Attempt: Merge Program A (geometry-from-algebra) and Program B (quantum-framework-final)

Status: Attemptive — substantive partial merge at one sublayer; thin methodological bridge at the deeper sublayer; NOT empty re-description (medium confidence) Last updated: 2026-06-08 Iteration: 3

This is an iteration-3 merge attempt. It takes two iteration-2 programs — Program A (geometry-from-algebra: modular / crossed-product, KMS thermality, area law, RT/generalized entropy) and Program B (operational / GPT reconstruction of QM plus the BMV non-classicality gate) — and asks whether a single, non-circular operational/algebraic substrate yields both the quantum framework and emergent causal/thermal/geometric structure. The discipline is the same as the iteration-2 synthesis (2026-06-08-unified-program-synthesis.md): resist manufactured unity, report the honest coherence verdict, and locate the obstruction precisely rather than papering over it.

---

Goal

Decide whether a GPT reconstruction (no-signaling + local tomography + purification + continuous reversibility $\Rightarrow$ finite-dimensional QM) upgraded to the infinite-dimensional / relativistic / Type $\mathrm{III}$ algebraic setting (Tomita–Takesaki modular flow $\Rightarrow$ KMS / thermal time and emergent causal order) can ground both pillars without circularity, and confront the agent-vs-no-agent (cosmological, closed-universe) category clash.

Three verdict buckets are admissible: substantive merge, thin methodological bridge, or empty re-description. The iteration-2 verdict ([REG-SYNTH-5], see ../HYPOTHESES.md) was "thin / methodological: shared tools (no-signaling / GHJW) + BMV as a logical gate." This note tests whether two 2025–2026 developments — an infinite-dimensional Born-form rigidity theorem and a covariant internal-observer crossed product — upgrade that verdict.

---

Setup (the two pillars and the candidate bridge)

Pillar B — operational / GPT core

A GPT assigns to each system a state space $\Omega$ (a compact convex subset of an ordered vector space $V$) and effects in the dual cone $V^*$, with bilinear probability rule $p = \langle e, \omega\rangle$. Composition is fixed by local tomography: $\dim V_{AB} = \dim V_A \cdot \dim V_B,$ which forces the tensor product and (with the remaining axioms) the complex field. The finite-dimensional reconstructions are theorems ESTABLISHED:

• Hardy 2001 (quant-ph/0101012): from $K = N^r$ plus a continuous reversible transformation between any two pure states.
• Chiribella–D'Ariano–Perinotti 2011 (PRA 84, 012311): from causality + purification + local tomography + perfect distinguishability + ideal compression + a single-system atomicity/reversibility postulate.
• Masanes–Müller 2011 (NJP 13, 063001): from analogous operational requirements.

Output: finite-dimensional QM, states $\rho \succeq 0$, $\mathrm{Tr}\,\rho = 1$, Born form $p = \mathrm{Tr}(\rho E)$, tensor-product composition.

Pillar A — algebraic / modular core

A local relativistic QFT subregion carries a Type $\mathrm{III}_1$ von Neumann factor $\mathcal{M}$ (Connes–Haag–Araki) ESTABLISHED: no trace, no minimal projections, no factorization $\mathcal{H} \ne \mathcal{H}_{\mathcal{M}} \otimes \mathcal{H}_{\mathcal{M}'}$, no density matrix. The vacuum $\lvert\Omega\rangle$ is cyclic and separating (Reeh–Schlieder), so Tomita–Takesaki applies: the closure of $S_0\, a\lvert\Omega\rangle = a^\dagger \lvert\Omega\rangle$ has polar decomposition $S = J\Delta^{1/2}$, giving the modular operator $\Delta$ and modular automorphism $\sigma_t(a) = \Delta^{it}\, a\, \Delta^{-it}, \qquad \sigma_t(\mathcal{M}) = \mathcal{M}.$ The state $\omega(\cdot) = \langle\Omega\rvert\,\cdot\,\lvert\Omega\rangle$ is KMS at inverse temperature $\beta = 1$ w.r.t. $\sigma_t$: $\omega\!\left(a\,\sigma_{i}(b)\right) = \omega(ba)$ ESTABLISHED. This is the "thermal time" / emergent-thermality structure (Connes–Rovelli, CQG 11, 2899, 1994). By Bisognano–Wichmann, for the Rindler wedge $\sigma_t$ is the boost — i.e. modular flow is geometric — but only because $\lvert\Omega\rangle$ is the symmetric vacuum on a fixed spacetime ESTABLISHED; geometricity holds only for special (algebra, state) pairs.

Candidate bridge — crossed product / internal observer (CLPW)

Adjoin an "observer" = a clock with Hamiltonian $q$ (bounded below) and form the crossed product $\hat{\mathcal{M}} = \mathcal{M} \rtimes_\sigma \mathbb{R},$ generated by $\mathcal{M}$ and the modular flow implemented relationally. Imposing the gravitational constraint (invariance under the global modular/boost flow — a Page–Wootters relational-time gauge condition) and a lower bound on the clock spectrum converts $\mathrm{III}_1 \to \mathrm{II}_\infty$ (or $\mathrm{II}_1$ for the dS static patch), restoring a semifinite trace $\mathrm{Tr}$, density matrices $\hat\rho$, a finite von Neumann entropy $S(\hat\rho) = -\mathrm{Tr}(\hat\rho \log \hat\rho)$, and $S(\hat\rho) = \frac{\langle A\rangle}{4G} + S_{\mathrm{out}} + \text{const} = S_{\mathrm{gen}}$ (Witten 2022, arXiv:2112.12828; CLPW 2022, arXiv:2206.10780; Chandrasekaran–Penington–Witten 2022, arXiv:2209.10454) ESTABLISHED given a fixed background with the relevant Killing symmetry.

Referee-corrected mechanism (load-bearing, ESTABLISHED). The Type II semifinite trace exists for any Type $\mathrm{III}_1$ factor $\mathcal{M}$ and any faithful normal state, purely algebraically (Takesaki duality / Connes–Takesaki flow of weights) — with no boundedness-below and no semiclassical input; the modular flow being outer is the generic $\mathrm{III}_1$ situation, and the crossed product is precisely what manufactures the trace from it. Boundedness-below of the clock and the semiclassical window are needed only for the physical reading — finite generalized entropy $S_{\mathrm{gen}}$, observer energy $q \ge 0$, and matching $\mathrm{Tr}$ to $A/4G + S_{\mathrm{out}}$. This is the iteration-2 ledger correction [A-CP-KILLING] (see ../ASSUMPTIONS_LEDGER.md). The naive phrasing "the clock is imposed to make the modular flow inner so Takesaki yields a trace" is false and is corrected throughout this note.

Proposed merge thesis (to be tested)

The shared substrate is a pair $(\mathcal{M}, \omega)$ — a von Neumann algebra with a faithful normal state, in standard form (cyclic–separating vector / purification) — plus the no-signaling / normal-steering (GHJW) structure. Claim to test:

• (i) the GPT axioms, suitably lifted, recover Pillar B on the normal state space $\mathcal{S}_*(\mathcal{M})$;
• (ii) modular theory on the same $(\mathcal{M}, \omega)$ delivers Pillar A (KMS thermality, geometric modular flow, crossed-product geometry/entropy);
• (iii) the GPT "agent / preparation / measurement" is replaced by the crossed-product internal observer, dissolving the cosmological category clash.

---

The attempt (carried as far as it honestly goes)

I carry the three sub-claims to their honest limit, then assemble the strongest merge that survives.

(i) GPT $\Rightarrow$ Born form, lifted to infinite dimensions — PARTIAL SUCCESS

The genuinely available lift is not the full Hardy/CDP reconstruction but only the Born-form rigidity. The relevant result (Torres Alegre, Causal Rigidity of Born-Type Probability Rules in Infinite-Dimensional Operational Theories, arXiv:2602.09056, 7 Feb 2026 [verified real; peer-review status unverified — preprint]): given

1. (NSS) no superluminal signaling via local measurements,
2. (NS) normal steering via purification = a $\sigma$-additive GHJW / Hughston–Jozsa–Wootters theorem, and
3. ($\sigma$-affinity) probabilities respect countable preparation mixtures,

any rule $P(\phi \mid \psi) = \Phi(\tau(\psi, \phi))$ with continuous nondecreasing $\Phi$, $\Phi(0)=0$, $\Phi(1)=1$, must satisfy $\Phi(p) = p,$ i.e. $\Phi = \mathrm{id}$: the linear Born rule is the unique no-signaling fixed point, established "via von Neumann algebras through normal state spaces and the GNS representation." This dovetails with the wiki's Born-rule treatment (2026-06-08-born-rule-and-quantum-linearity.md, KA-3 / A-3-REF), now in $\infty$-dim.

Two honest caveats. First, the paper is a rigidity result: §6 assumes a von Neumann-algebra representation already exists, then recovers $\tau = \lvert\langle\phi\rvert\psi\rangle\rvert^2$ — it does not derive the Hilbert space (see Where it breaks, BREAK 1). Second, "linearity" here is the post-processing functional $\Phi$, not the dynamical Schrödinger linearity of the Gisin–Polchinski no-signaling track; these are distinct and must not be conflated. Referee addendum: the Born form independently extends to Type II/III via Gleason–Christensen–Yeadon (Gleason's representation on all von Neumann algebras with no type-$\mathrm{I}_2$ summand) ESTABLISHED — so Pillar A is better-footed in Type III than the bare draft conceded.

Verdict (i): the Born form and post-processing linearity lift to $\infty$-dim normal state spaces. PARTIAL SUCCESS.

(ii) Modular theory $\Rightarrow$ Pillar A on the same $(\mathcal{M}, \omega)$ — SUCCESS, with a split tag

Tomita–Takesaki gives $\sigma_t$ and the KMS condition from $(\mathcal{M}, \omega)$ alone; Connes–Rovelli thermal time is $\sigma_t$ up to the Connes cocycle (state-independence modulo inner automorphisms = the canonical $\mathrm{Out}(\mathcal{M}) = \mathbb{R}$ for $\mathrm{III}_1$). This needs no agent.

But the components carry different epistemic status and must be split:

• the KMS condition at $\beta = 1$ is automatic ESTABLISHED;
• "thermal time $=$ physical time" is the Connes–Rovelli Thermal Time Hypothesis [SPECULATIVE/CONTESTED — identifying $\sigma_t$ with physical time risks presupposing a background temporal notion];
• "emergent causal/geometric modular flow" (that $\sigma_t$ is a boost / horizon flow) requires $\omega$ to be the symmetric vacuum (Bisognano–Wichmann / Hislop–Longo / CHM) — an input, not an output ESTABLISHED only for those special pairs; OPEN in general. For generic states modular flow is non-local.

So the robust co-derived content of (ii) is KMS + the Type II entropic structure (trace, finite $S_{\mathrm{gen}}$, semiclassical GSL), not "emergent causal flow," which silently inherits the BW/CHM circularity (../GAPS_AND_CONTRADICTIONS.md, the "reorganization vs generation" gap).

(iii) Internal observer dissolves the agent clash — PARTIALLY (relocation, not dissolution)

CLPW's observer is genuinely internal: a clock dof $q$ adjoined to $\mathcal{M}$, with the global modular flow imposed as a gravitational gauge constraint $\hat{H} = H_{\mathrm{mod}} + q = 0 \quad\text{(Page–Wootters relational time)},$ so the crossed-product algebra $\hat{\mathcal{M}} = (\mathcal{M} \otimes B(L^2\mathbb{R}))^\sigma$ is the algebra of boost-invariant (gauge-invariant) operators. The PW constraint is provably equivalent to Dirac constraint reduction (De Vuyst–Höhn–Kirklin arXiv:2412.15502; "Trinity" arXiv:1912.00033; Calcinari–Gielen, Quantum 9, 1610, 2025) ESTABLISHED. Note: the constrained generator $H_{\mathrm{mod}} = -\log\Delta$ is unbounded both ways and is not even affiliated to $\mathcal{M}$; boundedness-below is a property of the clock energy $q \ge 0$, not of the modular/boost flow.

The 2025 covariant-observer extension (arXiv:2511.00622, An algebra for covariant observers in de Sitter space) promotes the observer's geodesic to a dynamical dof transforming under $SO(1,d)$, yielding a quantum reference frame $L^2(SO(1,d))$ and a Type $\mathrm{II}_\infty$ trace "without requiring an external observer as background." The 2025–2026 closed-universe construction (Abdalla–Antonini–Iliesiu–Levine, arXiv:2501.02359, JHEP 02(2026)108) goes further: the global closed-universe Hilbert space is effectively one-dimensional, yet an internal observer of entropy $S_{\mathrm{Ob}}$ carries an emergent $\sim e^{S_{\mathrm{Ob}}}$-dimensional QM reducing to the modular crossed product at leading order (with a stated uniqueness theorem), valid only semiclassically (errors $\sim e^{-S_{\mathrm{Ob}}}$, big-crunch-capped) INFERENCE.

So the agent is demoted to an internal quantum reference frame — the CLPW bridge the goal names — but as a kinematic relocation within a semiclassical window, not a derivation that the GPT tomographic/purification axioms survive internally.

The merge, assembled

On a common $(\mathcal{M}, \omega)$ in standard form, under the no-signaling / normal-steering (GHJW) constraint:

• (ii) gives KMS thermality and the Type II entropic structure with no agent;
• (i) gives the Born form / post-processing linearity from no-signaling on the normal state space;
• (iii) replaces the GPT external agent by an internal crossed-product reference frame.

The bridge term is no-signaling / normal-steering (GHJW), which is simultaneously the load-bearing GPT axiom (Pillar B) and a microcausality consequence (Pillar A). Crucially, "non-circular" here means neither pillar's output is the other's input — both nonetheless share the presupposition $(\mathcal{M}, \omega)$ + purification/standard-form (Pillar A's "$\sigma$-additive normal steering via purification" and Pillar B's cyclic–separating vector are the same structural assumption in two costumes). This is a shared hidden substrate, not a vicious $A \vdash B$ or $B \vdash A$ loop. This is the strongest version of the merge that can be honestly written.

---

Where it breaks (the precise obstruction)

The construction breaks at three precisely-located steps, in increasing depth.

BREAK 1 — the tensor-product / composition axiom has no Type III lift

The part of GPT that does the framework work — local tomography $\dim V_{AB} = \dim V_A \dim V_B$, forcing the tensor product and the complex field — presupposes that composite systems factorize: $\mathcal{H}_{AB} = \mathcal{H}_A \otimes \mathcal{H}_B$, with local density matrices and finite-dim $V$. A Type $\mathrm{III}_1$ factor has no tensor factorization and no density matrices (Connes–Haag–Araki) ESTABLISHED. So "local tomography," the axiom that singles out QM's composition rule, is not even formulable on a single $\mathrm{III}_1$ factor. The 2026 causal-rigidity theorem is honest about this: it lifts only the Born-rule form, "does NOT fully reconstruct quantum theory (no Hilbert spaces derived)," and "centers on normal states; does NOT address Type III explicitly."

Referee caveat (the no-go is softer than "impossible"). Composition in AQFT is not built from a single factor but from nested/spacelike-separated inclusions. When the net satisfies the split property (Buchholz 1974 for the free field; Doplicher–Longo, CMP 1984), there is an intermediate Type I factor $N$ with $\mathcal{A}(O_1) \subset N \subset \mathcal{A}(O_2)$ for $O_1 \Subset O_2$, giving a genuine factorization $\mathcal{H} \cong \mathcal{H}_1 \otimes \mathcal{H}_2$ and statistical independence — at the cost of a "fuzzy collar." So a local-tomography-type statement is formulable for split inclusions. The honest classification is therefore domain-of-validity mismatch + missing-theorem gap, NOT a clean no-go. See [OP-GPT-TYPE3] below.

BREAK 2 — continuous reversibility and purification fail in infinite dimensions

Hardy's Axiom 5 (a continuous reversible transformation between any two pure states) and the CDP purification axiom have no valid verbatim $\infty$-dim form:

• not all purifications of a given state are connected by reversible (unitary) transformations — the canonical obstruction is the unilateral shift on $\ell^2(\mathbb{N})$, an isometry that is not unitary (it is not surjective), so the would-be reversible connection is one-way ESTABLISHED. (The bilateral shift on $\ell^2(\mathbb{Z})$ is unitary — the point is "norm-preserving $\ne$ reversible," not that the bilateral shift breaks anything.)
• the finite-dim characterization of purification fails because the set of orthogonal projections / pure-state space is not (sequentially) norm-compact ESTABLISHED.

Thus the two axioms that, in finite dim, force the reversible-transitive symmetry group $PU(d)$ (hence the Hilbert-space geometry) have no valid verbatim $\infty$-dim form.

Referee caveat. "Provably FAIL" overstates: CDP/Hardy were proved only in finite dimension and explicitly bracketed $\infty$-dim technicalities; "unproven there" is not "disproven there." And a working $\sigma$-additive purification surrogate does hold in $\infty$-dim (it is exactly the normal-steering axiom of arXiv:2602.09056). So BREAK 2 is "no valid verbatim transfer + genuinely open whether a surrogate reconstructs the framework," not a proof of impossibility. CONTESTED as stated

BREAK 3 — the internal observer is window-limited and may fail nonperturbatively

The physical $\mathrm{III}_1 \to \mathrm{II}$ reading (finite $S_{\mathrm{gen}}$, observer energy $q \ge 0$, the GSL) requires a clock Hamiltonian bounded below and a semiclassical, fixed-background regime. These inputs re-import precisely the "preferred state + fixed geometry" that Pillar A was supposed to generate — the Bisognano–Wichmann / CHM circularity in algebraic dress, now at the interpretation/entropy-matching stage (the bare algebra construction is geometry-free, per the corrected mechanism above). Giddings (arXiv:2505.22708) argues the $\mathrm{III} \to \mathrm{II}$ story is leading-order in gravitational dressing and may fail nonperturbatively ("algebraic spacetime disruption"). The closed-universe extension (arXiv:2501.02359) pushes the bridge into a background-independent setting only at leading order ($\sim e^{-S_{\mathrm{Ob}}}$ errors, big-crunch-capped). So the internal observer dissolves the agent clash only within a fixed-background, semiclassical, perturbative window — the same window in which Pillar A is conceded to be reorganization, not generation.

Mechanism correction (repeated for emphasis). The Type II trace itself is not what the clock buys — that is purely algebraic (Takesaki duality, any $\mathrm{III}_1$, any faithful normal state). What is imposed by hand and what Giddings threatens nonperturbatively is the physical observer (finite $S_{\mathrm{gen}}$ / GSL), not "Takesaki's trace." [A-CP-KILLING]

---

Furthest defensible claim

INFERENCE On a single shared substrate — a von Neumann algebra with faithful normal state $(\mathcal{M}, \omega)$ in standard form (cyclic–separating vector / purification), carrying the no-signaling + normal-steering (GHJW) constraint — two derivations run from the same input without either's output feeding the other:

• (a) the Born-rule form and $\sigma$-affine post-processing linearity on the normal state space (infinite-dim, via arXiv:2602.09056, 2026 — which recovers QM through the normal state space + GNS, i.e. presupposes, does not derive, the vNa/Hilbert structure; the form independently extends to Type II/III via Gleason–Christensen–Yeadon ESTABLISHED); and
• (b) the KMS condition ESTABLISHED plus a Type II semifinite trace, finite generalized entropy, and a semiclassical GSL via the crossed product $\mathcal{M} \rtimes_\sigma \mathbb{R}$ (the trace is purely algebraic for any $\mathrm{III}_1$ and any faithful normal state; the finite $S_{\mathrm{gen}}$, observer energy, and GSL additionally need a bounded-below clock + semiclassical window),

with the "observer" demoted to an internal crossed-product quantum reference frame, not an external agent.

But this shared substrate cannot derive the Hilbert-space framework itself (tensor-product composition, complex field): the GPT framework-fixing axioms — local tomography, purification, continuous reversibility — have no verbatim Type III formulation and have no valid verbatim $\infty$-dim form, and the crossed-product physical reading exists only in a fixed-background semiclassical perturbative window.

The merge is therefore substantive for the Born-form + KMS/entropic sublayer and a thin methodological bridge (shared no-signaling tool) for the framework-derivation sublayer.

Two language caveats the referee insists on: "thermal-time $=$ physical time" is SPECULATIVE/CONTESTED and "emergent causal/modular flow" inherits the BW/CHM circularity — so neither is part of the robust co-derived core (which is Born-form + linearity together with KMS + the Type II entropic trace structure).

---

If it fails / weaker true statement

If BREAK 1–2 were circumvented (an $\infty$-dim / Type III analog of local tomography + purification + reversibility recovering tensor composition and the complex field on normal state spaces of a $\mathrm{III}_1$ factor were found — e.g. a split-property-based or relational surrogate), the merge would upgrade from "Born-form sublayer" to "full framework," making the GPT reconstruction genuinely co-fundamental with modular geometry.

The honest weaker true statement if even the partial merge fails to count as more than re-description: the two programs share only no-signaling as a tool and BMV as a logical gate (the iteration-2 "thin bridge" verdict [REG-SYNTH-5]), and "single substrate" is an over-reading — $(\mathcal{M}, \omega)$ is a common notational home, not a common derivation. OPEN The merge sits between these two poles; the verdict below adjudicates which pole it is nearer.

Next move. Test whether the 2026 causal-rigidity theorem's "$\sigma$-affinity + normal steering" can be strengthened to fix composition (not just the probability rule) on $\mathrm{III}_1$ normal states. If it cannot even in principle (because composition needs factorization), BREAK 1 is a true no-go on a single factor; if split-property factorization suffices, BREAK 1 is only a missing-theorem gap.

---

Proposed registry items (with referee verdicts)

[REG-SYNTH-6] — hypothesis-refinement — keep (severity: MAJOR) — refined

Proposed statement. The two iteration-2 programs share a single algebraic-operational object $(\mathcal{M}, \omega)$ (von Neumann algebra + faithful normal state) under a no-signaling / normal-steering (GHJW) constraint, on which they non-circularly co-derive a common sublayer: (a) Pillar B's Born-rule form + linearity, lifted to $\infty$-dim normal state spaces (arXiv:2602.09056, 2026); and (b) Pillar A's KMS + (via an internal crossed-product clock) a Type II trace + finite $S_{\mathrm{gen}}$ + GSL (CLPW 2022). This upgrades the iteration-2 "thin methodological bridge" to a "substantive partial merge at the Born-form + thermal/entropic sublayer." [tag: inference]

Referee verdict — keep: true; severity: MAJOR. The referee flags a category equivocation: both cited sources take $(\mathcal{M}, \omega)$ as input scaffold, not co-derived output; arXiv:2602.09056 is a rigidity result that (its §6) assumes a vNa representation; the crossed-product leg takes $\mathrm{III}_1$ + $\omega$ as given and needs a geometric background for $S_{\mathrm{gen}}$ / GSL. "No external agent" is false (an observer/clock and a steering counterparty are both load-bearing). The GSL is Faulkner–Speranza 2024, semiclassical-only — NOT "CLPW 2022" (misattribution flagged). "Substantive partial merge" risks restating the iteration-2 thin-bridge verdict as if transcending it.

Refined statement (adopted). INFERENCE The two programs share a common formal arena $(\mathcal{M}, \omega)$ and a common theorem family (GHJW / Hughston–Jozsa–Wootters steering): Pillar B invokes it (normal steering via purification) to force the Born-rule form / post-processing linearity $\Phi = \mathrm{id}$ on $\infty$-dim normal state spaces (arXiv:2602.09056, 2026 — a rigidity result presupposing the operational + vNa scaffold), and Pillar A invokes the same GHJW ensemble-ambiguity to ground no-signaling, while separately using Tomita–Takesaki/KMS on $(\mathcal{M}, \omega)$ plus an adjoined-observer crossed product to obtain a Type II trace and (on a fixed Killing background, semiclassically) finite $S_{\mathrm{gen}}$ and a GSL (CLPW 2022 for the dS $\mathrm{II}_1$ finite trace; Faulkner–Speranza 2024 for the GSL). Neither program derives $(\mathcal{M}, \omega)$; each takes it as input and reads off different downstream structure; an internal observer and a steering counterparty are load-bearing, so "no external agent" is false. This is a sharper articulation of the shared-tools methodological bridge (now pinned to the $(\mathcal{M}, \omega)$ arena + GHJW family), correctly scoped — substantive only for the Born-form + KMS/entropic sublayer.

[OP-GPT-TYPE3] — open-problem — keep (severity: MAJOR) — refined

Proposed statement. Do the GPT framework-fixing axioms (local tomography / tensor-product composition, purification, continuous reversibility) admit any formulation on a Type $\mathrm{III}_1$ factor or its $\infty$-dim normal state space? Decide: true NO-GO (composition needs factorization, impossible on $\mathrm{III}_1$) or missing-theorem gap? [tag: open]

Referee verdict — keep: true; severity: MAJOR. The "true no-go / composition impossible on $\mathrm{III}_1$" horn is defeated by the split property: for nested regions a split-satisfying net supplies an intermediate Type I factor and hence genuine factorization + statistical independence (Buchholz; Doplicher–Longo). The Born form is already lifted to general-vNa normal states incl. Type III; a $\sigma$-additive normal purification surrogate exists; so the claim "composition needs factorization, impossible on $\mathrm{III}_1$" is false as a no-go premise. Recent work even gives a rich entanglement/composition theory for commuting Type $\mathrm{III}_1$ subalgebras (arXiv:2409.17739).

Refined statement (adopted). OPEN — domain-of-validity mismatch + missing-theorem gap; explicitly NOT a no-go Can the GPT framework-fixing axioms be reformulated as $\sigma$-additive / normal / split-based / relational surrogates on the normal state space of the hyperfinite $\mathrm{III}_1$ factor (or a Haag–Kastler net), so as to single out QM the way the finite-dim package does? Structural facts: (a) a single $\mathrm{III}_1$ factor has no trace / minimal projections / density matrix / factorization, so the finite-dim axioms are not statable verbatim ESTABLISHED; (b) the split property restores factorization + independence for nested regions ESTABLISHED; (c) the Born form lifts to Type III via arXiv:2602.09056 + Gleason–Christensen–Yeadon ESTABLISHED; (d) a $\sigma$-additive normal purification surrogate holds in $\infty$-dim ESTABLISHED; (e) CDP/Hardy were proved only finite-dim — "provably fail" is false as stated CONTESTED. OPEN: a single complete reconstruction using a split-based/relational composition surrogate + modular-compatible continuous-reversibility surrogate + $\sigma$-additive purification that uniquely fixes QM on the $\mathrm{III}_1$ normal state space. No no-go theorem currently forbids it.

[ASSUME-INTERNAL-CLOCK] — assumption — keep (severity: MINOR) — refined

Proposed statement. The crossed-product internal observer that dissolves the GPT external-agent clash is a semiclassical clock with a Hamiltonian bounded below, with the global modular/boost flow imposed as a Page–Wootters gauge constraint, valid in a fixed-background perturbative-in-$G_N$ window; these inputs re-import the preferred-state / fixed-geometry structure Pillar A aims to generate (BW/CHM circularity), and the $\mathrm{III} \to \mathrm{II}$ conversion may fail nonperturbatively (Giddings). [tag: inference]

Referee verdict — keep: true; severity: MINOR. Two corrections: (1) which Hamiltonian is bounded below — the clock energy $q \ge 0$, NOT the constrained modular generator $H_{\mathrm{mod}} = -\log\Delta$ (unbounded both ways, not affiliated to $\mathcal{M}$); "dissolves" overstates — it relocates the agent inside the system at semiclassical order. (2) The genuinely background-independent closed-universe case has a concrete leading-order construction (arXiv:2501.02359, JHEP 02(2026)108) reducing to the modular crossed product at leading order with intrinsic errors $\sim e^{-S_{\mathrm{Ob}}}$ — so the obstruction is not only Giddings' disruption but also the construction's own error budget.

Refined statement (adopted). INFERENCE + OPEN The crossed-product internal observer relocates (not dissolves) the GPT external-agent clash by replacing the external preparer/reader with an internal clock whose energy is bounded below ($q \ge 0$), imposing $H_{\mathrm{mod}} + q = 0$ as a Page–Wootters relational gauge reduction [ESTABLISHED: PW $=$ Dirac reduction $=$ crossed-product gauge-invariant algebra]. Boundedness-below applies to the clock, not the two-sided-unbounded modular generator. The fixed-geometry / preferred-state structure Pillar A aims to generate re-enters at the entropy/geometry-interpretation stage (the algebra construction is geometry-free by Connes–Takesaki), via the BW/CHM requirement that modular flow be a geometric boost — a real circularity in relational dress INFERENCE. The bridge is established only within a fixed-background, semiclassical, perturbative-in-$G_N$, observer-relative window; the closed-universe construction (arXiv:2501.02359) pushes it to leading order in a background-independent setting but only semiclassically. OPEN whether the bridge survives nonperturbatively (Giddings 2505.22708) and whether the GPT tomographic/purification axioms can even be instantiated by a single internal observer in a one-dimensional global Hilbert space.

---

Verdict

SUBSTANTIVE PARTIAL MERGE at one sublayer; thin methodological bridge at the deeper sublayer; NOT empty re-description. Precisely:

1. The Born-rule form + linearity (Pillar B's probabilistic core) and the KMS + Type II generalized-entropy structure (Pillar A) genuinely descend from a single shared object $(\mathcal{M}, \omega)$ in standard form under no-signaling / normal-steering, with neither pillar's output feeding the other and the "observer" demoted to an internal crossed-product quantum reference frame — so the agent/no-agent category clash is bridgeable for this sublayer (CLPW + covariant-observer 2025 + closed-universe 2025–26 + infinite-dim causal-rigidity 2026). This is a real, correctly-scoped upgrade to the iteration-2 thin-bridge verdict. Caveat: "emergent causal/modular flow" and "thermal-time $=$ physical time" are not in the robust core (BW/CHM circularity; Connes–Rovelli is contested).

2. BUT the framework-derivation core of Pillar B — local tomography, purification, continuous reversibility, which single out Hilbert space, the complex field, and tensor composition — has no verbatim Type III formulation and no valid verbatim $\infty$-dim form, and the crossed-product physical reading exists only in a fixed-background semiclassical perturbative window that re-imports the geometry Pillar A should generate. So the deep "derive the quantum framework and geometry from one substrate" merge does not go through; it is blocked at an identified, possibly-circumventable step [OP-GPT-TYPE3] (the split property softens the would-be no-go to a missing-theorem gap).

The honest answer to "single non-circular substrate for both pillars": YES for the Born-form + thermal/entropic sublayer, NO (with a precisely located, not-yet-no-go obstruction) for the Hilbert-space-framework sublayer. Not empty re-description, because the shared no-signaling principle does real, non-tautological work on both sides; not a full merge, because composition/tensor-product cannot (yet) be posed on a single Type $\mathrm{III}_1$ factor.

Confidence: medium.

---

Open subquestions

• OPEN Does any relational / algebraic / split-based surrogate for local tomography on a Type $\mathrm{III}_1$ normal state space recover tensor-product composition and the complex field, or is composition-needs-factorization a true no-go on a single factor? ([OP-GPT-TYPE3])
• OPEN Can the 2026 causal-rigidity theorem be strengthened from fixing the probability rule ($\Phi = \mathrm{id}$) to also fixing the composition rule on $\infty$-dim normal state spaces, without assuming factorization?
• OPEN/CONTESTED Does the $\mathrm{III} \to \mathrm{II}$ crossed-product trace (and hence the internal-observer dissolution) survive nonperturbatively in $G_N$, or does Giddings' algebraic spacetime disruption remove it precisely in the background-independent cosmological regime? ([OP-CP1], [ASSUME-INTERNAL-CLOCK])
• OPEN In a genuinely closed universe (no asymptotic boundary, no Killing vacuum), is there an $(\mathcal{M}, \omega)$ whose modular flow is geometric as an output, so the internal-observer construction is not parasitic on a pre-chosen symmetric vacuum (BW/CHM circularity)? — and can a single internal observer instantiate tomographically-complete preparations in a one-dimensional global Hilbert space (arXiv:2501.02359)?
• OPEN Is the peer-review status of arXiv:2602.09056 confirmed, and does its "normal steering via purification" secretly require a Type I/II structure, which would shrink even the partial-merge sublayer? unverified — preprint
• OPEN Can "no-signaling does real work on both sides" be made into a formal statement that the GHJW / normal-steering constraint is logically independent of both the Born form and the modular KMS condition, ruling out the "empty re-description" reading rigorously?

---

See also

• ../FINDINGS.md — consolidated findings (Born-form rigidity; crossed-product Type II entropy; reorganization-vs-generation).
• ../HYPOTHESES.md — H0, H2, H4 (algebra/entanglement before geometry); REG-SYNTH-5/6.
• ../OPEN_PROBLEMS.md — OP-CP1 (nonperturbative fate), OP-BORN-QUANT (quantitative-Born wall), OP-GPT-TYPE3 (this note).
• ../ASSUMPTIONS_LEDGER.md — A-2/A-3/A-4 (Born/linearity), A-CP-KILLING (interpretation-stage geometric input).
• ../GAPS_AND_CONTRADICTIONS.md — reorganization vs generation; BW/CHM circularity.
• ../EPISTEMICS.md — tag taxonomy; domain-of-validity-mismatch vs missing-theorem-gap vs no-go.
• ../domains/quantum-mechanics.md — A4/A5 (Gleason/Busch; linearity).
• ../domains/quantum-field-theory.md — Type $\mathrm{III}_1$ local algebras; modular theory.
• ../domains/general-relativity.md — horizons, Killing structure, generalized entropy.
• ../domains/cosmology.md — closed-universe observer, de Sitter static patch.
• ../domains/information-theory.md — no-signaling, GHJW, GPT reconstructions.
• ../domains/mathematics.md — von Neumann algebras, crossed products, Takesaki duality.
• Sibling iteration-2 notes: 2026-06-08-unified-program-synthesis.md, 2026-06-08-born-rule-and-quantum-linearity.md, 2026-06-08-algebraic-background-independence.md, 2026-06-08-de-sitter-and-emergent-spacetime.md, 2026-06-08-tabletop-quantum-gravity.md, 2026-06-08-entropic-and-emergent-gravity.md.