ¬ 1=0). More generally, if there exists a morphism n:B→Bsuch that there is no b: 1 →Bwith n◦b=b, then Theorem III.3 implies: Corollary III.4 (No weak surjectivity in the presence of negation) . If B admits an endomorphism n:B→Bwith no fixed point, then there is no weakly point-surjective map φ:A→BA. Proof. If such a φ existed, Theorem III.3 would force a fixed point of n , contradicting the hypothesis. 21 Interpretation: why total internal decoders are forbidden. Corollary III.4 is the categorical version of: If your system can represent all predicates internally and can evaluate them, then selfreference forces fixed points; if your predicate logic includes negation-like operations, total internal classification leads to contradiction. This is exactly the “weak surjectivity + evaluation ⇒ fixed points ⇒ contradiction with negation” slogan. The connection to our physics narrative is direct: •“Weak surjectivity” is the demand that an internal code classifies all semantic predicates. •“Evaluation” is the physical ability to implement decoding as an internal process. •“Negation” is the ability to form reflection queries that flip the decoder’s predicted answer. Hence, once the system internalizes its own decoding (self-reference), demanding a total internal decoder is structurally unstable. This is why we treat totality as a forbidden demand. Why universality can survive. Notice what Lawvere does not forbid. It does not forbid the existence of canonical encodings for restricted predicate classes, nor does it forbid initial objects or other universal constructions in a context. It forbids only the promotion of such constructions to a total internal classifier closed under negation/self-reference. This is the formal categorical content of “universality without terminality” previewed in Subsection II E. G. Key Proposition A (Semantic No-Go) We now distill the preceding subsections into a single statement that will serve as the logical backbone of the remainder of the paper. Cantor (Subsection III C), Gödel (Subsection III D), Turing (Subsection III E), and Lawvere (Subsection III F) are not merely historical analogies; they are instances of a single structural fact: total internal closure fails in sufficiently expressive self-referential systems. In physics language: once a system can internally represent descriptions of itself, act on those descriptions, and ask questions about the outcomes, no single globally valid internal decoder can decide all semantic predicates about the system. This is the precise sense in which the “information paradox” is better understood as a totality paradox. Set-up. Fix a system Sys together with: •a class of states Ω(the domain), •a family of semantic queries Qwith semantics J·K:Q × Ω→ A as in (18), •an encoding Einto an internal code system (Section II C), •and a notion of decoding Das an internal physical process. We further assume that Sys satisfies the diagonal trigger conditions of Subsection III A: representability (R), evaluation (E), and closure under reflection (C). Statement. Proposition III.5 (Semantic No-Go: anti-totality under self-reference) . Let Sys be a sufficiently expressive self-referential system satisfying (R)–(E)–(C). Then there exists no total internal decoder in the sense of Definition II.1 for the triple ( E,Q, Ω). Equivalently, there is no single, uniformly valid, internal decoding procedure that correctly decides all semantic predicates in Qabout Sys on the domain Ω. Interpretation in one line. Information about the system may exist and be preserved in correlations, but it cannot be totally internalized as a single globally coherent and uniformly decodable code once self-reference is present. 22 Proof idea (diagonal reduction). We now give a proof sketch that makes explicit the logical dependencies and connects Definition II.1 to the diagonal template of Subsection III B. The proof is not a new theorem; it is the categorical/semantic packaging of the classical diagonal arguments, adapted to decoding. Proof sketch. Assume, for contradiction, that a total internal decoder exists. Then by Definition II.1: (T1) there is a single decoder Dvalid uniformly on E(Ω), (T2) it preserves the semantics of all queries in Qon Ω, (T3) it is internal and physically implementable within the same system, (T4) Qis closed under reflection on the decoder’s own predicted answers. Using (R) and (E), the system can represent the query “apply the decoder to the encoding of this query and report the answer.” Using (C) and (T4), this reflected query is itself an element of Q and must therefore be decided correctly by D. Now form the diagonal (liar) query q∆∈ Q defined schematically by: q∆(ρ) := ¬“Doutputs ‘true’ on the encoding of q∆applied to ρ”,(41) where ¬ denotes the negation-like operation in the answer space (for sharp predicates, ¬ is boolean negation; for probabilistic predicates it is an appropriate complement operation, or one chooses a sharp subfamily where ¬is defined). Because q∆ is a valid query by closure (T4), correctness (T2) demands that D returns the correct truth value for q∆ on every ρ∈ Ω. But by construction (41) , the truth value of q∆ ( ρ )is defined to disagree with the decoder’s own prediction about q∆ on the same input. This yields the diagonal contradiction in the form “q∆is true” ⇐⇒ “q∆is not true”, the analogue of (28) in Subsection III B. Hence a total internal decoder cannot exist. Scope conditions and why this is a physics statement. The proposition is sometimes misunderstood as saying “no decoding is possible.” That is not the claim. The claim is that total internal decoding is impossible under self-reference. Decoding remains possible relative to a context: •restrict the query class Q(e.g. to a commuting subalgebra or a code algebra), •restrict the state domain Ω(e.g. to a code subspace or a semiclassical regime), •restrict closure under reflection (do not demand meta-queries about the decoder itself). These are exactly the controlled ways in which physics retains predictive power without terminal closure. This is why the modern holography program (QEC/EWR/islands) naturally implements contextual reconstruction rather than a single global decoder. Relation to the classical diagonal theorems. Proposition III.5 is the direct physics analogue of: •Cantor: no total encoding of P(X)by X[12]; •Gödel/Tarski: no total internal truth predicate [8, 26]; •Turing: no total halting decider [9]; • Lawvere: no weakly point-surjective representation of all predicates in the presence of negation [ 10 ]. In each case, representability + evaluation + closure under reflection yields a diagonal predicate that defeats total classification. In later sections we show that “no total holography” and “islands as decoder selection” are the gravitational instantiations of the same no-go structure. 23 IV. CATEGORICAL LOGIC OF INFERENCE: UNIVERSALITY, NATURALITY, CANONICAL CONSTRUCTIONS A. Universality: initial vs. terminal objects The diagonal results of Section III forbid terminal closure for self-referential systems (Proposition III.5), yet much of the progress in modern physics proceeds by identifying universal structures that are, in practice, highly predictive. To reconcile these facts we must separate two notions that are often conflated under the umbrella word “universal”: 1. Universality as minimal generation (initial objects, free constructions), 2. Universality as total collection (terminal objects, global classifiers). Only the second is threatened by diagonal logic; the first survives and is precisely what powers abductive inference in categorical form. A category as a space of descriptions. Let C be a category whose objects represent candidate descriptions (models, encodings, effective theories, or representations) and whose morphisms represent structurepreserving maps between them (coarse-grainings, inclusions, reductions, or reparameterizations). A universal object in C is one characterized by a universal property rather than by a choice of coordinates or ad hoc optimization. Standard references for these notions are [25, 27]. Initial objects: minimal generators Definition IV.1 (Initial object) . An object I∈ C is initial if for every object X∈ C there exists a unique morphism I→X: ∀X∈ C,∃! (I→X).(42) Intuition. Initial means: “the most economical starting point.” The unique map I→X expresses that any other description X can be obtained from I by adding structure, choices, or parameters, but there is essentially only one way to do so once X is fixed. Initial objects therefore encode minimality: they contain no optional embellishments, only what is forced. Abductive reading. In abductive logic, we seek a hypothesis that explains the observed constraints with the least additional structure. Categorically, “least additional structure” becomes initiality: An initial object is the canonical minimal explanation: every other explanation that works factors through it uniquely. This is why initiality narrows the search space: it turns “choose a good explanation” into “find the minimal generator through which all others factor.” Physics examples. •Free constructions: given a set of generators, the free group/free algebra is initial among groups/algebras receiving a map from the generators. The universal property guarantees canonicity (unique up to unique isomorphism), and computations become stable under reparameterization. •Universal encoding in a context: in our holography setting, an encoding theory E0 being initial in the fiber of all encodings reproducing the same observable data means E0 is the “best possible” encoding for that context: any other encoding that works must factor through E0 (this is the categorical form of abductive selection). Terminal objects: total collectors Definition IV.2 (Terminal object) . An object T∈ C is terminal if for every object X∈ C there exists a unique morphism X→T: ∀X∈ C,∃! (X→T).(43) 24 Intuition. Terminal means: “the most comprehensive sink.” Everything maps into T uniquely, so T acts like a final repository or universal receiver. Epistemic reading: terminality as totality. Terminality is the categorical shape of a total classifier: if T is terminal in an appropriate category of semantic representations, then all descriptions admit a unique map into T , meaning that T can be treated as the final, context-independent collector of information. In the present paper, this is exactly what is ruled out by diagonal logic in self-referential regimes: A “total internal decoder” is a terminalizing demand: it asks for a single global object/process into which all semantic predicates about the system collapse. Thus the anti-totality result of Section III is, categorically, an anti-terminal statement for the relevant “semantic” categories. Physics examples and caution. There are contexts where terminal objects are benign (e.g. a final “output type” in a restricted computational model), but in self-referential settings terminality is precisely where diagonal contradictions emerge. For instance: • A hypothetical global decoding object that assigns truth values to all predicates about itself would be terminal-like and is forbidden by the diagonal schema (Subsections III B–III G). • In black hole physics, demanding a single radiation-based decoder valid for all regimes is exactly a terminalizing demand and is the source of paradox; islands/QES replace it with a variational context selection rather than terminal closure. Initial vs. terminal: the precise distinction we need The distinction can now be stated sharply: Initiality captures minimal generative universality (best explanation, factorization core). Terminality captures total collecting universality (final closure, total decoder). Diagonal logic forbids terminality in sufficiently expressive self-referential settings, but does not forbid initiality. This is the categorical reason our program can be both rigorous and non-paradoxical: we pursue universal cores (initial constructions, canonical encodings) while explicitly rejecting the terminal demand of total internal decoding. That stance is what we later call universality without terminality. B. Canonical vs. arbitrary The distinction between canonical and arbitrary constructions is the second pillar of categorical abduction, alongside universality. Where initiality answers why this structure rather than a more complicated one, canonicity answers why this structure rather than one of many equally workable but choice-dependent alternatives. This subsection makes that distinction precise and explains why naturality emerges automatically from canonicity, rather than being an additional assumption. Why “canonical” is not a matter of taste. In informal usage, “canonical” is sometimes taken to mean “standard” or “convenient.” In category theory, it has a precise structural meaning: A construction is canonical if it is uniquely determined by a universal property and does not depend on arbitrary auxiliary choices (coordinates, bases, gauges, regulators, presentations). Equivalently, any two realizations of a canonical construction are related by a unique isomorphism that itself requires no choices. This notion is standard and foundational in category theory [25, 27]. Arbitrary constructions and hidden choices. By contrast, an arbitrary construction is one whose definition depends on extrinsic choices: •a choice of basis in a vector space, •a choice of coordinates on a manifold, •a gauge fixing, •a regulator or renormalization scheme, 25 •a preferred factorization of a Hilbert space. Such constructions may be perfectly valid and useful locally, but their results depend on the choices made. If those choices are changed, the construction itself must be redefined. Arbitrary constructions therefore do not provide stable explanations: they do not transport coherently under change of context. Canonical constructions as invariants. Canonical constructions, by contrast, are invariants of the structure. They are defined solely by how they interact with all other objects via morphisms. The standard slogan is: Canonical objects are characterized by what they do, not by how they are presented. This is why canonical constructions are indispensable in foundational work: they isolate what is forced by structure alone. Example: basis dependence vs. basis-free formulation. A simple example illustrates the point. Let V be a finite-dimensional vector space. Choosing a basis identifies V with Rn , but this identification is arbitrary. Any statement phrased in terms of a specific basis is not invariant. By contrast, constructions such as: •the dual space V∗, •the evaluation pairing V∗×V→R, •the determinant line det V, are canonical: they are defined without choosing a basis and are preserved under all linear isomorphisms. The basis-free formulation carries strictly more conceptual information because it makes explicit which statements are structural and which are artifacts of presentation. Canonicity as the source of naturality. A central categorical insight is that naturality is not an extra axiom imposed on canonical constructions; it is a consequence of canonicity. To see this, recall that a natural transformation between functors is one whose defining diagrams commute for all morphisms. If a construction depended on arbitrary choices, there would be no reason for such diagrams to commute: changing the choice could change the outcome. Conversely, if a construction is canonical—i.e. determined uniquely by a universal property—then there is no freedom for non-commutativity. The diagrams must commute because there is only one possible map compatible with the universal property. Formally: Canonicity forces naturality. This principle is explained in detail in standard category theory texts and is one of the reasons universal properties are so powerful [25, 28]. Abductive significance: why canonicity collapses the search space. From the perspective of abductive inference, the canonical/arbitrary distinction is decisive. If one allows arbitrary constructions, there are infinitely many candidates compatible with a given set of observations. Requiring canonicity eliminates this freedom: •auxiliary choices are disallowed, •only structures forced by the interaction pattern survive, •equivalent descriptions must be uniquely isomorphic. This is why abductive reasoning guided by canonicity does not involve testing a large number of hypotheses: almost all candidates are rejected immediately because they depend on hidden choices. Physics interpretation: canonical encodings vs. gauge artifacts. In physics, the distinction appears constantly: •gauge-dependent quantities vs. gauge-invariant observables, •regulator-dependent quantities vs. renormalization-group invariants, •coordinate-dependent descriptions vs. geometric invariants. In the present work, a canonical encoding is one that does not depend on a preferred observer, basis, or factorization, but is fixed by the requirement that it reproduces the relevant observables in a given context. Such encodings are robust under deformation of the context and therefore suitable as explanatory cores. 32 The probe/boundary category Probe (or Obs) Objects. An object X∈Probe represents a package of operationally accessible data. Examples include: •reduced density matrices or states on a boundary/radiation algebra; •correlation functionals or n-point functions (possibly in a specified regime); •anomaly and topological data (cohomology classes, index-theoretic invariants); •effective actions or generating functionals; •modular data and relative-entropy objects controlling distinguishability. Crucially, objects in Probe are contextual: they depend on what the observer can access (choice of region, algebra, scale, or code domain). Morphisms. Morphisms in Probe represent transformations of accessible data that are consistent with physical processing: •coarse-graining maps (channels, conditional expectations); •inclusions of algebras (expanding accessible region); •restriction maps (subregion selection); •equivalences of observable packages (e.g. quasi-isomorphisms, homotopy equivalences). Thus Probe is the right setting to express naturality of observables under deformation and coarse-graining. Why “Probe” rather than “boundary.” We emphasize the word “probe” because the same formalism applies beyond AdS/CFT: the relevant accessible data may come from detectors, radiation, subalgebras, or measurement records, not necessarily a geometric boundary. This generality will be essential when we interpret islands as decoder-selection and when we generalize beyond gravity (Section X). The role of Yoneda: probes determine meaning relative to a probe class A key conceptual point (already anticipated in Section II B) is that “what it means to know the system” depends on the family of probes one allows. Category theory encodes this dependence through the Yoneda viewpoint: an object is determined (up to isomorphism) by its behavior against all probes in a chosen category of test objects [25, 27]. In physics language: “Information about the system” is always relative to an algebra/probe family. This is not a weakness; it is the correct structural stance in a self-referential world, and it already foreshadows why a terminal global classifier of all bulk predicates is an over-demand. Preview: encoding functors. With Bulk and Probe specified, the next subsection defines the encoding functor U : Bulk →Probe that maps a bulk description to its accessible probe data. The remainder of Section V will then formulate and prove a No-Total-Holography statement: sufficiently expressive encodings cannot be promoted to terminal classifiers of all bulk predicates without running afoul of diagonal logic. B. Encoding functor U:Bulk →Probe Having specified the ambient categories Bulk and Probe in Subsection V A, we now define the central structural datum of categorical holography: the encoding functor U:Bulk −→ Probe.(49) This functor packages the statement “bulk information is encoded in accessible data” into a mathematically checkable object. Crucially, U is not assumed to be invertible, and it is not assumed to classify all predicates about the bulk. Those stronger demands are precisely what diagonal logic will forbid in later subsections. 33 Objects: Umaps bulk descriptions to probe data For each bulk object B∈Bulk , U ( B ) ∈Probe is the object encoding the operationally accessible information extracted from Bin a specified observational context. Examples include: •Subsystem restriction: U ( B )is the reduced state on a boundary region or radiation subsystem R(partial trace / restriction channel). •Algebra restriction: U ( B )is the state restricted to an accessible von Neumann subalgebra (conditional expectation / coarse-graining). •Observable package: U ( B )is a collection of correlation functions, an effective action, or anomaly data associated to B. •Code-domain encoding: U ( B )is an encoded boundary state in a quantum code model, with B representing an effective bulk code space. The formalism is deliberately broad: the only requirement is that the output lives in the probe category and that the assignment is functorial. Morphisms: functoriality is the formal content of “naturalness” For each morphism f:B→B0in Bulk, the functor Uassigns a morphism U(f) : U(B)−→ U(B0)(50) in Probe, such that identities and compositions are respected: U(idB) = idU(B), U(g◦f) = U(g)◦U(f). This is not a technicality: it is the precise mathematical meaning of the physical demand that the encoding be robust under deformations. If f represents an RG flow, a coupling deformation, or a change of accessible region, then U ( f )is the induced transformation of the probe data. Thus the functoriality of Uis exactly the naturality principle of Subsection IV C expressed in holographic terms. Observables as functors: contexts Fand G To relate bulk and probe descriptions precisely, it is useful to separate: •what data one declares to be the relevant “observable content” of the bulk, and •what data one declares to be the relevant “observable content” of the probe/encoding. Categorically, this is expressed by context functors. We introduce two functors: F:Bulk −→ Obs, G :Probe −→ Obs,(51) where Obs is a category of “observable packages” appropriate to the problem (cochain complexes, correlator families, effective actions, modular/entropy data, etc.). The role of F is to select which aspects of B we insist on matching (the explanatory target), while the role of G is to extract the corresponding observable package from the probe object U(B). Relative holography as a natural equivalence. In this language, a holographic statement takes the form of a natural equivalence α:F≃ ==⇒G◦U, (52) meaning that for each B∈Bulk there is an equivalence αB : F ( B ) ≃G ( U ( B )) in Obs compatible with morphisms in Bulk . This is precisely a naturality condition: the “bulk observable content” and the “probe observable content” agree in the chosen context, and this agreement commutes with deformations. 34 Why contexts matter. The point of separating F and G is conceptual hygiene. The equivalence (52) is context-dependent: it need not hold for all conceivable observables, only for those selected by F and G . This avoids the terminality trap: one never claims that U captures all semantic predicates about B , only those in the specified context. Diagonal logic will later show that promoting (52) to a global, context-independent classifier is generically forbidden. Encoding versus decoding The functor U formalizes encoding: it maps a bulk description to accessible probe data. A decoder would be a right inverse or reconstruction procedure attempting to recover bulk predicates from probe data. However, the semantic no-go (Proposition III.5) implies that in self-referential regimes no such decoder can be global and terminal. Instead, one expects: •relative reconstruction on restricted domains (code subspaces, wedges), •regime-dependent decoder selection (islands/QES), •and, in the HCCB picture, coherence defects that prevent global strictification. Thus the correct role of U is to provide a universal encoding backbone, while all decoding is necessarily contextual. Takeaway. The encoding functor U is the minimal mathematical structure needed to express holography as a statement about functorial accessibility of bulk information. The separation into context functors F, G makes explicit that holography is always relative to a chosen observable class, and therefore already points toward “universality without terminality.” The next subsection makes precise what a forbidden total holography claim would amount to in this language. C. What “total holography” would require Having introduced the encoding functor U : Bulk →Probe and the context functors F, G in Subsection V B, we can now state precisely what it would mean to demand total holography. This subsection is intentionally diagnostic: we formulate the strongest plausible “bulk-from-boundary” claim and isolate the exact point at which it becomes a forbidden terminality demand in the sense of Section II D. The reader should keep in mind that we do not advocate total holography; we are preparing to show why it is generically inconsistent in self-referential regimes. From relative holography to total holography. Relative holography, as expressed by a natural equivalence (52) , only asserts agreement of selected observables in a chosen context. Total holography would upgrade this to a statement about all semantic predicates about the bulk: From the boundary/probe data U ( B )one can internally decide every predicate about the bulk object Bin a single, globally valid way. To formulate this in a way that is mathematically checkable, we must specify what we mean by “predicates” and by “decide.” Predicates as subobjects or propositions A canonical categorical way to model predicates is via subobjects. Assume for the moment that Bulk and Probe admit a reasonable notion of subobjects (e.g. are well-powered and have a stable class of monomorphisms). Then the predicates on an object X can be represented by the poset Sub ( X )of subobjects of Xup to isomorphism. This is standard categorical logic [29, 30]. Bulk predicates. For B∈Bulk, let PredBulk(B) := SubBulk(B).(53) Probe predicates. For X∈Probe, let PredProbe(X) := SubProbe(X).(54) The underlying intuition is that a subobject S ,→B represents a yes/no property of B (membership in S) and generalizes projection predicates in quantum logic (Subsection II B). 35 Total holography as a global predicate classifier The strongest natural requirement. A very strong but clean formulation of total holography is the existence of a natural family of isomorphisms ΘB: PredBulk(B)∼ −−→ PredProbe(U(B)),natural in B∈Bulk.(55) Naturality means that for every morphism f : B→B0 in Bulk the induced squares commute after transporting subobjects along fand along U(f). In words, (55) says: Every bulk predicate is represented (and decidable) as a probe predicate on U ( B ), in a choicefree way that commutes with all deformations. This is exactly a terminality-style demand: it asserts that boundary data internally classifies all bulk semantic distinctions. Relation to a “total internal decoder.” Observe that (55) has the same structure as Definition II.1: •it is uniform (one rule, natural in B), •it is correct (an isomorphism of predicate spaces), •it is internal (predicates live inside Probe), •it is closed under reflection if Probe can express predicates about (55) itself. Thus total holography is precisely the bulk-boundary instantiation of total internal decoding. Why the reflection step is unavoidable in self-referential holography The diagonal obstruction requires closure under reflection. In the holographic setting, reflection becomes unavoidable when: 1. the probe data includes observers/decoders as physical subsystems, and 2. the query class includes questions about the decoding itself (e.g. which reconstruction map applies, whether a given operator is reconstructible, whether a given island is selected). In evaporating black hole settings this closure is natural: the radiation is produced by the system and decoding is a physical process acting on the radiation; hence the system can in principle query its own decoding map. Under such conditions, (55) is exactly the kind of terminal closure that diagonal logic forbids. Total holography as a terminality demand It is useful to say this explicitly: Total holography is terminality: it asserts the existence of a single global internal classifier of all bulk predicates from probe data. This is the bulk-boundary analogue of: •a total truth predicate in arithmetic (forbidden by Gödel/Tarski), •a total halting decider (forbidden by Turing), • a weakly point-surjective representation of all predicates in the presence of negation (forbidden by Lawvere). Accordingly, one should expect total holography to fail precisely in the regimes where self-reference becomes unavoidable. 36 What we keep instead. The correct replacement is relative holography: •choose a context (F, G)selecting which observables/predicates are relevant, •allow universal encodings within that context (initial objects in a fiber), •forbid terminal global closure across all contexts. In the next subsection we state and prove the corresponding No-Total-Holography result: once the probe category is expressive enough for reflection, a global predicate classifier (55) is inconsistent with diagonal logic. D. Key Proposition B (No-Total-Holography) We now state the bulk–boundary analogue of the semantic no-go (Proposition III.5): under self-reference, there can be no total holographic classifier of bulk predicates from boundary/probe data. This result formalizes the intuitive claim that holography is universal but not terminal: it may provide canonical encodings and robust reconstruction in restricted contexts, yet it cannot yield a single globally valid decoder that decides all bulk semantic distinctions. The proposition below is deliberately framed to match the diagonal trigger conditions (R)–(E)–(C) of Subsection III A and the definition of “total holography” in Subsection V C. Set-up recap. We have: •a bulk category Bulk and a probe category Probe (Subsection V A), •an encoding functor U:Bulk →Probe (Subsection V B), • a notion of predicates on objects, modeled (for definiteness) by subobjects Pred ( X ) = Sub ( X ) (Subsection V C), • and a “total holography” demand expressed as a natural predicate classifier Θ B : PredBulk ( B ) ∼ = PredProbe(U(B)) (Equation (55)). We emphasize that the categorical form of the argument is independent of the particular choice “predicates = subobjects”; any sufficiently expressive internal predicate logic works similarly, as in Lawvere’s fixedpoint theorem (Subsection III F) [10]. Proposition V.1 (No-Total-Holography).Assume the following. (H1) (Predicate expressivity). The probe category Probe admits a sufficiently expressive internal predicate logic (e.g. a topos/subobject classifier structure, or CCC-style evaluation) so that predicates about U(B)include predicates about the encoding/decoding process itself [10, 29, 30]. (H2) (Self-reference / closure). The admissible query class is closed under reflection: it includes predicates that refer to the outcomes of would-be reconstruction/decoding procedures acting on probe data (the analogue of condition (C) in Subsection III A). (H3) (Total holography demand). There exists a natural family of isomorphisms Θ B : PredBulk ( B ) ∼ −→ PredProbe ( U ( B )), natural in B , i.e. a global internal classifier of all bulk predicates from probe data (Equation (55)). Then (H1)–(H3) are mutually incompatible in a self-referential regime: no such global predicate classifier Θcan exist. Equivalently, if Probe is expressive enough to internalize its own reconstruction, then U cannot be a total classifier of bulk predicates. Consequently, holography can at best be contextually universal: universal (canonical) encodings and reconstruction maps may exist on restricted domains (code subspaces, sectors, specified regions/algebras), but no single terminal, context-independent decoder/classifier exists across all predicates and contexts. Proof sketch (diagonal reduction). Assume (H1)–(H3). By (H3), every bulk predicate P∈PredBulk ( B ) corresponds to a unique probe predicate Θ B ( P )on U ( B ). By (H1), the probe predicate logic is rich enough to form predicates that talk about the behavior of Θ B itself, i.e. predicates that refer to the classifier’s own output. By (H2), such reflected predicates are admissible and must themselves be classified by ΘB. 37 Now apply the diagonal schema of Subsection III B. Construct a probe predicate D on U ( B )that is defined to disagree with the classifier’s own prediction about Don its own name: schematically, D:= ¬“Θ−1 B(D)holds”,(56) where ¬ denotes the negation/complement operation available in the internal predicate logic. Because (H2) allows reflection and (H3) purports to classify all predicates, D must be in the image of Θ B ; let P∆ := Θ −1 B ( D )be the corresponding bulk predicate. But then the truth value of D is defined to be the negation of the truth value of P∆ , which by correctness of the classifier must coincide with the truth value of D itself. This yields the diagonal contradiction D↔ ¬D in the internal logic, exactly as in Cantor’s set, Gödel’s sentence, and Turing’s machine. Therefore Θcannot exist. Interpretation: why this is not anti-holography. Proposition V.1 does not deny holographic encoding. It denies the terminal strengthening of holography into a total internal classifier. In fact, the proposition predicts the structure we see in practice: •Code-subspace restriction. Reconstruction is proven on a code domain (QEC view of AdS/CFT) rather than on the full Hilbert space. •Region dependence. Which bulk algebra is reconstructible depends on the chosen boundary region (entanglement wedge reconstruction). •Sector/context dependence. Different semiclassical regimes, sectors, or observer contexts admit different reconstruction maps; islands are a geometric realization of a decoder transition. All of these are manifestations of contextual universality: universal cores exist within contexts, while terminal closure across all contexts is forbidden. Relation to diagonal logic and to later sections. This proposition is the bulk–boundary counterpart of the semantic no-go (Proposition III.5) and is ultimately a corollary of the Lawvere fixed-point mechanism [ 10 ]. In the next subsection, we will show how the QEC/EWR framework provides exactly the kind of contextual universality allowed here, and how the island/QES prescription can be interpreted as a variational selection of the appropriate decoding context rather than as a fixed global decoder. E. How this reframes “bulk reconstruction puzzles” The holography literature contains several long-standing “puzzles” that arise when one implicitly expects bulk reconstruction to behave like a terminal, globally defined decoding map. Examples include: •state dependence and the status of interior operators, •apparent non-factorization of degrees of freedom across regions, •the existence of operators that are not reconstructible from a given boundary region, •and tensions between different reconstructions that appear mutually inconsistent. Proposition V.1 reframes these puzzles by identifying the hidden assumption: the demand for total holography, i.e. a global internal classifier of all bulk predicates from boundary data. Once that demand is abandoned, the “puzzles” become expected structural features. In this subsection we articulate this reinterpretation precisely and introduce a central concept: diagonal remainders. From “why can’t we reconstruct?” to “what must fail?” The naive question “why can we not reconstruct every bulk operator from the boundary?” presupposes that reconstruction should be terminal. Diagonal logic suggests the correct question is instead: Given anti-totality, which operators/predicates are structurally non-reconstructible from a fixed probe context, and how are the allowed reconstructible algebras selected? In other words, the relevant issue is not the failure of reconstruction per se, but the classification of its inevitable domain-of-validity and its inevitable obstructions. 38 Non-reconstructible operators as diagonal remainders Definition (informal). Fix a probe context (boundary region, accessible algebra, code subspace, semiclassical regime), and let U : Bulk →Probe be the encoding functor. A bulk predicate/operator is reconstructible in this context if its semantic content is preserved by the encoding in the sense of Section II C. A bulk predicate/operator is a diagonal remainder if it cannot be reconstructed within that fixed context without promoting the reconstruction to a total internal classifier, i.e. without violating Proposition V.1. Why diagonal remainders must exist. The core diagonal schema of Subsection III B guarantees that any fixed total classifier induces an evading predicate constructed by self-reference. In the holographic setting, any attempt to build a single, fixed boundary decoding map that decides all bulk predicates in a self-referential regime will generate an obstruction: a bulk predicate whose meaning depends on the decoder’s own output and hence cannot be assigned consistently by that decoder. Therefore: Non-reconstructible operators are not anomalies; they are required remainders of anti-totality. The precise content of the remainder depends on the chosen context (the probe algebra), but its existence is structural. Where diagonal remainders appear physically Diagonal remainders manifest in several familiar guises: (i) Region dependence (entanglement wedge boundaries). Entanglement wedge reconstruction asserts that operators in EW ( R )are reconstructible from a boundary region R on an appropriate code domain. Operators outside EW ( R )are not reconstructible from R . From our viewpoint, this is not merely geometric: it is a controlled instance of contextual universality. The existence of non-reconstructible operators outside the wedge is expected; demanding otherwise would amount to a fixed global decoder. (ii) State dependence and code-subspace dependence. A reconstruction map defined on one semiclassical code subspace need not extend to a larger domain without contradiction. Attempting to force a single reconstruction valid on all states is precisely the terminality demand forbidden by Proposition V.1. Thus state dependence can be read as a boundary manifestation of anti-totality: the theory must restrict the domain to preserve coherence. (iii) Edge modes, centers, and sectorization. In gauge theories and gravity, localization of algebras introduces centers and edge degrees of freedom. This is another form of diagonal remainder: additional sector data is required to glue descriptions consistently. From the categorical perspective, these are precisely the “extra labels” that appear when terminal factorization is forbidden. (iv) Islands as decoder transitions. In evaporating black holes, the island/QES prescription changes the effective reconstructible region associated to radiation. In our language, this is a controlled handling of diagonal remainders: rather than demanding a fixed decoder and accumulating contradiction, the theory selects a new context (a new wedge) in which reconstruction is coherent. How to work with remainders: classify rather than eliminate Once non-reconstructible operators are recognized as structurally necessary, the correct research strategy changes: •Do not attempt to eliminate remainders by ad hoc extensions that effectively reintroduce totality. • Instead, classify remainders by context: determine which predicates are reconstructible in which probe families and what coherence data relates overlapping reconstructions. This is precisely a categorical/descent viewpoint: global descriptions are obtained not by a single terminal object but by gluing compatible local reconstructions, with obstructions encoded as higher coherence data. This viewpoint is also the natural bridge to HCCB, where coherence defects become dynamical and appear as CP flows and history dependence. 39 Takeaway. The correct moral of No-Total-Holography is not pessimism but clarity: Bulk reconstruction puzzles are the expected fingerprints of anti-totality. Non-reconstructible operators are diagonal remainders, not failures of holography. With this reframing, islands and entanglement wedge reconstruction cease to look ad hoc: they are precisely the structured ways physics avoids the forbidden demand of terminal, context-independent decoding. VI. ENTANGLEMENT WEDGE RECONSTRUCTION AS RELATIVE UNIVERSALITY A. Quantum error correction view of bulk reconstruction The purpose of this subsection is to state, in a form that is both logically clean and operationally precise, what the modern “bulk reconstruction = quantum error correction” paradigm actually asserts. This paradigm is the canonical example of universality without terminality: reconstruction exists and is robust within a specified context (code subspace, region, and algebra), while the anti-totality results of Section III forbid any promotion of reconstruction to a single global decoder valid for all predicates, all states, and all observer contexts. The correct statement is therefore a relative universality: universal within a controlled regime, non-terminal globally. Encoding as an isometry and the code subspace. Let Hphys denote the physical Hilbert space of the boundary theory (or, more generally, of the degrees of freedom accessible to a putative decoder). A code subspace is a subspace Hcode ⊆ Hphys,(57) whose states admit a controlled bulk interpretation (e.g. semiclassical bulk effective field theory excitations on a background). Equivalently, one may represent the encoding by an isometry V:Hcode ,→ Hphys,(58) and identify encoded physical states as ρphys =V ρcodeV†. This is the basic QEC posture of holography emphasized in [ 13 ]: bulk effective degrees of freedom are not identified with the entire boundary Hilbert space, but with a distinguished subspace (or subalgebra) thereof. Noise as restriction to an accessible region. To express subregion reconstruction, factorize (at least conceptually) Hphys ∼ =HR⊗ H ¯ R,(59) where R is the boundary region (or radiation subsystem) accessible to the decoder. Restriction to R is then a channel (a “noise map”) NR(ρ) := Tr ¯ R(ρ),(60) and the operational question is: what aspects of the code can be recovered from the noisy output NR(V ρV †)? Recovery maps and relative decoding. A (Schrödinger-picture) recovery map is a CPTP map RR:S(HR)−→ S(Hcode),(61) or, equivalently, a map that reconstructs the relevant code information from region R . The strongest possible recovery condition (perfect state recovery on the code) would be: RRNR(V ρV †)=ρ∀ρ∈ S(Hcode).(62) In holography, one seldom needs (and cannot expect) such a global condition for all states of the full boundary theory. Instead, the meaningful content is that recovery holds on a specified code domain and for a specified class of observables/predicates. This is precisely the “relative” aspect of reconstruction. 40 Operator-algebra quantum error correction (OAQEC). Holography is most naturally formulated at the level of algebras of observables, not entire states. Let Abulk be an algebra of bulk effective operators (e.g. those supported in a candidate entanglement wedge), represented on Hcode . The OAQEC question is: does there exist a representation of Abulk on the accessible boundary region R that reproduces all code matrix elements? In the Heisenberg picture, this becomes the existence of a *-representation πR:Abulk −→ B(HR)(63) such that for every O∈ Abulk, V†πR(O)⊗1¯ RV=Oon Hcode.(64) Equation (64) is the precise operational meaning of “ O is reconstructible from R on the code subspace.” In this form, entanglement wedge reconstruction is simply an OAQEC statement for a particular choice of Abulk and region R[14]. Error correction conditions (why this is nontrivial). In standard QEC, perfect correction is characterized by the Knill–Laflamme conditions. In OAQEC, the corresponding conditions are expressed in terms of commutants and the action of the noise channel on the code; these conditions precisely capture the idea that the noise cannot distinguish different elements of the protected algebra. One convenient general formulation is given in [ 31 ]. The important conceptual point for our purposes is that OAQEC makes explicit that: Reconstruction is always relative to (i) a code domain, (ii) a noise model (here, restriction to R), and (iii) a protected algebra Abulk. This is exactly the categorical anti-totality stance: one never asserts a single global decoder that works for all predicates on the entire state space. Relative entropy and canonical recovery (Petz). When relative entropy monotonicity is saturated for a channel and a family of states, the Petz recovery map provides a canonical reconstruction (“recovery by universality”) relative to that context [ 17 ]. This further illustrates the logic of the paper: recovery can be canonical and robust in a regime without being globally terminal. Why this already obeys anti-totality. The QEC/EWR framework does not even attempt totality in the sense of Definition II.1: •it restricts to a code domain Hcode rather than all states; •it restricts to a boundary region Rand a specific noise model NR; •it reconstructs a specific protected algebra Abulk, not all bulk predicates. Therefore, the “non-reconstructible operators” that appear outside EW ( R )or outside the code domain are not failures; they are the expected diagonal remainders of anti-totality (Subsection V E). Takeaway. Bulk reconstruction in modern holography is best understood as operator-algebra QEC: a universal (canonical) encoding within a code subspace, together with recovery maps that are valid on specified regions and for specified algebras. This is the paradigmatic example of contextual universality that survives diagonal constraints. In the next subsections, we will connect this explicitly to entanglement wedge reconstruction and then to islands as a variational selection of the decoding context. B. Region dependence A defining feature of entanglement wedge reconstruction (EWR) is that it is intrinsically region dependent: which bulk operators can be reconstructed depends on which boundary region (or radiation subsystem) is accessible to the decoder. This subsection explains why region dependence is not a technical limitation or an inconvenient subtlety, but a structural necessity once totality is forbidden. In the language developed earlier, region dependence is the bulk–boundary manifestation of contextual universality. From global boundary to boundary regions. In the earliest formulations of AdS/CFT, it was often heuristically stated that “the bulk is encoded in the boundary theory.” Taken literally, this suggests a global statement: access to the entire boundary Hilbert space should suffice to reconstruct all bulk operators. Modern holography refines this picture by considering subregions of the boundary and asking what bulk information is accessible to observers restricted to those regions. This refinement leads directly to region dependence. 41 Entanglement wedge reconstruction. Given a boundary region R , one defines its entanglement wedge EW ( R )as the bulk domain of dependence bounded by R and the corresponding (quantum) extremal surface [ 4 , 14 ]. The EWR theorem states that, on an appropriate code subspace, any bulk operator supported within EW(R)can be reconstructed from operators acting only on R: Obulk ⊂EW(R) =⇒Obulk is reconstructible from R. (65) Operators supported outside EW(R)are not reconstructible from Ralone. Region dependence as an operator-algebra statement. In the operator-algebra QEC formulation (Subsection VI A), region dependence is expressed as follows. To each boundary region R one associates an algebra AR of bulk operators that are correctable against erasure of ¯ R . Different regions correspond to different protected algebras: R16=R2=⇒ AR16=AR2.(66) There is no single bulk algebra Abulk that is simultaneously reconstructible from all boundary regions. This is not a failure of the formalism; it is the expected outcome of OAQEC [13, 14]. Why region dependence is unavoidable under anti-totality. Suppose, contrary to EWR, that region dependence were absent: that the same bulk algebra Abulk were reconstructible from any sufficiently large boundary region by a single, region-independent decoder. Such a statement would amount to a form of total holography: a single internal decoding rule that classifies the same bulk predicates regardless of which probe context (boundary region) is used. Proposition V.1 shows that this demand is generically inconsistent once the probe category is expressive enough to internalize reconstruction itself. Thus region dependence is not an accident; it is how holography avoids the forbidden terminality demand. Inclusion relations and monotonicity. Region dependence has a precise structural order. If R1⊆R2 , then typically EW(R1)⊆EW (R2),(67) and correspondingly AR1⊆ AR2. This monotonicity reflects the functorial nature of reconstruction: enlarging the accessible region enlarges the protected algebra. Categorically, boundary regions form a poset (or category) under inclusion, and the assignment R7→ AR is a covariant functor into the category of algebras. This functoriality is the concrete expression of naturality under morphisms (Subsection IV C). Multiple reconstructions and apparent redundancy. A striking feature of EWR is that the same bulk operator may be reconstructible from multiple, distinct boundary regions. This “redundancy” is often described as a hallmark of holographic codes. From the present perspective, it is simply the statement that universality holds in multiple overlapping contexts: the same semantic predicate can be represented in more than one probe context, provided one does not demand a single terminal classifier that unifies all contexts simultaneously. Diagonal logic predicts precisely this pattern: overlapping local universality without global closure. Relation to bulk reconstruction puzzles. Several puzzles in the literature—such as apparent inconsistencies between reconstructions from different regions, or the question “which reconstruction is the real one?”—arise from forgetting that region dependence is structural. There is no unique “the” reconstruction map for a bulk operator; there is a family of contextual reconstructions, each valid in its own region and code domain. Attempting to force a unique global answer reintroduces the terminality demand that diagonal logic forbids. Takeaway. Region dependence is not a limitation of entanglement wedge reconstruction; it is its conceptual strength. It is the concrete holographic realization of universality without terminality: Which boundary region reconstructs which bulk algebra is a structural feature dictated by anti-totality, not a defect to be cured. In the next subsection, we will show how this region dependence naturally extends to state dependence and code-subspace dependence, completing the picture of relative universality in modern holography. 48 C. Page curve as a decoder transition The Page curve is traditionally presented as a statement about entropy evolution under global unitarity: the fine-grained entropy of Hawking radiation cannot grow indefinitely; it must eventually turn over and decrease [2]. In the present framework, the Page curve acquires a sharper conceptual meaning: the Page transition is a decoder transition : a variationally forced change of the decoding context required to avoid the forbidden demand of a fixed global decoder. This subsection explains that statement in detail and makes explicit how the island formula realizes it. Two regimes, one fixed probe system. Fix the radiation subsystem (or boundary region) R accessible to an exterior observer. The probe system is fixed: the observer always has access to R . What changes is not the probe system, but the decoding context associated to R : which bulk region is treated as reconstructible from Rin that regime. No-island decoding context: early times At early times, the semiclassical picture suggests that R is entangled with degrees of freedom that remain effectively inside the black hole. If one adopts the no-island context ( I = ∅ ), the generalized entropy becomes Sgen(R;∅) = Sbulk(R),(77) so the radiation entropy is governed by the bulk entanglement entropy of the radiation region alone. Because the radiation is approximately thermal at leading order, Sbulk ( R )grows as radiation accumulates. In this regime, the no-island decoder is consistent: it provides a coherent semiclassical description, and the area cost of introducing an island would outweigh any entanglement benefit. Thus the variational principle selects I=∅. Island decoding context: late times At late times (in particular after the Page time), global purity and the Page argument imply that the radiation must begin to purify itself: the fine-grained entropy of R must stop growing and begin to decrease [ 2 ]. If one insists on keeping the no-island context fixed, one obtains an ever-growing entropy, which becomes inconsistent with unitarity-compatible expectations and leads to AMPS-type tensions when combined with horizon entanglement structure [ 3 ]. In the present language, this is exactly the failure mode of a fixed decoder: it is an attempt to maintain a single decoding context across regimes in a maximally self-referential setting, effectively reintroducing a terminal closure demand. The island/QES prescription provides an alternative decoding context by allowing I6=∅. Then Sgen(R;I) = Area(∂I) 4GN +Sbulk(R∪I),(78) and the inclusion of I can drastically reduce the bulk entropy term because R∪I now captures interior partners responsible for the growth of Sbulk ( R ). The price is the area term. After the Page time, the reduction in Sbulk ( R∪I )can outweigh the area cost, making the island context the variationally preferred decoding context. This is precisely the mechanism by which the island formula produces a Page curve in controlled models. The Page transition as saddle crossing The island formula S(R) = min Iext hSgen(R;I)i(79) implies that the entropy is given by the smallest of the competing extremal saddles. The Page transition is the point where two saddles exchange dominance: Sgen(R;∅)≈Sgen(R;I?).(80) 49 Before this time, the no-island saddle dominates; after it, the island saddle dominates. This is a standard variational phase transition in a saddle-point evaluation, and it is completely analogous to phase transitions in thermodynamics (where different minima of a free-energy functional exchange dominance). Decoder-transition interpretation. The key conceptual point is that each saddle corresponds to a different decoding context: •no-island saddle ↔no-island decoding context (small wedge), •island saddle ↔island decoding context (enlarged wedge including I?). Thus the saddle crossing is literally a decoder transition: the theory switches which bulk algebra is treated as reconstructible from Rin order to remain coherent and avoid terminal fixed-decoder contradictions. Why the transition is forced by anti-totality We can now state precisely why the transition is not an arbitrary semantic move. Fixed-decoder insistence is a terminality demand. Insisting that the decoding context remain fixed (no island for all times) is equivalent to demanding that one and the same reconstruction map from R apply across all regimes. In self-referential settings, such a fixed global decoder is exactly what diagonal logic forbids (Section III): it is a form of terminal closure. In black holes, this insistence manifests as AMPS-type contradictions [3]. The island transition is the minimal consistency repair. The variational rule chooses the minimal change of context that restores coherence: introduce an island only when the entanglement cost forces it, and only in the amount constrained by the QES extremality condition. Therefore the Page curve is not merely an entropy plot: it is the signature that the system has avoided forbidden terminality by transitioning to a new, coherent decoding context. Takeaway. The Page curve can be read as the time-profile of a decoder-selection process: no-island → island saddle crossing is a variationally forced change of decoding context, implementing contextual universality in a maximally self-referential regime. D. Key Proposition C (Islands as a physical regularization of non-totality) We now state, in proposition form, the main conceptual result of Sections VII A–VII C: island formation is not an optional interpretive gloss; it is a concrete physical mechanism that replaces a forbidden terminal decoding demand by a regime-dependent, variationally selected decoding context. This proposition is the gravity-side counterpart of: •the semantic no-go (Proposition III.5), •and the No-Total-Holography statement (Proposition V.1). Proposition VII.1 (Islands as variational decoder selection) . Consider an evaporating black hole system with an accessible radiation subsystem R and a semiclassical regime in which the generalized entropy functional (73) is meaningful. Assume: (I1) (Contextual reconstruction) Bulk reconstruction is meaningful only relative to a context (region, code domain, algebra), as in operator-algebra QEC/EWR (Section VI) [13, 14]. (I2) (Anti-totality) No single fixed, globally valid internal decoder/classifier exists that decides all semantic predicates about the system from a fixed probe subsystem in a self-referential regime (Sections III and V; Propositions III.5 and V.1). (I3) (QES/island semiclassical entropy rule) The fine-grained entropy of R is computed by the min–ext island prescription (74) derived via replica methods in controlled models [5]. Then island formation implements a physical regularization of non-totality: The theory replaces a forbidden fixed global decoder for R by a regime-dependent decoding context selected by a variational principle. Concretely, candidate islands correspond to candidate reconstruction wedges/algebras for R , and the min–ext rule selects the context in which decoding is coherent in that regime. 50 In particular, the Page transition is the saddle crossing between distinct decoding contexts (no-island → island), and should be understood as a decoder transition rather than as an information-loss event. Argument (structural, with explicit logical dependencies). The proof is a logical reduction, not a rederivation of the replica computation. By (I1), reconstruction is only meaningful relative to a context; in particular, there is no a priori reason to expect a fixed context to remain valid across all regimes. By (I2), insisting on a single fixed decoder valid for all times and states is an over-demand: it would amount to a terminal internal classifier, forbidden by diagonal logic in self-referential systems. Now consider the semiclassical entropy computation in (I3). The island functional (73) assigns to each candidate island I a generalized entropy value Sgen ( R ; I ). As explained in Subsection VII B, the choice of I is equivalently the choice of a decoding context for R (which wedge/algebra is treated as reconstructible from R on the code domain). The min–ext prescription (74) selects the island (hence the decoding context) that minimizes Sgen subject to quantum extremality. If one attempts to hold the decoding context fixed (no island) across all times, then after the Page time the no-island saddle yields an ever-growing Sbulk ( R )inconsistent with unitary-compatible Page behavior [ 2 ] and leads to AMPS/monogamy tensions [ 3 ]. Thus the fixed-decoder insistence forces contradiction or incoherence, exactly as predicted by (I2). The island saddle provides the minimal variationally allowed change of context that restores coherence: include I when and only when doing so lowers the generalized entropy. Therefore island formation is precisely the physical mechanism by which the theory avoids a forbidden fixed global decoder, replacing it by contextual decoding selected by a variational principle. Remarks (what this does and does not claim). • The proposition does not claim that interior degrees of freedom “move” into the radiation. Rather, it claims that the consistent decoding context changes: what is treated as reconstructible from R is regime-dependent. • The proposition does not assume the existence of a total decoder; it explicitly replaces that forbidden demand by a variational selection of contexts. • The proposition is compatible with both global unitarity and with the HCCB viewpoint where unitarity is patchwise and global strictification fails; in both cases islands act as coherence repairs rather than as totalizing decoders. Takeaway. Islands/QES are best understood not as “mysterious interior inclusions” but as a variational decoder-selection law that regularizes non-totality in a maximally self-referential gravitational regime. E. Relation to AMPS/monogamy The AMPS firewall argument is often presented as a trilemma: one must abandon (i) unitarity, (ii) semiclassical effective field theory at the horizon, or (iii) the equivalence principle [ 3 ]. In the framework developed in this paper, AMPS is more naturally interpreted as a diagnostic of terminality: the contradiction arises because one implicitly demands a simultaneous and global decoding of interior information from incompatible contexts. Islands/QES resolve the tension not by violating monogamy, but by refusing the forbidden demand of a single total internal decoder and replacing it with contextual, regime-dependent decoder selection. Set-up: three subsystems and the monogamy constraint. Consider the standard AMPS partition at late times: •R= early Hawking radiation already emitted, •B= a late Hawking mode just outside the horizon, •A= the interior partner mode of Bin a semiclassical near-horizon description. Semiclassical effective field theory near the horizon predicts that A and B are highly entangled (approximately in a vacuum-like state). This suggests that Bis nearly maximally entangled with A. On the other hand, unitarity and Page-curve reasoning imply that after the Page time the newly emitted mode B must be correlated with the early radiation R in such a way that the radiation purifies itself [2]. In the strongest form, one expects Bto be highly entangled with (a subsystem of) R. The monogamy of entanglement then creates the apparent contradiction: a system cannot be simultaneously maximally entangled with two independent systems. This is the core AMPS tension. 51 Hidden assumption: a single global factorization and a single global decoder. In our language, the AMPS contradiction is not “caused by monogamy”; monogamy is simply a constraint. The contradiction arises because one silently assumes: (A) a fixed tensor-factorization of the Hilbert space that treats A , B , and R as independent subsystems in a single global description, (B) a fixed decoding semantics that simultaneously assigns meaning to A as an interior mode defined by horizon EFT and to Bas a mode purified by Rin a post-Page-time decoding. Together, (A) and (B) are precisely the terminality demand we have been isolating: they require that all contexts (the near-horizon EFT context and the radiation-decoding context) embed into one global semantics without conflict. But anti-totality says such a terminal embedding is generically forbidden in self-referential regimes (Section III; Proposition III.5). AMPS as “simultaneous total decoding across incompatible contexts.” We can now restate the paradox in the language of this paper: AMPS is the contradiction obtained by insisting on simultaneous total decoding of interior information across incompatible contexts: the horizon-EFT context (where A exists as the partner of B) and the radiation-decoding context (where Bmust be purified by R). In categorical terms, one is attempting to strictify two locally valid functorial descriptions into a single global object without acknowledging higher coherence obstructions. In diagonal terms, one is demanding terminal closure of semantic predicates about interior modes. How islands resolve AMPS in our framework. Islands/QES resolve the AMPS tension by implementing adecoder transition rather than a global total decoder. After the Page time, the entanglement wedge of the radiation includes an island region that contains the degrees of freedom that would otherwise be labeled as Ain the horizon-EFT context. Operationally, this means: • in the radiation-decoding context, the “partner” of B is not an independent A external to R , but a degree of freedom that lies in the entanglement wedge of R, • hence the purification of B by R does not require B to be maximally entangled with an independent subsystem distinct from R. In short, islands change the context so that the semantic identification of “the partner of B ” is no longer incompatible with post-Page-time purification. What changes: semantics and context, not monogamy. In this view, monogamy is not violated. What changes is: • the identification of subsystems (which degrees of freedom count as inside/outside the decoder context), •and the semantic meaning of “interior mode A” across regimes. This is exactly the anti-totality stance: refuse a single global semantics that treats all identifications as simultaneously valid. Connection to diagonal remainders. The AMPS tension can also be read as an instance of diagonal remainder structure: attempting to maintain a fixed global decoder across incompatible contexts generates non-reconstructible or inconsistent predicate assignments. The island transition is then the minimal coherence repair: rather than forcing a global assignment (which triggers contradiction), the theory changes decoding context so that the relevant predicates become coherently decidable within the selected regime. Takeaway. AMPS/monogamy does not force information loss or firewall metaphysics. It forces a recognition of anti-totality: The paradox arises from demanding simultaneous total decoding across incompatible contexts. Islands resolve it by selecting the regime-appropriate decoding context, thereby avoiding the forbidden terminal closure while preserving local semiclassical consistency. 52 VIII. HCCB: PHYSICAL MECHANISM FOR NON-TOTALITY A. Coherence vs. symmetry vs. totality The diagonal results of Section III (Key Proposition III.5) and the No-Total-Holography result of Section V (Proposition V.1) establish a semantic constraint: terminal, context-independent closure is generically forbidden in sufficiently expressive self-referential systems. The remaining question is physical: How can physics remain predictive and consistent if terminal closure is forbidden? Higher Categorical Coherence Breakdown (HCCB) is the proposed mechanism: the world maintains coherence closure rather than terminal totality. This subsection clarifies the foundational distinctions between coherence,symmetry, and totality, because confusing these is precisely how one reintroduces the forbidden demand in disguised form. Three notions that must not be conflated. We distinguish: (1) Coherence (closure under composition). Coherence is the minimal consistency condition: whenever there are multiple admissible ways to compose or compare descriptions, those ways must agree up to controlled equivalence. Categorically, coherence is encoded by commuting diagrams and higher coherence constraints (pentagon/hexagon-type conditions) [ 25 ]. Physically, coherence means: local descriptions can be glued on overlaps without contradiction. (2) Symmetry (invariance under transformations). Symmetry is stronger than coherence. A symmetry principle supplies canonical identifications by an action of a group, groupoid, or higher group on the space of descriptions. Symmetry often implies coherence (since invariance yields canonical maps), but coherence does not require symmetry. Coherence can exist even when there is no global symmetry acting on the full description. (3) Totality (terminal closure). Totality is stronger still: it is the demand that all semantic predicates about the system be decidable or representable within a single, globally valid internal description. In categorical form, totality corresponds to terminal classifiers or terminal decoders (Subsection V C), and in logical form it is a global truth/halting predicate (Section III). Diagonal logic shows that totality is the specific form of “closure” that becomes inconsistent under self-reference [10]. Why “symmetry is sufficient but not necessary.” The guiding methodological insight, which we will use repeatedly, is: Symmetry is a sufficient but not necessary condition for coherent gluing; coherence/closure is the minimal consistency requirement and symmetry emerges as a limit. This statement is not merely philosophical; it is a structural fact about how constraints arise in categories: • If a global symmetry acts, then many diagrams commute “for free” because there is a canonical map supplied by the symmetry action. •But diagrams can commute without a global symmetry, simply because the construction is forced by a universal property or because the coherence constraints are satisfied in a weaker (higher) sense. Mac Lane’s coherence theorems already illustrate this principle: associativity can be strictified in many settings, but the existence of associativity constraints does not require a symmetry group; it requires only coherence data [25]. Coherence closure as the physically correct replacement for totality. Diagonal arguments do not say “physics is inconsistent.” They say: A single terminal semantics is an over-demand. HCCB responds by replacing terminal semantics with coherence closure: •Local (contextual) descriptions remain valid. Within a fixed context (a chosen algebra of observables, code subspace, regime, observer standpoint), quantum mechanics can remain linear and unitary, and predictions are standard. 53 •Gluing of contexts is the nontrivial part. When one changes context (e.g. changes the accessible region, changes the code subspace, crosses a Page transition, changes an observer chain), one must compare descriptions. Coherence closure demands that overlapping descriptions agree where they overlap, but it does not demand that they collapse into a single global description. •Totality would demand more than gluing. Totality would demand that all these contexts embed into one global structure that decides every predicate about the system (a terminal decoder). Anti-totality forbids that demand. Why this is a mechanism, not a retreat. HCCB is not “giving up” on unitarity; it is reclassifying where unitarity belongs. It says: Unitarity is a property of local charts (contexts), not necessarily of global strictification. The new physics appears not in the local charts themselves, but in the coherence data that relates them: holonomy/defect structure in the gluing rules produces operational consequences (sectorization, history dependence, and CP effective dynamics) without requiring a fundamental abandonment of local quantum mechanics. This is developed in detail in the HCCB framework [ 11 ] and will be integrated into the holography/islands narrative in the subsequent subsections. Coherence closure and predictability. Finally, it is essential to note the epistemic payoff: Relaxing terminal totality does not weaken prediction; it strengthens prediction within admissible contexts by preventing inconsistent global demands and by preserving structured memory in the coherence data. This is the mechanism-level version of “universality without terminality”: we keep canonical universal cores wherever they exist, but we refuse terminal closure where diagonal logic forbids it. The remainder of Section VIII will formalize how higher categorical coherence constraints encode this gluing and how their breakdown yields the operational phenomena that replace total decoding. B. Contexts as a higher category / groupoid To make HCCB mathematically explicit, we must formalize what we mean by a “context” and how different contexts relate. The key methodological move is to stop treating “the theory” as a single static object and instead treat it as a family of local descriptions connected by structure-preserving maps and higher coherence data. Category theory—and, in particular, higher category theory—is the natural language for this, because it is precisely the mathematics of “things, transformations, and transformations between transformations.” Why contexts are unavoidable in physics. In practice, every physical description involves choices: •a choice of accessible algebra or subsystem (region R, radiation, observer), •a choice of coarse-graining scale (RG context), •a choice of semiclassical background (code subspace context), •a choice of gauge/coordinate presentation (presentation context), •a choice of decoding/reconstruction scheme (holography context). The anti-totality results tell us we cannot expect all of these choices to collapse into a single terminal global semantics. Instead, the correct requirement is coherence closure: overlapping contexts must agree where they overlap, but they need not globally strictify. A context groupoid (1-categorical skeleton) A minimal model is a groupoid Cof contexts: •Objects c∈Ob(C)are contexts. •Morphisms f:c→c0are admissible transitions (changes of context). 54 • Every morphism is invertible (up to isomorphism), reflecting that context changes are re-descriptions rather than irreversible dynamics. This is already enough to express “global” vs “local”: global structure would mean a single object into which all contexts embed compatibly; local structure means chart-wise definitions with transition data. However, many physically important transitions are not strictly invertible: coarse-graining, restriction to a subalgebra, and partial tracing are not invertible. Thus, in general, C is better modeled as a category (not necessarily a groupoid), or as a bicategory where invertibility holds only up to higher equivalence. Higher category viewpoint: contexts, transitions, and coherence The HCCB principle is fundamentally higher-categorical. We model contexts and their relations as a (weak) 2-category (bicategory) Ctx: (0) Objects: contexts. An object c∈Ob ( Ctx )specifies a local semantic chart in which quantum mechanics is well-defined: a Hilbert space Hc or algebra Ac , a chosen code domain, and the relevant observable/predicate family. (1) 1-morphisms: transitions between contexts. A 1-morphism f : c→c0 represents a transformation relating the two local descriptions: •an embedding/restriction of observable algebras, •a coarse-graining channel, •a change of code subspace, •a change of semiclassical background, •a reconstruction map between wedges. Composition of 1-morphisms corresponds to composing transitions: changing context in steps. (2) 2-morphisms: coherence between transitions. A 2-morphism α : f⇒g between two 1morphisms f, g : c→c0 is the data that certifies that two different transitions are equivalent as ways of relating contexts. Physically, 2-morphisms encode: •gauge equivalences between presentations, •equivalences between two coarse-graining routes, •homotopies between reconstruction prescriptions, •consistency transformations between two decoding maps that agree on overlap. This is exactly the layer where “memory” and “history dependence” live: if two paths of context change are not coherently equivalent, the system retains information about which path occurred. Standard references for bicategories and higher categorical coherence include [25, 34]. Coherence constraints: why higher morphisms matter In a strict category, associativity holds on the nose: ( h◦g ) ◦f = h◦ ( g◦f ). In a bicategory, associativity holds only up to a specified coherent isomorphism (the associator), and these associators must satisfy higher coherence constraints (pentagon identities). The physical meaning is direct: If a system can be described by multiple overlapping contexts, then “changing context” must be consistent under composition, and the consistency itself is a structured object. HCCB asserts that these higher coherence constraints can fail globally even when local coherence holds: there can exist nontrivial “holonomy” in the context space, meaning that composing transitions around a loop yields a nontrivial 2-morphism defect rather than the identity. This is the categorical analogue of curvature: local charts exist, but no global flattening exists. 55 Global vs local semantics in this language The distinction between global unitarity and HCCB-limited unitarity can now be expressed precisely: •Global unitarity as strictification: there exists a single global context c∗ such that all other contexts embed into c∗ coherently, and all 1and 2-morphisms can be strictified into identity coherence (no holonomy defects). In such a situation, one effectively has a single global Hilbert space/semantics. •HCCB-limited unitarity: contexts admit local unitary/linear semantics, but the 2-morphism coherence data has nontrivial defects, preventing strictification. Operationally, this yields path dependence and CP effective dynamics when moving between contexts. This higher-categorical formulation is the mechanism-level counterpart of the anti-totality constraint: terminal semantic closure would require global strictification, which coherence defects forbid. Takeaway. Modeling contexts as a higher category provides a mathematically explicit home for: •contextuality (many objects), •decoding/reconstruction transitions (1-morphisms), •path dependence and memory (2-morphisms and their holonomy), •and the failure of terminal closure (nontrivial coherence defects). The next subsections will show how these coherence defects manifest physically as patchwise unitarity with global gluing failure, CP flows, and sectorization, thereby making HCCB a concrete physical mechanism for non-totality. C. Patchwise unitarity Having modeled contexts as objects of a higher category in Subsection VIII B, we now clarify a central physical consequence of HCCB: unitarity is preserved locally, but need not globalize. This subsection formalizes the notion of patchwise unitarity and explains why it is the correct replacement for global unitarity once terminal closure is forbidden. What “patchwise” means. Apatch is a context c∈Ctx in the sense of Subsection VIII B: a choice of accessible algebra, code domain, semiclassical background, and decoding semantics. Patchwise unitarity asserts that: Within each context c , the effective dynamics is linear and unitary (or unitarily implementable) on the corresponding Hilbert space or algebra. This is a statement about local charts, not about their global gluing. Local unitary evolution in a fixed context Fix a context c with Hilbert space Hc (or algebra Ac ). Patchwise unitarity means that time evolution within this context is implemented by a one-parameter unitary group (or automorphism group) Uc(t) : Hc−→ Hc, Uc(t1+t2) = Uc(t1)Uc(t2),(81) or equivalently, in the Heisenberg picture, by a *-automorphism αc t : Ac→ Ac . All the familiar machinery of quantum mechanics applies within c : superposition, interference, conservation of probability, and unitary symmetry. Physical examples. • In semiclassical gravity, c may be a fixed background geometry with small quantum fluctuations; EFT evolution on that background is unitary. • In holography, c may be a fixed code subspace and boundary region; operator-algebra QEC ensures that bulk evolution restricted to that code is unitarily represented (Subsection VI A). • In measurement theory, c may correspond to a fixed observer chain with a specified pointer basis; within that chain, evolution is unitary prior to coarse-graining. 56 Why local unitarity does not imply global unitarity The crucial point is that local unitary dynamics in each context does not imply the existence of a single global unitary acting on a universal Hilbert space. To obtain global unitarity one would need to: 1. embed all Hcinto a single Hilbert space Hglobal, 2. identify all context changes as unitary transformations on Hglobal, 3. and require that all coherence 2-morphisms strictify to identities. This is precisely the strictification or terminal closure demand that diagonal logic and HCCB forbid in self-referential regimes. Nontrivial coherence data (Subsection VIII B) obstructs this global embedding even when each local chart is perfectly unitary. Patch transitions and effective non-unitarity Consider a transition between contexts f:c→c0. Physically, fmay represent: •restriction to a smaller accessible algebra, •inclusion of additional degrees of freedom, •coarse-graining or tracing out, •a decoder change (e.g. an island transition). Such transitions are generally not invertible and therefore cannot be implemented by a unitary on Hc . Instead, the induced map on states is a completely positive trace-preserving (CPTP) map Φc→c0:S(Hc)−→ S(Hc0),(82) which can always be dilated to a unitary on a larger space (Stinespring), but not within the restricted context itself. This is not a failure of quantum mechanics; it is the operational signature of moving between contexts. Connection to open-system dynamics. When a sequence of context transitions is coarse-grained in time, the effective evolution appears non-unitary and is well-described by a GKLS/Lindblad generator in appropriate limits [ 35 ]. In the HCCB picture, such effective non-unitarity does not signal a breakdown of local quantum mechanics; it signals the presence of coherence defects in the gluing of contexts. Patchwise unitarity and predictability A common worry is that abandoning global unitarity undermines predictability. Patchwise unitarity resolves this concern: •Predictions within a context remain as sharp and reliable as in ordinary quantum mechanics. • Transitions between contexts are governed by physically controlled CPTP maps, whose informationtheoretic properties (monotonicity, contractivity) are well understood. • Apparent non-unitarity arises only when one insists on a single global description across incompatible contexts. Thus HCCB preserves what is operationally required for physics while refusing the forbidden global strictification. Relation to “global unitarity undecidability.” From this viewpoint, “global unitarity” is not a directly testable physical property but a claim about the existence of a strictification of the context bicategory into a single Hilbert space. Whether such a strictification exists may be undecidable from within the theory, even if local unitarity holds everywhere. This perspective reframes debates about “fundamental unitarity” as questions about coherence data rather than as empirical contradictions. 57 Takeaway. Patchwise unitarity is the precise physical content of HCCB: Quantum mechanics remains locally unitary in each admissible context, while global unitarity is replaced by coherence closure across contexts. This is how physics implements anti-totality without sacrificing local predictability. In the next subsections we will show how coherence defects manifest operationally as sectorization, memory, and completely positive effective dynamics. D. HCCB: higher coherence obstruction / holonomy defect In Subsection VIII C we established patchwise unitarity: each context admits a linear/unitary semantics. This subsection identifies the precise mechanism by which global unitarity/linearity may fail without local inconsistency: a higher coherence obstruction, equivalently a holonomy defect in the higher category of contexts. This is the mathematical core of Higher Categorical Coherence Breakdown (HCCB) [11]. From local charts to global strictification. The foundational distinction is analogous to differential geometry: •A manifold can be described locally by coordinate charts. •A single global chart exists only if the manifold is globally trivial in an appropriate sense. •Curvature/holonomy obstructs global flattening. In HCCB, the “charts” are quantum contexts (Subsection VIII B), and the analogue of curvature is the failure of higher coherence data to trivialize. The result is: local linear/unitary charts exist, but there is no globally strictified semantics in which all charts become one. Coherence data in a bicategory of contexts Let Ctx be a bicategory of contexts: objects c , 1-morphisms f : c→c0 (context transitions), and 2-morphisms α : f⇒g (coherence between transitions). In a bicategory, composition of 1-morphisms is associative only up to a specified associator 2-isomorphism. Thus, for composable 1-morphisms f, g, h there are two composite 1-morphisms (h◦g)◦fand h◦(g◦f)and a canonical 2-isomorphism ah,g,f : (h◦g)◦f⇒h◦(g◦f).(83) The associators must satisfy the pentagon identity (Mac Lane coherence) [ 25 ]. This expresses a basic form of higher coherence: different parenthesizations yield coherently equivalent composites. HCCB is not a violation of coherence axioms. The statement “coherence breakdown” does not mean the bicategory axioms fail. Rather, it means that when contexts are interpreted physically, there can exist nontrivial holonomy in the coherence data: around closed loops of context transitions, the composite is not coherently equivalent to the identity in a trivial way. This is the analogue of having a nontrivial connection with curvature. Holonomy defect around a loop Consider a loop of context transitions c0 f1 −→ c1 f2 −→ c2 f3 −→ · · · fn −→ cn=c0.(84) Form the composite 1-morphism F:= fn◦ · · · ◦ f2◦f1:c0→c0. If the context space were globally strictifiable (terminal closure), one would expect F to be coherently equivalent to the identity 1-morphism idc0 in a way that can be trivialized globally. In a higher categorical setting, one asks whether there exists a 2-isomorphism Ω : F⇒idc0,(85) and, crucially, whether such Ωcan be chosen coherently across all loops and all compositions. A holonomy defect is the situation in which: 64 Context skeleton and transport data. For concreteness, we begin with the 1-categorical skeleton: a groupoid C of contexts. Objects are contexts c∈Ob ( C ); morphisms f : c→c0 are invertible context changes. (We discuss the higher-categorical refinement at the end of the subsection.) Apatchwise unitary semantics assigns: 1. a Hilbert space Hcto each context c, 2. for each morphism f:c→c0, a unitary transport map Uf:Hc−→ Hc0,(94) 3. and (optionally) local unitary time evolutions Uc(t)within each Hc(Subsection VIII C). Projective functoriality and the cocycle ω.Global strictification would require strict functoriality: Ug◦f=UgUf. In HCCB-limited settings, one expects only projective functoriality: UgUf=ω(g, f)Ug◦f, ω(g, f)∈U(1),(95) for composable f : c0→c1 , g : c1→c2 . Here ω ( g, f )is the coherence factor measuring the failure of strict composition. (More generally ω ( g, f )may take values in central unitaries in an operator algebra; the U(1) case already captures the essential obstruction structure.) Associativity of composition implies the 2-cocycle condition: for composable f, g, h, ω(h, g ◦f)ω(g, f) = ω(h◦g, f)ω(h, g).(96) Thus ωdefines a cohomology class [ω]∈H2(C, U(1)),(97) in groupoid cohomology (or equivalently in the cohomology of the nerve of C ). Standard references for the cocycle/trivialization viewpoint are [39]. Gauge of transports and trivialization. We may rephase the transport unitaries by a 1-cochain λ(f)∈U(1): Uf7→ U0 f:= λ(f)Uf.(98) Under this change, the cocycle transforms by a coboundary: ω(g, f)7→ ω0(g, f) := λ(g)λ(f) λ(g◦f)ω(g, f).(99) Therefore, the obstruction to strict functoriality is exactly the cohomology class [ω]: [ω] = 0 ⇐⇒ there exists a rephasing making ω0= 1. Global unitarity as strictification We can now state a clean criterion. Proposition IX.1 (Strictification criterion for global unitarity) . Assume a patchwise unitary transport system {Uf} over the context groupoid C satisfying (95) with coherence cocycle ω . Then the following are equivalent: (i) (Vanishing coherence class)[ω]=0in H2(C, U(1)). (ii) (Strict functorial transport) There exists an equivalent choice of transports {U0 f} (related by rephasing (98)) such that U0 g◦f=U0 gU0 ffor all composable f, g. (100) 65 (iii) (Global trivialization) There exists a single Hilbert space H∗ and unitary embeddings Jc : Hc→ H∗ such that for every f:c→c0, U0 f=J−1 c0Jc(on the image of Jc), (101) so that all context changes are realized by strict unitary identifications inside one global semantics. Proof sketch. (i) ⇒ (ii): if [ ω ] = 0, then ω is a coboundary, so there exists λ such that ω0 in (99) is identically 1. This yields strict transport (100). (ii) ⇒ (iii): choose a reference context c0 and set H∗ := Hc0 . For any c , pick a morphism p : c→c0 and define Jc := U0 p . Because transport is strict and C is a groupoid, Jc is independent of the choice of p up to the (now trivial) coherence. This defines a coherent family of embeddings realizing (101). (iii) ⇒ (i): if a global trivialization exists, then transports compose strictly by construction, hence ω = 1 and [ω] = 0. Interpretation. Proposition IX.1 formalizes exactly what “global unitarity” means in our framework: Global unitarity is the existence of a global trivialization of the context transport system, equivalently the vanishing of a coherence/holonomy class [ω]. If [ ω ] 6 = 0, then no global strictification exists: one may have patchwise unitarity in each context, but there is no single global Hilbert space semantics compatible with all context transitions. That is precisely HCCB-limited unitarity. Higher-categorical refinement When contexts form a bicategory (rather than a groupoid) with nontrivial 2-morphisms, the obstruction class can live one degree higher (often as a 3-cocycle measuring associator holonomy). The precise cohomological degree depends on the chosen model of higher coherence. The conceptual content remains unchanged: Global unitarity corresponds to strictification/trivialization of higher coherence data; HCCB corresponds to a nontrivial coherence class obstructing such strictification. We use the groupoid/2-cocycle presentation here because it makes the strictification criterion fully explicit with minimal overhead; the higher-categorical generalization follows the same pattern. C. Why internal certification is impossible in general Having defined “global” precisely (Subsection IX A) and formalized global unitarity as strictification/trivialization of a coherence class [ ω ](Subsection IX B), we now address the key epistemic question: Why can an internal observer (or any internal procedure) not, in general, certify whether global unitarity holds? The answer has two logically independent components: 1. an operational limitation (restricted access to algebras and finite data), and 2. a diagonal limitation (self-reference forbids total internal certification). Either limitation already undermines terminal certainty; together they strongly motivate HCCB-limited unitarity as the correct stance. (I) Operational limitation: accessible algebras are always restricted Accessibility is context-dependent. By construction, any internal agent inhabits a context c with an accessible algebra Ac (Sections II and VIII). Even if there exists some larger algebra Aglobal , the agent’s operational content is limited to Ac and to transformations implementable within that context. This is not a matter of technological limitation; it is the definition of “internal observer.” Consequently, the agent can only ever observe a restricted set of correlators, expectation values, and channel actions. 66 Global unitarity is not a local observable. Global unitarity, in our precise sense, is the existence of a strictification/trivialization over the entire context groupoid (Subsection IX B). That is a global property of the entire transport/coherence system, not a property of any single local algebra Ac . No finite list of local expectation values determines a global cohomology class [ ω ]in general. Even in ordinary geometry, local measurements cannot determine global topology without global access; similarly, local operational data cannot certify global strictification without global access to the context space. Open-system ambiguity: unitarity vs environment. Operationally, any restricted description is governed by CPTP maps, not unitaries (Subsection VIII E). Given only restricted data on Ac , an apparent nonunitary evolution can always be explained either as: •unitary evolution on a larger system with an environment (Stinespring dilation), or •intrinsic effective non-unitarity due to coherence defects between contexts. From within a restricted algebra, these two explanations are generally underdetermined by data. Thus operational access alone cannot certify global unitarity. (II) Diagonal limitation: the system cannot certify its own closure The operational limitation already prevents certainty in practice. The deeper reason is logical: certification of global closure is itself a self-referential predicate, and diagonal logic forbids its total internal decidability. Certification as a semantic predicate. Consider the predicate: “The coherence class [ω]vanishes, hence global strictification exists.” This is a semantic predicate about the system’s own global structure: it asserts the existence of a single global semantics into which all contexts embed. Therefore, it belongs to the class of “information about the system” predicates discussed in Section II B. Internal certification is a total-decoder demand. To internally certify global unitarity would mean that there exists an internal procedure which, from within the system, decides this predicate correctly in all relevant cases. But this is exactly the structure forbidden by diagonal logic: once the system can represent and evaluate its own descriptions and can ask reflection queries about those evaluations, no total internal classifier exists (Proposition III.5). In short: A system cannot, in general, certify its own terminal closure from within, because the certification predicate itself becomes subject to diagonal reflection. Analogy with Gödel/Turing. The pattern is identical to: • Gödel/Tarski: a sufficiently expressive consistent theory cannot define (and hence cannot decide) its own truth predicate [8, 26]. • Turing: a sufficiently expressive computational model cannot decide halting for all programs, including those that refer to the decider [9]. Global unitarity certification is of the same logical type: it is a global semantic property that would require a total internal decider if one demanded certainty in all cases. The self-reference loop is physical. In the black-hole setting, this is not merely an analogy. Decoding is a physical process; queries about the decoding are physical queries; the system can incorporate observers who attempt to certify global closure. Hence the diagonal loop is operationally meaningful, and the impossibility is physical: a total global-certification procedure is an over-demand in a self-referential world. Synthesis: why “global unitarity” should be treated as non-certifiable Combining (I) and (II), we obtain the core epistemic message: Even if global unitarity were true, internal agents generally cannot certify it: operational access is restricted to local algebras, and diagonal logic forbids total internal deciders of terminal closure predicates. 67 Therefore, physics should be formulated in a way that does not require such certification. This is precisely what HCCB provides: local/patchwise unitarity with coherence closure across contexts, allowing nontrivial holonomy classes [ ω ]without inconsistency. In this view, global unitarity becomes a contingent global property (equivalent to [ ω ] = 0) that may hold in special cases, but it is not a universally certifiable axiom. Takeaway. The non-certifiability of global unitarity is not a philosophical resignation but a rigorous consequence of (i) operational restriction and (ii) diagonal anti-totality. It provides the correct epistemic backdrop for islands as decoder selection and for HCCB as a physical mechanism of non-totality. D. Undecidability schema This subsection states an undecidability schema that complements the epistemic non-certifiability discussed in Subsection IX C. The point is not to claim that every physically realizable context system gives an undecidable decision problem; rather, the point is structural: Once the space of contexts is allowed to be sufficiently expressive (in the same sense that allows Turing-complete computation), the question “does the coherence class [ ω ]vanish?” can become algorithmically undecidable in families of finitely presented models. This provides a mathematically sharp sense in which “global unitarity” (identified with [ ω ]=0in Subsection IX B) is, in general, not something one should expect to be decidable by any uniform internal certification procedure. Why an undecidability schema (and not a universal theorem). Our claim is deliberately scoped: • We do not claim that [ ω ]=0is undecidable for every physically meaningful model. Physics may impose additional regularity constraints. • We do claim that there exist natural families of finitely presented context systems (e.g. finitely presented groupoids with specified cocycles) for which the decision problem “is [ ω ] = 0?” is undecidable, by reduction from known undecidable problems. This is the standard posture in logic and computation: undecidability results are statements about classes of formal objects, not about every special subclass one might care about. Decision problem: strictification/trivialization of [ω] Fix a class of finitely presented context groupoids C (or groupoids presented by generators and relations) together with U(1)-valued 2-cocycles ωas in (95)–(97). Consider the decision problem: Input: a finite presentation of (C, ω). Question: does [ω]=0in H2(C, U(1))? Output: YES if [ω] = 0 (global strictification exists), NO otherwise. In the language of Subsection IX B, this is the decision problem “does a global trivialization/strictification (global unitary semantics) exist?” Undecidability claim (schema) Proposition IX.2 (Undecidability schema for vanishing coherence class) . There exist families of finitely presented context groupoids C and finitely specified cocycles ω such that the decision problem “is [ ω ] = 0?” is algorithmically undecidable. In particular, there is no Turing machine which, given an arbitrary finite presentation from such a family, halts and correctly decides whether [ω]=0. 68 Interpretation. This proposition says that, in sufficiently expressive families, global strictification is as hard as the halting problem. It strengthens (but does not replace) the non-certifiability argument of Subsection IX C: • non-certifiability: internal observers cannot, in general, certify global closure from restricted operational access and diagonal limitations; • undecidability schema: even as a purely formal decision problem on finite presentations, [ ω ]=0 may be undecidable in expressive families. Reduction idea (sketch; full details deferred) We briefly indicate why Proposition IX.2 is plausible and standard in spirit. Encoding computation into finite presentations. A classical route to undecidability is to encode Turing computation into finitely presented algebraic objects. There are well-known undecidability theorems for finitely presented groups/semigroups, notably the unsolvability of the word problem and related decision problems [40, 41]. The underlying reason is that finite presentations can simulate computation. From computation to cohomology obstructions. Given such an encoding, one constructs a family (CM, ωM)indexed by Turing machines Mwith the property: Mhalts on a specified input ⇐⇒ [ωM]=0.(102) If such a construction exists, then a decision procedure for [ ω ] = 0 would decide the halting problem, which is impossible [ 9 ]. The construction uses the fact that strictification corresponds to solving a coherence equation (finding a 1-cochain λ such that ω = δλ ), and this equation can be arranged to have a solution iff a simulated computation halts. Where the proof lives. A complete reduction requires careful specification of the chosen presentation formalism and the precise mapping M7→ ( CM, ωM ). We therefore present the reduction as an appendixlevel schema: Appendix D will outline a concrete construction strategy and state the precise assumptions under which (102) holds. Physics reading The physics relevance is conceptual rather than computational. If global unitarity corresponds to vanishing of a coherence class [ω]over the space of contexts, then: “Global unitarity” is a global strictification property whose verification may be non-decidable in expressive families and non-certifiable from within restricted operational contexts. This reinforces the methodological stance of HCCB: physics should be formulated in terms of coherence closure and contextual semantics, not in terms of a terminal global strictification axiom that is neither operationally accessible nor algorithmically decidable in general families. E. Experimental distinguishability Sections IX A–IX D make two statements that can sound pessimistic if read carelessly: (i) “global unitarity” is a global strictification property over the context space, and (ii) it may be non-certifiable from within and even undecidable in expressive families. This subsection addresses the natural question: Can global unitarity be experimentally distinguished from HCCB-limited (patchwise) unitarity? The answer has two parts. 1. In principle, yes in sufficiently controlled simulators that allow deliberate navigation through context space and measurement of holonomy/path dependence. 2. In practice for astrophysical black holes, likely no with any foreseeable technology, because the required control over Hawking radiation and interior-sensitive correlators is far beyond reach. However, the conceptual lesson remains important: physics should not rely on a notion that cannot be operationally certified in the regimes where self-reference is maximal. 69 What would distinguish global from HCCB-limited unitarity? By Proposition IX.1, “global unitarity” in our sense is equivalent to [ ω ] = 0, i.e. trivial holonomy of the context transport system. HCCB-limited unitarity corresponds to [ ω ] 6 = 0, i.e. nontrivial coherence holonomy. Thus the distinguishing signature is not “unitary vs non-unitary evolution” on a restricted subsystem (which is always CPTP); it is: path dependence of effective evolution/decoding under closed loops in context space. Concretely, if two different context paths γ1, γ2 : c→c0 induce different effective channels Φ γ16 = Φ γ2 (Equation (92) ), then strictification fails and HCCB-limited behavior is present. If all such differences can be eliminated by a global trivialization (a choice of embeddings Jc ), then the system is global in the strictified sense. In principle: controlled simulators and holonomy tests Why simulators are the right arena. Laboratory quantum platforms allow one to engineer: •a tunable “environment” and accessible subsystem, •explicit coarse-graining and decoding procedures, •repeated preparation and controlled protocol sequences, •and careful process tomography of effective channels. These capabilities are precisely what is needed to probe path dependence and context holonomy. Operational holonomy test (protocol-level statement). Choose a set of contexts {ci} and transitions {fi : ci→ci+1} that implement different but nominally equivalent coarse-graining/decoding routes. Run two different sequences γ1, γ2 with the same endpoints (same initial and final nominal context) and reconstruct the induced effective CPTP maps Φ γ1, Φ γ2 via standard quantum process tomography. If Φ γ16 = Φ γ2 beyond experimental uncertainty and beyond trivial gauge choices (rephasing/strictification), this is a direct operational signature of coherence holonomy and hence of HCCB-limited unitarity. Relation to non-Markovianity diagnostics. Holonomy-induced path dependence is closely related to quantum non-Markovianity: if effective evolution depends on history (path), then CP-divisibility typically fails. Thus one can use non-Markovianity witnesses (e.g. information backflow measures) as partial diagnostics, though they are not equivalent to holonomy in full generality [ 38 ]. The conceptual distinction is that HCCB emphasizes the geometric origin of memory: nontrivial gluing of contexts rather than merely environmental correlations. Quantum gravity analogues in toy models. In principle, one could attempt analogous tests in holographyinspired toy systems: random unitary circuits, tensor-network codes, SYK-like models coupled to baths, or other scrambling systems where “radiation” and “interior” degrees of freedom can be engineered. The goal would be to probe whether decoder selection exhibits holonomy/path dependence under protocol loops, thereby distinguishing a strictified global picture from HCCB-limited contextuality. (We treat these as future directions rather than as claims of existing experiments.) In practice: astrophysical black holes For astrophysical black holes, experimental distinguishability is overwhelmingly constrained. Control and access limitations. To test global unitarity directly one would need: •near-complete control of Hawking radiation degrees of freedom, •the ability to perform complex joint measurements on early and late radiation, •and enough control to implement different “decoder paths” and compare them. Even ignoring computational complexity, this level of control is not remotely feasible. Moreover, the radiation is extraordinarily weak for macroscopic black holes, and decoherence with the environment is unavoidable. 70 What remains meaningful. What remains meaningful is not direct certification, but consistency conditions: Page-curve behavior in controlled models, internal consistency of semiclassical gravity with islands/QES, and the coherence-first stance advocated here. Thus, for black holes, “experimental distinguishability” should be interpreted primarily as: •indirect support for contextual decoding mechanisms (islands/QES), •consistency of entanglement-wedge reconstruction as a QEC statement, •and the absence of any need to demand terminal global decoding. Takeaway In summary: Global unitarity versus HCCB-limited unitarity is experimentally distinguishable in principle via holonomy/path dependence tests in controlled simulators, but is likely inaccessible for astrophysical black holes. This is not a weakness of the framework; it is a reminder that terminal closure is not an operationally meaningful demand in maximally self-referential regimes. Physics remains predictive by relying on coherence closure and contextual decoding, not on global strictification that cannot be certified in the relevant regimes. X. BEYOND BLACK HOLES: SAME STRUCTURE IN OTHER QUANTUM CONTEXTS A. AQFT Type-III algebras Black holes provide a maximally self-referential arena where anti-totality becomes hard to ignore. However, the underlying constraint is not uniquely gravitational. A particularly clean non-gravitational realization is Algebraic Quantum Field Theory (AQFT), where the local observable algebras are typically Type III von Neumann algebras. Type III structure forces a conceptual shift away from “global density matrices” and toward relational information measures (modular/relative entropy), which aligns naturally with the coherence-first, anti-terminal framework developed in this paper. Local algebras and nets. In AQFT, one associates to each spacetime region O a von Neumann algebra A ( O )of observables localized in O , forming a net that satisfies isotony, locality, and covariance [ 42 ]. The fundamental object is the net itself, not a global Hilbert-space factorization. States are positive normalized linear functionals on these algebras (or represented on a Hilbert space via the GNS construction). Type-III phenomenon: no trace, no density matrix for local restrictions. A central structural result is that, in relativistic QFT under standard assumptions, local algebras A ( O )are typically Type III. Among other consequences, Type III factors admit no faithful normal trace and thus do not support the naive “density matrix on the local algebra” intuition familiar from finite dimensions [ 42 ]. Operationally, this means: • there is no globally meaningful “reduced density matrix” for a local region in the usual trace-class sense, • entanglement entropy of sharp spatial subregions is generically UV-divergent and not an intrinsic algebraic quantity, •subsystem factorization is not a primitive notion; the net of algebras is. This is a concrete and rigorous example of why totality demands are suspect: the most naive terminal object one might hope for—a global trace-class reduced state that encodes all local information—does not exist in the intended category. 71 Relational information and the primacy of relative entropy. Although local density matrices are not available in the naive way, the theory remains perfectly predictive. The correct invariants are relational quantities, especially relative entropy and modular data. Relative entropy D ( ρkσ )is defined for normal states on von Neumann algebras and obeys the same conceptual data processing properties as in finite dimensions; it is the proper measure of distinguishability and information content in the algebraic setting (Subsection II A). Thus AQFT already enforces a key lesson of this paper: Operational information is not primarily a global “state object” but a relational, contextdependent structure of distinguishability between states on accessible algebras. This aligns directly with our insistence that “information about the system” must be interpreted relative to contexts and probe algebras. Modular structure and non-terminal semantics. The Tomita–Takesaki modular theory associated to a faithful normal state on a von Neumann algebra provides a canonical modular automorphism group. We do not develop modular theory here, but we note its conceptual relevance: it provides canonical dynamics and information structure intrinsic to the algebra and state without appealing to a terminal global factorization. In this sense, AQFT exemplifies universality without terminality: canonical structures exist (modular flow, relative entropy), but terminal global objects (density matrices/traces for local algebras) do not. Connection to anti-totality and HCCB. Type III structure provides a concrete operator-algebraic realization of anti-totality: the attempt to force a single global density-matrix semantics for local physics fails mathematically. Yet coherence closure persists: local algebras glue consistently as a net, and operational physics is encoded in functorial restrictions to subalgebras. This is precisely the mechanismlevel picture of HCCB: local charts (contexts) are consistent, but naive global strictification into a single terminal semantics is obstructed. Operationally, the agent’s dynamics is described by channels and conditional expectations, not by a single global unitary acting on a factorized Hilbert space. Takeaway. AQFT Type III algebras show that the black-hole lesson is not exotic: even in ordinary QFT, the mathematically correct notion of information is relational and contextual, and terminal closure in the form of global density matrices/traces for local physics is unavailable. This provides strong independent support for our coherence-first principle and for the idea that universality can survive while terminality fails. B. Gauge theories and edge modes Gauge theories provide a second non-gravitational arena where the anti-totality theme becomes mathematically sharp. The relevant phenomenon is not exotic: it is the failure of naive subsystem factorization in the presence of constraints (Gauss laws) and the consequent necessity of additional boundary degrees of freedom (edge modes) or center data to define local subsystems consistently. From the perspective developed in this paper, this is an archetypal instance of: universality without terminality: local/patchwise descriptions exist and glue coherently, but no single terminal factorization or total internal subsystem decoder exists. Why factorization is nontrivial in gauge theories. In ordinary finite-dimensional quantum systems one often begins with a tensor factorization H∼ =HR⊗ H ¯ R and defines entanglement entropy by partial trace. Gauge theories obstruct this naive move because the physical Hilbert space is not a simple tensor product of region Hilbert spaces: it is defined by a constraint (Gauss law) that couples degrees of freedom across the boundary of a region. Equivalently, the gauge-invariant local operator algebra typically has a nontrivial center associated to boundary fluxes. Local algebras and centers Algebraic statement. Let AR denote a candidate algebra of gauge-invariant observables “localized in R .” In gauge theory one encounters the fact that AR and A¯ R do not generate the full physical algebra by simple tensor product, and that ARoften has a nontrivial center: Z(AR) := AR∩ A0 R6=C1.(103) 72 Physically, central elements correspond to boundary charge/flux data that is shared between R and ¯ R . This destroys the naive notion that “ R ” is a standalone subsystem with a unique reduced state and a unique entanglement entropy independent of choices. Implication: multiple inequivalent subsystem notions. Because the center is nontrivial, there is not a unique choice of “the” local algebra. Different prescriptions (sometimes described as “electric center” vs “magnetic center”) correspond to different operational choices about what boundary data is included among the observables. Casini, Huerta, and Rosabal analyze these issues systematically and show how entanglement entropy for gauge fields depends on this algebraic choice. This is a concrete, mathematically clean manifestation of contextuality: “information about the system” is relative to the chosen probe algebra. Edge modes and extended Hilbert spaces Necessity of extensions. One way to restore a useful notion of locality is to enlarge the description by including additional boundary degrees of freedom. Donnelly and Freidel show that defining local subsystems in gauge theory and gravity requires an extended phase space in which edge modes live on the boundary and carry the information needed to glue regions consistently. In the extended description, one recovers a factorization-like structure, but only at the cost of adding exactly the “missing” boundary data. Interpretation: edge data as coherence/gluing data. From our viewpoint, edge modes are not optional decorations; they are the concrete physical realization of coherence closure: the additional boundary degrees of freedom encode the compatibility constraints needed to glue local charts. This is precisely the same structural role played by islands in the black hole setting: islands adjust the decoding context so that the gluing of entanglement wedges remains coherent; edge modes adjust the subsystem description so that gauge constraints glue consistently. Anti-totality reading: why non-factorization is expected Non-factorization as refusal of terminal closure. A naive demand for a unique subsystem factorization in gauge theory is a terminality demand: it asks for a single globally valid decomposition that would allow all local questions to be answered by fixed subsystems. Gauge constraints forbid this, and the center/edge-mode structure is the mechanism that replaces it. Thus, non-factorization is not a bug; it is the gauge-theory analogue of our anti-totality principle. Diagonal remainders and sectorization. The presence of a center produces sector labels (flux sectors) that cannot be eliminated by local operations. This is a concrete instance of “diagonal remainders” discussed earlier: there are semantic distinctions (sector membership) that are not decidable by purely local operators in a fixed algebra without including the appropriate boundary data. The correct response is not to demand total decoding, but to classify sectors and glue them coherently—exactly the coherence-first philosophy of HCCB. Connection to HCCB. Gauge theories therefore offer a clean laboratory for HCCB ideas: local unitarity holds within a chosen algebraic context, but global strictification into a single terminal subsystem picture fails; operationally one obtains contextual CP flows when restricting observables and sector-dependent descriptions. Takeaway. Gauge theories show, in a mathematically controlled setting, the same structural lesson as black holes: Local predictability survives, but terminal totality fails; coherence is maintained by additional gluing data (centers/edge modes) rather than by a single globally factorized semantics. This strengthens the claim that the black-hole “totality paradox” is not a special feature of gravity, but a universal feature of self-referential constrained systems. C. Measurement chains (Wigner, Frauchiger–Renner) Measurement theory provides a non-gravitational arena where self-reference is not an abstract metaphor but an operational feature: observers are physical systems, measurement outcomes are physical records, and one can (at least conceptually) treat observers as quantum systems whose states are described by 73 other observers. This is precisely the setting in which terminal totality demands become inconsistent. The Wigner’s friend scenario and its sharpenings (notably Frauchiger–Renner) therefore fit naturally into the anti-totality/HCCB framework developed in this paper. Two levels of description. A minimal measurement chain contains: •a system Sto be measured, •an observer/apparatus F(“the friend”) who measures Sand records an outcome, • and a higher-level observer W (“Wigner”) who treats the joint system ( S + F )as a quantum system. The essential tension is not about whether unitary evolution holds locally—it can hold in each context— but about whether one can demand a single global semantics that makes all contexts simultaneously and terminally compatible. Wigner’s friend as context dependence Wigner’s original discussion highlights that the “friend” F experiences a definite outcome (a context in which a classical record exists), while Wigner W can, in principle, assign a coherent superposition to the combined system (S+F). These are two distinct contexts: •the friend context: a coarse-grained algebra in which the record is definite, •the Wigner context: a finer-grained algebra in which S+Fevolves unitarily. The crucial point is that both contexts can be locally consistent, but their relation is not captured by a terminal global description without additional coherence structure. In the language of this paper, the tension appears precisely when one demands: a single terminal truth/decoding map that simultaneously assigns definitive predicates to the friend’s record and maintains the global unitary superposition in Wigner’s description. This is a measurement-theoretic instance of the totality paradox. Frauchiger–Renner: no total self-description Frauchiger and Renner sharpened the Wigner-friend logic by considering a multi-agent measurement chain and imposing a set of seemingly reasonable assumptions about how agents use quantum theory to reason about each other’s predictions. Their conclusion is that: Quantum theory cannot consistently describe the use of itself. In our terms, Frauchiger–Renner is a direct demonstration that a total, globally consistent, agentindependent semantics is an over-demand once self-reference is present. More precisely: •each agent’s local description can be coherent, • but insisting on a single global semantics that simultaneously validates all agents’ inferences (including inferences about inferences) leads to contradiction. This is exactly the same logical structure as diagonal no-go results: closure of the query class under reflection (“what does the other agent predict about what I predict?”) triggers inconsistency with terminal closure. Anti-totality reading Measurement chains therefore illustrate the distinction between: •patchwise unitarity: each context admits a unitary description of the degrees of freedom it treats as quantum (Subsection VIII C), •non-terminal gluing: no single global description can simultaneously incorporate all contexts with a terminal truth/decoder map once the agent structure is internalized. The correct response is not to deny quantum mechanics, but to adopt a coherence-first stance: contexts glue only up to controlled coherence data, and some global totalizations are forbidden. 80 A minimal principle: coherence closure. We can formulate the coherence-first stance as a minimal principle: Coherence Closure Principle. A fundamental theory must provide local/sectorial descriptions together with transition maps and higher coherence data such that overlapping descriptions agree on overlaps up to coherent equivalence; no further demand of terminal global decoding or strictification is required. This principle is deliberately weaker than “global unitarity” as a terminal claim, but it is strong enough to constrain physics: it dictates the admissible forms of context change (channels, embeddings, decoder selection), and it explains why certain global questions are ill-posed. Why coherence closure is the right “replacement” for total decoding. Total decoding demands that a single internal procedure decide all semantic predicates about a system from within. Coherence closure demands only that when two contexts overlap, they agree on what they both can meaningfully talk about. This is exactly what EWR/QEC provides (agreement on a code algebra in overlapping regions), and what islands/QES enforce by selecting the context in which such agreement remains coherent. Thus coherence closure is not an alternative to the modern picture; it is the principle that makes the modern picture conceptually inevitable. Operational manifestation: CP dynamics and contextual decoding. Coherence-first physics predicts that when one restricts to an accessible algebra and moves between contexts, the operational dynamics is CPTP and generally history dependent (Subsections VIII E and IX A). Markovian semigroups are special limits, not axioms. This is not a breakdown of physics; it is a rigorous consequence of gluing local charts without terminal closure. Relation to HCCB. Higher Categorical Coherence Breakdown is the proposed mechanism-level formalization of the coherence-first principle: local unitary charts exist (patchwise unitarity), but global strictification is obstructed by higher coherence holonomy, yielding CP flows, sectorization, and memory. Thus HCCB is not an additional hypothesis on top of coherence-first physics; it is the natural way coherence-first physics becomes nontrivial in self-referential regimes. Why this increases predictive power rather than reducing it. A terminal global description would force an impossible compression of context and history into a single state/decoder, which either produces contradiction (diagonal collapse) or destroys relevant structure. Coherence closure instead preserves precisely the information that matters for local prediction: contextual universal cores, structured memory in higher coherence data, and stable behavior under admissible morphisms. Thus abandoning terminality is not surrender; it is a way to preserve maximal predictability consistent with logic. Takeaway. Coherence-first physics replaces the misleading expectation of terminal totality by a minimal, structurally stable requirement: Physics should demand coherence closure (consistent gluing of contexts), not total symmetry or total decoding (terminal closure). This is the conceptual principle that unifies diagonal logic, modern holography/islands, and HCCB into a single framework. C. Why string theory cannot be terminal but may be universal (conditionally) A common sociological framing in high-energy theory is that string theory is (or was intended to be) a “theory of everything.” In the language developed in this paper, that slogan is ambiguous, because it conflates two categorically distinct aspirations: 1. Terminality: a single, context-independent, globally strictified description that internally closes over all semantic predicates (a terminal decoder / total classifier). 2. Universality: a canonical framework that generates and organizes a broad class of consistent quantum-gravitational descriptions, with different regimes related by universal properties and natural equivalences (initial cores in appropriate fibers). Our results strongly constrain the first aspiration and clarify the only coherent sense in which the second may be pursued. The correct mature target is therefore: Universality without terminality. 81 Why string theory cannot be terminal (and why no theory can). Terminality is the demand for a single context-independent global semantics: one Hilbert space/algebra, one global decoding, one strictified gluing across all contexts, and closure under reflection. The diagonal spine (Section III) shows that sufficiently expressive self-referential systems cannot admit terminal internal classifiers; the HCCB mechanism (Section VIII) shows how physics can realize this as nontrivial coherence holonomy obstructing global strictification. These constraints do not single out string theory; they apply to any candidate UV completion of quantum gravity. Therefore, string theory cannot be a terminal “theory of everything” in the strong sense of a globally strictified, context-independent decoder of all information about the system. A practical corollary is conceptual: expecting a single final formulation with a unique set of fundamental degrees of freedom is an over-demand. Dualities, code-subspace descriptions, and regime-dependent reconstructions are not anomalies to be removed; they are exactly the kind of non-terminal structure forced by self-reference and coherence closure. What string theory may still be universal over.The anti-totality constraint does not forbid universality; it forbids terminality. String theory can still be universal in the abductive/categorical sense if, for a specified class of quantum-gravitational regimes, it provides: • canonical generating principles for consistent models (e.g. extended objects such as D-branes and their worldvolume dynamics [43]), • nonperturbative completeness in certain asymptotic settings via holographic duality (e.g. AdS/CFT [44–46]), • and a duality web showing that apparently distinct descriptions factor through a common framework in different limits. In particular, AdS/CFT provides an existence proof of a complete quantum description of gravitational physics in AdS asymptotics, and it is precisely within this setting that the modern QEC/EWR/island structures have been most sharply formulated. Thus, conditionally (over AdS-like regimes), string theory plausibly functions as a universal framework in the sense that many consistent gravitational descriptions factor through string-theoretic constructions, often via holography. Why “universality” here is necessarily conditional. The claim “string theory is universal” would require quantifying over “all consistent quantum gravities,” which is itself a totality-style claim and is not something one should expect to be provable in a terminal, context-independent way (Section IX). Our framework therefore suggests a disciplined posture: • universality claims must be relative to a specified class of contexts (asymptotics, observables, code domains), • and should be understood as initial/core factorization properties rather than as global terminal closure. In this sense, “string theory as a universal framework” is an abductive hypothesis: it is a proposal that a broad and physically important class of consistent gravitational contexts factor through a canonical core, not a claim of terminal completeness. Duality and holography as evidence for non-terminal universality. Historically, the internal evolution of string theory itself provides evidence for this reframing: •dualities repeatedly dissolve the idea of a unique “fundamental” description, •holography replaces naive locality by contextual reconstruction, • and the modern islands/QES picture replaces fixed global decoding by variational decoder selection. Rather than interpreting these as failures to reach a terminal ToE, our framework interprets them as signs that the correct mathematical shape is being approached: a universal but non-terminal theory. Takeaway. The appropriate conceptual conclusion is therefore: String theory cannot be terminal in the sense of providing a single context-independent global decoder of all information, but it may still be universal in a conditional sense: as a canonical framework whose constructions and dualities provide universal cores for important classes of quantum-gravitational contexts (notably AdS holography). This is the mature target: universality without terminality. It dissolves false dichotomies in the sociology of quantum gravity (“ToE or failure”) and replaces them with the only shape consistent with diagonal constraints and coherence-first physics. 82 D. Why this helps islands The island/QES prescription has been one of the most productive technical advances in the modern black-hole information discussion. Nevertheless, even sympathetic readers sometimes treat islands as “magic” or as a clever semiclassical trick whose deeper meaning is unclear. The framework developed in this paper helps in a very specific way: it explains why islands are not an optional patch but a necessary decoder-selection repair once totality is recognized as the hidden over-demand. The apparent mystery islands create. In the naive narrative, one fixes a subsystem R (the radiation) and asks for the entropy S ( R ). The island formula (74) then seems to assert that an interior region I “counts as part of R ”. Without a guiding principle, this can feel like semantic fiat: why should an interior region be included in a quantity supposedly defined outside the black hole? Our diagnosis is that this sense of mystery arises because one silently assumes a terminal global semantics: •the subsystem decomposition is fixed once and for all, •the meaning of “interior operators” is fixed once and for all, •the decoding map from radiation to interior is fixed once and for all. But these are exactly the assumptions of terminal totality (Definition II.1) that diagonal logic forbids (Proposition III.5). Islands as the gravity realization of anti-totality. Once one accepts anti-totality and No-TotalHolography (Propositions III.5 and V.1), there is a single inevitable conclusion: No fixed global decoder can remain valid across all regimes in a maximally self-referential system. But after the Page time, insisting on a fixed no-island decoding context yields an ever-growing entropy inconsistent with unitary-compatible expectations [ 2 ] and triggers monogamy/AMPS tensions [ 3 ]. In our language, this is the failure mode of terminality: a fixed decoder is attempting to function as a total classifier across incompatible contexts. Islands then appear not as a trick but as the controlled alternative: they implement a regime-dependent decoding context selected by a variational principle. This is exactly Proposition VII.1. Decoder selection makes the “interior inclusion” inevitable. The key conceptual improvement is that it replaces the misleading question: “Why does an interior region become part of the radiation?” by the correct one: “Which decoding context is coherent in this regime, given that total decoding is forbidden?” Once this is recognized, the island term is no longer a semantic surprise: it is the statement that, in the post-Page regime, the only coherent decoding context is one in which the entanglement wedge of R includes an island. This is precisely what the minimization of generalized entropy selects [5]. The generalized entropy functional as a decoder-cost functional. The decoder-selection reinterpretation of Subsection VII B explains why the formula takes the form it does. A candidate island I is a candidate decoding context; the functional Sgen(R;I) = Area(∂I) 4GN +Sbulk(R∪I) balances: •the gravitational cost of extending the decoding context (area term), •the entanglement cost of refusing to include interior partners in the decode (bulk entropy term). Minimization then selects the context that repairs coherence with the least structural change. This is exactly how one should expect physics to respond to anti-totality: replace a forbidden terminal closure by a controlled, minimal, variationally selected contextual rule. 83 Why this removes the “ad hoc” worry. From this perspective, islands are not an added postulate. They are the concrete semiclassical implementation of a necessity: a maximally self-referential system cannot maintain a fixed global decoding semantics, so it must select decoding contexts. The replica derivations show how semiclassical gravity implements this necessity, while the diagonal/HCCB framework explains why the necessity exists in the first place. Compatibility with both global unitarity and HCCB-limited unitarity. A further benefit is conceptual robustness. The island prescription is often presented as enforcing global purity, hence global unitarity. Our framework shows that islands can be understood more generally: even if unitarity is only patchwise (HCCB-limited), islands still function as coherence repairs: they select the decoding context in which operational entropy bookkeeping and reconstruction remain self-consistent. Thus islands are not tied to a dogmatic global-unitarity axiom; they are tied to coherence closure. Takeaway. The main conceptual help is therefore: Islands are not magic; they are the necessary decoder-selection repair demanded by anti-totality in a maximally self-referential gravitational regime. This framing clarifies why islands have the precise variational structure they do, why the Page curve is a decoder transition, and why AMPS/monogamy is a diagnostic of terminality rather than of information loss. E. What this suggests for future “Einstein-like” physics One of the guiding analogies developed in our discussions is the Kant → Einstein pattern: a structural impossibility (no absolute spacetime knowledge) is followed by a physical reformulation that implements the constraint as an invariant principle (relativity and field equations). In this paper we have argued that black holes (and self-referential quantum systems more generally) reveal an analogous structural boundary for information: terminal total decoding is forbidden. The natural question is then: What would count as the “Einstein-like” physics that implements anti-totality as a dynamical, predictive principle? We suggest two complementary directions: (i) decoder-selection laws as variational principles, and (ii) coherence-defect dynamics as effective field equations in the HCCB framework. (I) Decoder-selection laws as variational principles The island/QES prescription already provides a concrete model of what an Einstein-like law could look like: not a terminal decoder, but a selection rule that chooses the correct decoding context in a given regime by extremizing a functional. In gravity, this functional is the generalized entropy, Sgen(R;I) = Area(∂I) 4GN +Sbulk(R∪I), and the law is the min–ext rule (74) selecting the coherent reconstruction wedge. The conceptual upgrade provided by this paper is to interpret such rules as instances of a general pattern: In the absence of terminal decoding, physics supplies a variational principle that selects the decoding context in which coherence closure is maintained. In other words, the law is not “decode everything,” but “select the context in which the decodable algebra is coherent.” A future Einstein-like development would be to identify analogous decoder-selection functionals beyond the island setting: for example, selection principles for: • which algebra should be treated as accessible in gauge/gravity localization problems (edge-mode choices), •which code subspace is stable under a given deformation, •which context is operationally meaningful for a given observer chain in measurement theory. These would generalize the island logic to a broader “coherence repair” principle applicable across quantum systems. 84 (II) Coherence-defect dynamics as effective field equations (HCCB direction) The second, deeper direction is to promote the coherence obstruction itself to a dynamical object. Section VIII framed HCCB as nontrivial higher coherence holonomy: a nonvanishing class [ ω ]obstructs global strictification (Subsection IX B). Operationally, such obstructions manifest as CP flows, history dependence, and sectorization (Subsection VIII E). This suggests a new kind of “Einstein equation” for information/coherence: not a field equation for a metric, but a field equation for the coherence defect. Coherence defect as a dynamical source. In differential geometry, curvature is not merely an obstruction; it is a dynamical object sourced by stress-energy. Analogously, in HCCB one may treat coherence curvature/holonomy as sourced by physical processes: measurement, coarse-graining, decoding transitions, horizon-induced observer splitting, or other mechanisms that internalize self-reference. An Einstein-like program would then ask: What dynamical laws govern the evolution of the coherence defect class [ ω ]and its operational shadow (CP flow, memory kernels, sector formation)? What such equations might look like. At a schematic level, one expects: • an evolution law for the effective channel family { Φ t} on accessible algebras, including memory kernels (non-Markovianity) and constraints ensuring complete positivity [37, 38]; • a constitutive relation linking those operational channels to coherence holonomy in the context bicategory (Subsection VIII D); • and, in regimes where Markovian closure is valid, an effective GKLS generator with parameters determined by coherence-defect data (Subsection VIII E). In this sense, the “field equation” would be an equation for the connection/holonomy data on context space, with CP evolution as its observable consequence. Why this is a genuine physics program. This direction is not merely interpretive. It suggests: •new invariants to compute (coherence-defect entropy, holonomy classes), •new kinds of conservation laws (coherence closure constraints), • and potentially new experimentally testable signatures (path dependence/holonomy in controlled simulators; Section IX E). In other words, it turns anti-totality from a constraint into a dynamical theory of how coherence is maintained when terminal closure is forbidden. Synthesis We therefore suggest the following “Einstein-like” research program: 1. Identify variational decoder-selection functionals that select coherent contexts (islands are the gravitational prototype). 2. Promote coherence-defect (holonomy) data to dynamical variables and derive effective equations governing their evolution and operational consequences (CP flows, memory kernels, sectorization). This program would make precise and predictive the core philosophical shift of this paper: physics is not a terminal decoder of reality, but a coherence-closed system of contextual encodings and transitions. XII. OUTLOOK AND OPEN PROBLEMS A. Quantitative measure of self-referentiality A central theme of this paper is that the decisive obstruction to terminal closure is not mystical but structural: self-reference triggers diagonal constraints, which in physics manifest as contextual decoding, islands, and HCCB-limited unitarity. So far we have treated “self-referentiality” qualitatively, via the 85 minimal trigger conditions (R)–(E)–(C) in Subsection III A. A natural next step—and an important open problem—is to make self-referentiality quantitative: to define an index that can be estimated in models and that correlates with the onset of decoder transitions (e.g. island formation) and with the strength of HCCB effects (holonomy, memory, CP flow). Goal of this subsection. We propose a research program for defining a self-reference index Σ(Sys;Ctx) that depends on a system and a specified family of contexts, and decomposes into three axes: R= representability, E= evaluation/self-application, C= closure under reflection. The index should be: •operationally meaningful (estimable from channel/algebra data), •context-sensitive (because self-reference is not absolute), •predictively useful (correlates with wedge/decoder transitions and coherence defects). Axis R: representability capacity Axis R measures how richly the system can internally represent predicates about itself. A minimal operational proxy is the expressive capacity of the internal code space or record space relative to the predicate family of interest. In finite-dimensional toy models, one may measure this by: •the dimension of a “name” register or description algebra, •the size of a family of internally representable predicates on the state space, •the effective rank of a representation map Name : Q0→ N. In algebraic settings, R could be formulated in terms of the size/structure of the accessible center, the space of superselection labels, or the complexity of the internal logical language supported by the probe algebra (e.g. the richness of subobjects/projections). A promising direction is to define R as the maximal size of a predicate family Q0 that admits an injective internal naming scheme consistent with the observable algebra constraints. Axis E: evaluation power and controllability Axis E measures whether the system can apply its internal descriptions as procedures: whether it can perform decoding/evaluation operations on internally stored codes. Operationally, Eshould quantify: •the controllability of the code degrees of freedom by internal dynamics, •the ability to implement recovery/decoding channels on the code, •the ability to couple the “name register” to the rest of the system as an interpreter. In black-hole settings, E corresponds to how much control an external agent has over Hawking radiation degrees of freedom (which is practically tiny for astrophysical black holes but conceptually meaningful). In toy models and simulators, Eis the practical implementability of decoding circuits. A potential quantitative proxy is the maximal fidelity of implementation of a family of evaluation channels Eval compatible with the system’s dynamics and locality constraints, or a bound on the set of CPTP maps implementable by internal operations. 86 Axis C: closure under reflection Axis C measures the extent to which the admissible query class is closed under reflection: whether the system can pose queries about the outcomes of its own decoding/evaluation. This is the true diagonal trigger. A quantitative Cshould capture: •whether the probe algebra can represent predicates about decoding maps, •whether the system can operationally access and condition on its own predicted answers, •whether meta-queries (“what does the decoder output?") can be encoded as ordinary queries. In categorical terms, C is related to the availability of internal “evaluation” structures and negation-like operations in the internal logic (Lawvere/Tarski type conditions). In operational terms, C is related to whether the agent can form closed loops of context transitions whose outcomes affect later choices. A composite self-reference index We therefore propose defining a composite index, for example: Σ(Sys;Ctx) := FΣR,ΣE,ΣC,(105) where Σ R, Σ E, Σ C are normalized measures on the three axes and F is a monotone aggregator (e.g. product, minimum, or weighted sum depending on the intended interpretation). The essential qualitative requirement is: Diagonal instability should become likely only when all three axes are simultaneously large: representability is rich, evaluation is implementable, and reflection closure is effective. Connection to wedge/decoder transitions The island/Page transition can be read as the point where a fixed decoding context becomes incoherent and the theory selects a new context by a variational principle. We suggest that Σshould correlate with this transition: • In early evaporation, the effective self-reference index relevant to the radiation decoder is low: the radiation does not yet support the necessary representability/evaluation/closure to force a decoder transition. • After Page time, the effective index increases: the radiation subsystem becomes large enough (representability), the decoding problem becomes meaningful (evaluation), and reflection closure is unavoidable in the self-referential setting (closure). • The transition in Σshould coincide with or predict the saddle crossing in the generalized entropy minimization. This provides a concrete research target: construct toy models (random circuits, holographic codes, JT-like models coupled to baths) where one can compute both a candidate Σand the decoder transition, and test quantitative correlation. Connection to HCCB strength HCCB predicts that higher coherence holonomy and CP path dependence become significant when the system is sufficiently self-referential and context-rich. Thus Σshould also correlate with: •the magnitude of holonomy defects (measured by deviation from strictification), •the degree of path dependence of effective channels (Equation (92)), •non-Markovianity measures (CP-divisibility failure) [38]. A successful quantitative program would therefore unify: diagonal instability thresholds, island decoder transitions, and HCCB coherence curvature as facets of one underlying quantitative self-reference index. 87 Takeaway. Defining and computing a self-reference index Σis a concrete open problem with clear payoffs: it would turn the present qualitative framework into a quantitative predictive tool, linking logical non-totality to physical transitions (Page/islands) and to coherence-defect dynamics (HCCB). B. Coherence-defect entropy A second concrete open problem is to define an intrinsic quantitative invariant that measures how much “gluing obstruction” is present when one tries to combine local/contextual descriptions into a globally strictified semantics. In the present framework, this obstruction is encoded by nontrivial higher coherence data (e.g. a holonomy/cohomology class [ ω ] 6 = 0 in Section IX) and manifests operationally as path dependence, sectorization, and CPTP dynamics between contexts (Section VIII). We propose that this obstruction should admit an entropy-like quantification: a coherence-defect entropy. Motivation: gluing costs show up as “extra” boundary/sector data. Across several domains discussed earlier, the failure of naive global strictification is repaired by introducing additional structure at boundaries or overlaps: • In gauge theories, the obstruction to naive factorization is repaired by edge modes or center data; entanglement depends on this extension. • In AQFT, local algebras are Type III; naive density-matrix factorization fails and one must work with relational (modular/relative-entropy) data [42]. • In evaporating black holes, islands/QES implement a decoder-context transition and the generalized entropy includes an explicit “area cost” that can be read as the price of a context change. In each case, the theory remains predictive, but only after paying a price in additional boundary/sector/- coherence information. This is precisely what an entropy-like invariant should capture. Conceptual definition: minimal cost to restore strict gluing Let Ctx be the context bicategory and let [ ω ]denote the coherence/holonomy class obstructing global strictification (Section IX B). A coherence-defect entropy Sdef should measure the minimal “extra structure” needed to trivialize the obstruction, i.e. to pass to an extended description in which [ ω ]becomes trivial and gluing becomes strict. One abstract way to express this is: Sdef := inf extensions ECost(E)s.t. [ω]becomes trivial in the extended system.(106) Here an “extension” could mean: •enlarging local algebras by adding center/edge degrees of freedom, •enlarging the code domain by adding an “edge” Hilbert factor, •enlarging the context bicategory by adding additional 2-morphisms that trivialize loops. The cost functional should be operationally meaningful (measurable from accessible physics), for example as a minimal entropy, mutual information, or log-dimension of an added sector space. Gauge-theory prototype: center/edge entropy as a defect measure Gauge theories provide a concrete prototype. A common way to obtain a factorized subsystem description is to enlarge the Hilbert space by including edge modes on the entangling surface. In such extended descriptions, the algebra has a nontrivial center, and the reduced state decomposes into a direct sum over superselection sectors labeled by boundary fluxes. The associated “center entropy” (or edge-mode entropy contribution) is a natural candidate for Sdef in that setting. In our language: The edge/center entropy is the quantitative price paid to restore coherent gluing (local subsystems) in the presence of gauge constraints. 88 Gravity prototype: the area term as coherence-defect entropy In the island/QES formula, the generalized entropy functional Sgen(R;I) = Area(∂I) 4GN +Sbulk(R∪I) already has the exact structure expected of a coherence-defect measure. In the decoder-selection interpretation (Section VII), choosing an island is choosing a decoding context; the area term penalizes enlarging the decoding context and can be interpreted as the “price” of a context repair. This suggests a concrete conjectural identification: Sgrav def (R)∼Area(∂I?) 4GN ,(107) where I? is the variationally selected island. On this view, the area term is not merely geometric; it is an entropy of coherence repair: the minimal cost needed to restore a coherent decoding context for the radiation. This interpretation aligns naturally with the idea that islands are not “magic inclusions” but a controlled, variationally minimal context transition. Operator-algebra/AQFT perspective: defect as non-factorizability cost In AQFT, Type III structure prevents naive density-matrix factorization on local algebras [ 42 ]. A coherence-defect entropy in this setting should quantify the minimal extension required to obtain a Type Ilike factorization (if possible) or, more realistically, quantify the amount by which naive factorization-based entropic notions fail and must be replaced by relational quantities. One possibility is to define Sdef via the minimal relative-entropy “gap” between two gluing prescriptions, or via the minimal information that must be added (a center) to achieve a chosen factorization property. This remains open and would connect naturally to modular theory and to relative-entropy control in holography. Connection to HCCB: defect entropy as curvature magnitude HCCB interprets global failure of strictification as a coherence-curvature/holonomy phenomenon (Subsection VIII D). A coherence-defect entropy should therefore be related to: •the size (or complexity) of the holonomy class [ω], •the magnitude of path dependence of effective channels Φγ(Equation (92)), •the degree of sectorization (how many inequivalent gluing sectors are present), •non-Markovian memory strength (as an operational shadow of holonomy). In this sense, Sdef would function as an “order parameter” for HCCB: zero in strictifiable regimes, nonzero when coherence defects are present. Concrete open problems. A fully developed coherence-defect entropy program would need to answer: 1. Define a cost functional Cost ( E )in (106) that is operationally meaningful and model-independent. 2. Show that Sdef reduces to known edge/center entropy terms in gauge theory and to the area term in gravitational island transitions. 3. Relate Sdef to measurable signatures: channel path dependence, non-Markovianity, or relativeentropy inequalities. 4. Establish monotonicity or extremality principles (“defect entropy is minimized by the physical decoder-selection law”). 89 Takeaway. A coherence-defect entropy would provide exactly what is currently missing for a quantitative HCCB research agenda: a scalar (or family of scalars) measuring “how far” a system is from global strictification, and how much contextual structure must be retained to preserve coherence closure. Its existence would unify the area term in islands, edge-mode entropy in gauge theories, and non-factorizability in AQFT under a single quantitative principle. C. Formal HCCB mathematics The HCCB mechanism has been presented in this paper at a level sufficient to connect diagonal anti-totality to operational CP dynamics, sectorization, and decoder transitions. A major open problem is to place HCCB on a fully formal mathematical footing: to specify precisely which cohomological (or higher-categorical) class controls strictification, how it is computed in concrete models, and how it relates to familiar obstruction phenomena in quantum field theory such as anomalies and descent. Goal. We seek a precise mathematical package of the following form: 1. a well-defined higher category (or stack) of contexts Ctx, 2. a functorial assignment of local quantum data (Hilbert spaces/algebras and unitary dynamics) to objects of Ctx, 3. a coherence datum (associators, 2-morphisms, higher homotopies) defining gluing, 4. an obstruction class [Ω] such that: [Ω] = 0 ⇐⇒ global strictification exists (global unitary semantics), and [Ω] 6= 0 ⇐⇒ HCCB-limited unitarity (nontrivial holonomy/defect). Which cohomological class controls strictification? In Section IX B we presented a groupoid-level model where the obstruction to strict functorial transport is a U (1)-valued 2-cocycle class [ ω ] ∈H2 ( C, U (1)). This captures the simplest “projective” obstruction. However, for a genuine bicategory (or (∞,1)-category) of contexts, the obstruction can shift degree: • In a bicategory, the associator coherence typically defines a 3-cocycle-type obstruction to strictification (analogous to how monoidal categories are classified by 3-cocycles in some settings). • In ( ∞, 1)-categorical settings, obstructions can live in higher cohomology groups or as classes in the cohomology of the nerve of the context ∞-groupoid. Thus a central task is to identify the correct mathematical home of the obstruction class. Possible candidates include: 1. groupoid cohomology of the context space with coefficients in a central unitary group, 2. cohomology of a classifying space BCtx, 3. obstruction theory for strictification of pseudofunctors into Hilb or vNAlg, 4. higher stacky cohomology classes controlling descent. A minimal deliverable for a formal HCCB theory is a theorem of the form: A global strictification exists if and only if a specific obstruction class vanishes. This is the coherence analogue of familiar topological results (global section exists iff the bundle class is trivial). 96 2. Gödel diagonal lemma (one page) Gödel’s diagonal lemma is the engine behind self-referential sentences in arithmetic [ 8 ]. We state it in the standard form used in incompleteness proofs. Set-up. Let Tbe a sufficiently expressive, effectively axiomatized theory of arithmetic, with a coding p·qof formulas/sentences by natural numbers (Gödel numbering). Lemma A.2 (Diagonal lemma) . For any formula F ( x )of Twith one free variable x , there exists a sentence Gsuch that T`G↔F(pGq).(A1) Interpretation. The lemma says: any property F of codes can be “self-applied” to yield a sentence G that asserts Fof its own code. This is the formal realization of (R)–(E)–(C) in the main text. From diagonal lemma to incompleteness (sketch). Choose F ( x ) ≡ ¬ProvT ( x )where ProvT is the provability predicate. Then Gsatisfies G↔ ¬ProvT(pGq), i.e. “ G is not provable.” Consistency of Timplies G is not provable, hence Tis incomplete (details in standard texts). Physics moral. A total internal truth predicate would be a terminal classifier of semantic facts about the system. Gödel/Tarski show such a terminal closure is forbidden once self-reference is present. 3. Turing halting diagonal machine Turing’s halting theorem is the computational form of diagonal non-totality [9]. Theorem A.3 (Halting problem) . There is no total algorithm Halt ( M, x ) ∈ { 0 , 1 } that decides for every program Mand input xwhether Mhalts on x. Proof. Assume Halt exists. Define the diagonal machine D that on input M : (i) computes Halt ( M, M ); (ii) if it is 1, loops forever; if it is 0, halts. Now run D on itself. If Halt ( D, D )=1, then D ( D )does not halt; contradiction. If Halt(D, D)=0, then D(D)halts; contradiction. Physics moral. “Hard to decode” (complexity) is different from “no total decoder.” Turing shows the latter is a structural impossibility once self-reference is allowed. 4. Lawvere fixed point theorem (diagram) Lawvere provides a categorical unification of the diagonal mechanism in cartesian closed categories [ 10 ]. Set-up. Let C be a cartesian closed category. For objects A, B there is an exponential BA and evaluation ev : BA×A→B. Weak point-surjectivity. A morphism φ : A→BA is weakly point-surjective if for every g : A→B there exists a point a: 1 →Asuch that g= ev ◦(φ◦a×idA). Theorem A.4 (Lawvere fixed point) . If φ : A→BA is weakly point-surjective, then every h : B→B has a fixed point. Proof sketch with the diagonal diagram. Define g:A→Bby g:= h◦ev ◦(φ×idA)◦∆, where ∆ : A→A×A is the diagonal. By weak point-surjectivity choose a : 1 →A such that g= ev ◦(φ◦a×idA). Then the point b:= ev ◦(φ◦a×a):1→B satisfies h◦b=b. 97 Diagrammatic intuition. The diagonalization is exactly the insertion of ∆and evaluation: A A ×A BA×A B B B ∆ g φ×id ev h(A2) If B admits a negation-like endomorphism with no fixed point, Theorem A.4 implies no such weak point-surjective φcan exist; this is the categorical version of diagonal inconsistency. Physics moral. Lawvere makes explicit what is often hidden in physics arguments: “total representability + evaluation” forces fixed points and hence forbids terminal classifiers in the presence of negation/reflection. This is the categorical spine behind our No-Total-Holography statement. Appendix B: Categorical formalism: contexts, functors, and coherence This appendix collects the categorical formalism implicitly used throughout the manuscript and makes it explicit in one place. Its purpose is not to introduce new results, but to fix notation, terminology, and logical dependencies so that references in the main text (categorical holography, diagonal logic, HCCB, strictification, and undecidability) are unambiguous and technically precise. 1. Contexts as a category or bicategory We model physical descriptions as contexts. A context represents a choice of: •accessible algebra or subsystem, •code subspace or semiclassical background, •decoding/reconstruction prescription, •observer or operational standpoint. Context category. At minimum, contexts form a category (often a groupoid) C(contexts) whose: •objects c∈Ob(C)are contexts, • morphisms f : c→c0 are admissible context changes (restriction, extension, coarse-graining, decoding change). In many physical settings, C is naturally a groupoid (changes are invertible up to equivalence), but non-invertible morphisms are allowed when modeling irreversible coarse-graining. Higher structure. When multiple context changes compose only up to equivalence, C should be treated as a bicategory (or weak 2-category): •1-morphisms: context transitions, •2-morphisms: coherence data relating different transition paths. This higher structure is essential for HCCB and for coherence holonomy. 2. Quantum semantic charts To each context cwe associate a quantum semantic chart: Q(c)∈ {Hilbert spaces, operator algebras, or pairs} For example: •a Hilbert space Hc, •or a von Neumann algebra Acacting on Hc, •possibly together with a specified code domain or state family. 98 Local dynamics. Within each context, dynamics is assumed to be linear and unitary (or *-automorphic): Uc(t) : Hc→ Hc This is the notion of patchwise unitarity used throughout the paper. 3. Transport functors and naturality Context changes induce transport of quantum data. Transport. For each morphism f:c→c0in C, we associate a map Q(f) : Q(c)−→ Q(c0) which is: •unitary if fis an equivalence, •CPTP (channel) if frepresents restriction or coarse-graining. Naturality. Naturality means that physical constructions commute with admissible context changes. Categorically, Qis a (pseudo)functor Q:C → Hilb,vNAlg,or CPTP. This is the precise meaning of “robust under deformation” used in the main text. 4. Coherence data and projective composition In general, transport is not strictly functorial. Instead, for composable morphisms f, g one may have Q(g)◦ Q(f) = ω(g, f)Q(g◦f), where ω(g, f)is a central unitary (often U(1)-valued). Coherence cocycle. Associativity implies that ωsatisfies a cocycle condition, defining a class [ω]∈H2(C, U(1)) (or higher-degree analogues in bicategorical settings). This class measures the obstruction to strictification. 5. Strictification and global unitarity Strictification. Strictification means replacing the pseudofunctor Q by an equivalent strict functor for which all coherence factors are trivial. Physical meaning. Strictification is equivalent to the existence of: •a single global Hilbert space H∗, •embeddings Jc:Hc→ H∗, •such that all context changes are realized as strict unitary conjugations. This is precisely what we call global unitarity in the terminal sense. If [ω]6= 0, strictification is impossible: one has only HCCB-limited unitarity. 6. Relation to diagonal logic The categorical formalism here is not independent of diagonal arguments. Diagonal logic forbids total internal classifiers; categorically, this forbids terminal objects or global truth functors. HCCB provides the geometric/physical realization of this logical constraint: diagonal non-totality appears as nontrivial coherence holonomy obstructing strictification. 99 7. Role in the main text This appendix underlies: •Section V (encoding functors, No-Total-Holography), •Section VIII (coherence holonomy and patchwise unitarity), •Section IX (global unitarity vs non-certifiability), •Appendices B and C (strictification and undecidability). Readers primarily interested in physics may safely treat this appendix as a reference dictionary. Readers interested in formalization may view it as the minimal categorical scaffold on which the entire argument is built. Appendix C: Formal criterion for global unitarity over a context groupoid This appendix provides a self-contained formalization of the strictification criterion used in Section IX. The goal is to make mathematically explicit what is meant by: “Global unitarity is the existence of a context-independent strictification/trivialization of patchwise unitary semantics over the context groupoid.” We work in a groupoid-level model because it is the simplest setting in which the obstruction class [ ω ] and its vanishing criterion can be written down explicitly. The higher-categorical refinement (bicategories and 3-cocycles) follows the same pattern but requires heavier machinery; the groupoid model already captures the essential physics of holonomy and path-independence. 1. Patchwise unitary semantics: data and functoriality Let C be a (small) groupoid of contexts. Objects are contexts c∈Ob ( C ), and morphisms f : c→c0 are invertible context changes (re-descriptions) between contexts. Apatchwise unitary semantics over Cconsists of: 1. a Hilbert space Hcfor each context c; 2. for each morphism f:c→c0, a unitary transport map Uf:Hc−→ Hc0;(C1) 3. (optionally) for each context c , a unitary time evolution Uc ( t )acting on Hc (patchwise unitarity) as in Subsection VIII C. Strict functoriality versus projective functoriality. If the assignment c7→ Hc and f7→ Uf were strictly functorial, then Ug◦f=UgUffor all composable f, g. (C2) In many physically relevant situations (and in the HCCB picture), strict functoriality is too strong; one expects only projective functoriality: UgUf=ω(g, f)Ug◦f, ω(g, f)∈U(1),(C3) where ω(g, f)is a coherence factor measuring the failure of strict composition. Remark (central extensions). One may replace U (1) by central unitaries in a von Neumann algebra or by a more general coefficient group. The U (1) case already illustrates the strictification obstruction with minimal overhead. 100 2. The coherence cocycle and its cohomology class Cocycle condition from associativity. Associativity of composition implies that ω must satisfy the 2-cocycle condition: for composable morphisms f, g, h, ω(h, g ◦f)ω(g, f) = ω(h◦g, f)ω(h, g).(C4) Thus ωis a U(1)-valued 2-cocycle on the groupoid C. Gauge transformation (rephasing) and coboundaries. A change of phase convention for the transport maps corresponds to choosing a 1-cochain λ(f)∈U(1) and defining Uf7→ U0 f:= λ(f)Uf.(C5) Under this change, the cocycle transforms as ω(g, f)7→ ω0(g, f) := λ(g)λ(f) λ(g◦f)ω(g, f),(C6) i.e. by multiplication with a coboundary. Therefore ωdetermines a cohomology class [ω]∈H2(C, U(1)),(C7) which is invariant under rephasing. Standard references for group/groupoid cohomology and cocycles include [39]. 3. Strictification criterion as a proposition We now state the formal strictification criterion used in the main text. Proposition C.1 (Global unitarity ⇐⇒ vanishing of [ ω ]) . Let C be a context groupoid and let { ( Hc, Uf ) } be a projective unitary transport system satisfying (C3) with cocycle ω and class [ ω ] ∈H2 ( C, U (1)). Then the following are equivalent: (i) [ω]=0. (ii) There exists a rephasing {U0 f}as in (C5) such that the transport becomes strictly functorial: U0 g◦f=U0 gU0 ffor all composable f, g. (iii) There exists a single Hilbert space H∗ and unitary embeddings Jc : Hc→ H∗ such that for every f:c→c0, U0 f=J−1 c0Jc(on the image of Jc),(C8) so that all contexts are realized as compatible charts of one global unitary semantics. Proof sketch. (i) ⇒ (ii): If [ ω ]=0, then ω = δλ−1 for some 1-cochain λ , and (C6) yields ω0≡ 1. Thus U0 gU0 f=U0 g◦f. (ii) ⇒ (iii): Choose a reference context c0 and set H∗ = Hc0 . For each c , choose a morphism p : c→c0 and define Jc:= U0 p. Strict functoriality implies independence of choices and yields (C8). (iii) ⇒ (i): If a global trivialization exists, then transport is strict by construction, so ω≡ 1and [ω]=0. Physics interpretation. Proposition C.1 makes precise what is meant by “global unitarity” in the sense of this paper: Global unitarity is not a statement about a single reduced subsystem’s dynamics; it is the statement that the entire patchwise/contextual transport system strictifies into one contextindependent unitary semantics. If [ ω ] 6 = 0, then local/patchwise unitarity can still hold in each context, but no global strictification exists; this is the precise mathematical meaning of HCCB-limited unitarity. 101 Higher-categorical refinement. In a full bicategory of contexts, the obstruction can appear at higher degree (e.g. a 3-cocycle associated to associators), but the conceptual structure is identical: global strictification exists iff a higher coherence class vanishes. The groupoid case treated here is the minimal model sufficient to define and discuss globality, holonomy, and strictification in a form accessible to physicists. Appendix D: Undecidability schema This appendix expands Subsection IX D by giving a more explicit outline of how undecidability can enter the strictification problem. We emphasize the scope carefully: • The goal is a schema: a standard reduction pattern showing that, in sufficiently expressive families of finitely presented context systems, deciding [ω]=0can be algorithmically undecidable. •We do not claim that every physically realized context family lies in such an expressive class. • We do not attempt a full formal reduction with all technical details; we provide the structure of the reduction and point to the classical undecidability inputs. This level of detail is appropriate for a physics paper: it makes explicit what is meant by “undecidability” here and why it is not a mere rhetorical flourish. 1. The decision problem Fix a class of finitely presented context groupoids C together with a finitely specified U (1)-valued 2-cocycle ω(Appendix C). Consider the decision problem: Input: a finite presentation of a context groupoid C and a finite description of a cocycle ω on C. Question: is [ω] = 0 in H2(C, U(1))? Output: YES if [ω] = 0 (global strictification exists), NO otherwise. In the language of the main text, this is the decision problem “does a context-independent global unitary semantics exist?” Finite presentation model. A concrete model of input data is: •a finite presentation of a groupoid (or group) by generators and relations, •a finite table or formula specifying ωon generators and extending by the cocycle condition. (There are multiple equivalent formalizations; the choice does not affect the conceptual reduction pattern.) 2. Classical undecidability inputs The standard route to undecidability in algebraic decision problems is to use the fact that finite presentations can encode computation. Two classical results serve as canonical inputs: •Novikov’s theorem: the word problem is undecidable in general for finitely presented groups [40]. • Boone’s theorem: further establishes the undecidability of the word problem and related decision problems [41]. A third, conceptually simpler input is Turing’s halting problem itself [9]. All three instantiate the same principle: sufficiently expressive finite presentations can simulate universal computation. 102 3. Reduction outline: from computation to cocycle triviality We outline a reduction of the following form. Given a Turing machine M (with a fixed input convention), construct a finitely presented context groupoid CMand a cocycle ωMsuch that: Mhalts ⇐⇒ [ωM]=0.(D1) If such a construction exists, then any algorithm deciding whether [ ω ] = 0 would decide the halting problem, which is impossible [9]. Step 1: encode computation into a finite presentation The first step is to encode the computation of M into a finitely presented algebraic object. There are multiple known routes: • encode the computation into a finitely presented group GM such that a word wM is trivial in GM iff Mhalts (a standard construction in the theory of undecidability of group problems); •encode the computation into a finitely presented semigroup/monoid with an analogous property; •encode the computation directly into a groupoid presentation if groupoids are preferred. We do not reproduce the full construction here; the existence of such encodings is the content behind [40, 41]. Step 2: convert the encoded problem into a cocycle triviality condition The strictification problem for a cocycle is the existence of a 1-cochain λsuch that ω=δλ, (D2) i.e. a solution of the coherence equation that removes projective phases. Thus, to reduce a word/halting problem to [ ω ] = 0, one must construct ωM so that solving (D2) is equivalent to solving the encoded computation constraint. A conceptual route is: • choose CM so that certain compositional loops correspond to computational steps or to acceptance/halting conditions, • define ωM to be nontrivial precisely on those loops that represent “non-halting” behavior, and trivial otherwise, •arrange that a global trivializing cochain λexists iff the computation halts. Intuitively, ωM acts like a “coherence obstruction flag” attached to the computation: halting corresponds to the obstruction being removable by a consistent global gauge choice. Step 3: conclude undecidability If the mapping M7→ ( CM, ωM )can be made effective (computable), then (D1) implies that deciding [ ω ]=0is at least as hard as the halting problem. Therefore, no algorithm can decide [ ω ]=0for all inputs in the corresponding family. 4. Physics interpretation and scope The undecidability schema should be interpreted with the same care as all undecidability results in physics: 103 •It does not say “global unitarity is undecidable in our universe.” • It says: if one allows sufficiently expressive families of context systems, then the strictification predicate [ω] = 0 can be undecidable in those families. • This supports (but does not replace) the main epistemic claim of the paper: global strictification is not something one should treat as universally certifiable from within a self-referential system. In the present program, this provides a mathematically sharp sense in which “global unitarity” is not an operational axiom but a contingent global property, potentially inaccessible even in principle for expressive context families. Takeaway. The undecidability schema reinforces the coherence-first stance: physics should be organized around contextual coherence closure and decoder-selection laws, not around terminal global closure predicates whose decision may be non-algorithmic in general families. Appendix E: Islands as decoder selection in operator-algebra QEC language This appendix reformulates the “islands as decoder-selection” interpretation of Section VII in the language of operator-algebra quantum error correction (OAQEC). The goal is to make the link between: •entanglement wedge reconstruction as algebraic recoverability, •islands/QES as a selection of the recoverable algebra (via a variational rule), •and relative entropy equalities as the sharp information-theoretic criterion for recoverability, as explicit as possible. We emphasize that the appendix does not introduce new physics; it provides a precise operator-algebra translation of the main text. 1. Set-up: code subspace, regions, and algebras Let Hphys be the boundary/radiation Hilbert space and Hcode ⊆ Hphys a code subspace with encoding isometry V : Hcode → Hphys (Subsection VI A). Let R be a boundary region (or radiation subsystem) and write Hphys ∼ =HR⊗ H ¯ R. The “noise” channel is the erasure of ¯ R: NR(ρ) = Tr ¯ R(ρ). We consider a bulk operator algebra Abulk represented on the code subspace. The central question is: for which algebras Abulk does there exist a representation on Rthat reproduces all code matrix elements? 2. Recovery map conditions (OAQEC) Reconstructibility of an algebra. An algebra Abulk is said to be reconstructible from R on the code if for each O∈ Abulk there exists an operator e OR∈ B(HR)such that V†(e OR⊗1¯ R)V=Oon Hcode.(E1) This is exactly Equation (64) in the main text. Recovery channel formulation. Equivalently, one can require the existence of a recovery channel RR:S(HR)→ S(Hcode)such that for all code states ρ, RRNR(V ρV †)=ρ(on the protected code domain). (E2) State recovery implies operator recovery for the full matrix algebra on the code; OAQEC is more general because it corrects a subalgebra rather than the full algebra. 104 Correctability condition (commutant form). OAQEC admits a characterization in terms of the commutant of the protected algebra and the noise channel; a standard unified formulation is given in [ 31 ]. The key structural point is that correctability depends on: •the code domain (via V), •the noise model (here, erasure of ¯ R), •and the algebra one wishes to protect. This is contextuality in precisely the sense required by anti-totality. 3. Relative entropy equalities as recoverability criteria A particularly sharp and physically meaningful recoverability criterion is formulated in terms of relative entropy. Recall that relative entropy is monotone under channels: D(ρkσ)≥D(Φ(ρ)kΦ(σ)). Saturation of this inequality signals recoverability. Boundary–bulk relative entropy equality. In holography, a central result is that, in appropriate codesubspace regimes, boundary relative entropy on R equals bulk relative entropy in the entanglement wedge. At a schematic level: D(ρRkσR) = DρEW (R) σEW (R),(E3) for code states ρ, σ . Equation (E3) is an information-theoretic expression of entanglement wedge reconstruction: it says the distinguishability relevant to bulk operators in EW ( R )is exactly captured by the boundary region R. Petz recovery as canonical decoder (contextual universality). If relative entropy is saturated for a channel and a suitable family of states, then the Petz map provides a canonical recovery channel that reverses the coarse-graining on that family [ 17 ]. Thus relative entropy equality is not merely a diagnostic; it provides a canonical candidate decoder in the context of interest. This illustrates the paper’s main theme at the level of information theory: universality (canonical recovery) within a context, without terminal global decoding. 4. Islands as selection of the recoverable algebra We now connect the OAQEC viewpoint to islands. Island choice determines the wedge and hence the algebra. In the main text, a candidate island I defines a candidate entanglement wedge for R , denoted EWI ( R ), and hence a candidate bulk algebra Abulk(EWI(R)). Thus the choice of Iis the choice of which algebra is to be treated as reconstructible. Variational rule selects the context where relative-entropy reconstruction is coherent. The island/QES prescription selects I? by minimizing the generalized entropy. Operationally, this selects the wedge in which bulk entropy bookkeeping and relative-entropy control are consistent with semiclassical gravity. In this OAQEC language, islands implement: a variational selection of the recoverable algebra for the fixed probe subsystem R. Instead of demanding that R reconstruct the same algebra for all regimes (a terminality demand), the theory selects the algebra that is correctable in that regime. Connection to Page transition. The Page transition is the point at which the recoverable algebra for R changes: before the transition, the no-island wedge yields the correct relative-entropy bookkeeping; after the transition, a larger wedge including I? is required. Thus the Page curve is a decoder transition in the strongest algebraic sense: the protected algebra changes by variational selection. 105 5. Takeaway The OAQEC formulation makes the main text’s interpretation maximally explicit: •Bulk reconstruction is algebraic recoverability relative to a code and region. •Relative entropy equality is the sharp information-theoretic signature of recoverability. •The Petz map provides a canonical recovery in contexts where data processing saturates. • Islands/QES should be read as selecting which wedge/algebra is correctable from R in a given regime, thereby implementing a variational decoder-selection law. This formulation clarifies why islands are not an ad hoc semantic move: they are the semiclassical implementation of contextual recovery demanded by anti-totality. [1] S. W. Hawking, “Breakdown of Predictability in Gravitational Collapse,” Phys. Rev. D 14 (1976) 2460–2473. doi:10.1103/PhysRevD.14.2460. [2] D. N. Page, “Average Entropy of a Subsystem,” Phys. Rev. Lett. 71 (1993) 1291–1294. doi:10.1103/PhysRevLett.71.1291. arXiv:gr-qc/9305007. [3] A. Almheiri, D. Marolf, J. Polchinski and J. Sully, “Black Holes: Complementarity or Firewalls?,” JHEP 02 (2013) 062. doi:10.1007/JHEP02(2013)062. arXiv:1207.3123 [hep-th]. [4] S. Ryu and T. Takayanagi, “Holographic Derivation of Entanglement Entropy from AdS/CFT,” Phys. Rev. Lett. 96 (2006) 181602. doi:10.1103/PhysRevLett.96.181602. arXiv:hep-th/0603001. [5] A. Almheiri, R. Mahajan, J. Maldacena and Y. Zhao, “The Page curve of Hawking radiation from semiclassical geometry,” arXiv:1908.10996 [hep-th]. [6] C. E. Shannon, “A Mathematical Theory of Communication,” Bell System Technical Journal 27 (1948) 379–423, 623–656. [7] M. M. Wilde, Quantum Information Theory, 2nd ed., Cambridge University Press (2017). [8] K. Gödel, “Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I,” Monatshefte für Mathematik und Physik 38 (1931) 173–198. [9] A. M. Turing, “On Computable Numbers, with an Application to the Entscheidungsproblem,” Proc. Lond. Math. Soc. 2 42 (1936) 230–265; erratum 2 43 (1937) 544–546. doi:10.1112/plms/s2-42.1.230. [10] F. W. Lawvere, “Diagonal arguments and cartesian closed categories,” Theory Appl. Categ. 15 (2006) 1–13; reprint of Lecture Notes in Mathematics 92 (1969) 134–145. [11] A. T. Patrascu, “Higher-order categorical coherence breakdown: a geometric framework for nonlinear quantum mechanics and its applications to strongly correlated electron systems,” Eur. Phys. J. B 98 (2025) 210. doi:10.1140/epjb/s10051-025-01062-6. [12] G. Cantor, “Über eine elementare Frage der Mannigfaltigkeitslehre,” Jahresbericht der Deutschen Mathematiker-Vereinigung 1(1891) 75–78. [13] A. Almheiri, X. Dong and D. Harlow, “Bulk Locality and Quantum Error Correction in AdS/CFT,” JHEP 04 (2015) 163. doi:10.1007/JHEP04(2015)163. arXiv:1411.7041 [hep-th]. [14] X. Dong, D. Harlow and A. C. Wall, “Reconstruction of Bulk Operators within the Entanglement Wedge in Gauge-Gravity Duality,” Phys. Rev. Lett. 117 (2016) 021601. doi:10.1103/PhysRevLett.117.021601. arXiv:1601.05416 [hep-th]. [15] G. Penington, “Entanglement Wedge Reconstruction and the Information Paradox,” arXiv:1905.08255 [hep-th]. [16] H. Umegaki, “Conditional expectation in an operator algebra. IV. Entropy and information,” Kodai Mathematical Seminar Reports 14 (1962) 59–85. [17] D. Petz, “Sufficient subalgebras and the relative entropy of states of a von Neumann algebra,” Commun. Math. Phys. 105 (1986) 123–131. doi:10.1007/BF01212345. [18] G. Birkhoff and J. von Neumann, “The logic of quantum mechanics,” Annals of Mathematics 37 (1936) 823–843. [19] S. Kochen and E. P. Specker, “The problem of hidden variables in quantum mechanics,” J. Math. Mech. 17 (1967) 59–87. [20] R. B. Griffiths, “Consistent Histories and the Interpretation of Quantum Mechanics,” J. Stat. Phys. 36 (1984) 219–272. doi:10.1007/BF01015734. [21] M. Gell-Mann and J. B. Hartle, “Quantum mechanics in the light of quantum cosmology,” in W. Zurek (ed.), Complexity, Entropy and the Physics of Information, Addison-Wesley (1990); arXiv:gr-qc/9404013. [22] W. F. Stinespring, “Positive Functions on C∗-Algebras,” Proc. Amer. Math. Soc. 6(1955) 211–216. [23] K. Kraus, States, Effects, and Operations: Fundamental Notions of Quantum Theory, Lecture Notes in Physics 190, Springer (1983). [24] M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, Cambridge University Press (2000).