Iteration-4 Synthesis: Convergence, Not Direction — the dS Cardinality No-Go

Status: Active synthesis — coherence verdict UNCHANGED at PARTIAL; structural argument SHARPENED Last updated: 2026-06-08 Iteration: 4

This note folds the iteration-4 attacks and the synthesis-verdict review into one record. The honest one-line headline: iteration 4 moved no headline claim — it reinforced all three — by closing the last open escape route from iteration 3, settling one open problem as a sharp disjunction, and promoting one obstruction toward a near-no-go; the marginal yield is now a sharper obstruction or a self-correction, not new positive structure. INFERENCE, high

The three headline claims are unchanged: (1) coherence is PARTIAL; (2) the geometry-from-algebra program ENCODES but does not GENERATE geometry; (3) the central wager is not-yet-physics — it has no distinguishing near-term test. What changed is precision and the closing of escape routes — convergence behavior, the third consecutive iteration at this verdict. INFERENCE, high

---

Did the standing verdict change?

NO. Unchanged at the headline level, sharpened at the structural level. INFERENCE, high

The standing verdict holds across iterations $2 \to 3 \to 4$: PARTIAL coherence; geometry-from-algebra encodes, does not generate, geometry; the wager is not-yet-physics. Iteration 4 did not move any of these — it reinforced all three. The central diagnosis ("encodes, does not generate") is now backed by a much harder argument than iteration 3 supplied.

The de Sitter first-law obstruction is upgraded from iteration-3's hedged OPEN sharpened obstruction with one un-foreclosed escape to a scoped near-no-go: the one escape (the $SO(d+1,1)$ static-patch modular-Hamiltonian family) was actually executed and fails on three independent, individually-sufficient grounds. The decisive one is a hard cardinality mismatch — a finite-dimensional isometry orbit ($\dim\mathcal O = 2d = 6$ in 4d) supplying at most $\dim SO(d+1,1) = 10$ scalar Killing-charge constraints, which cannot Radon-invert to the infinite-dimensional local field $\delta G_{\mu\nu}(x)$ (6 independent functions on $M^4$ after Bianchi). [INFERENCE, high — rests on the established FGHMVR Radon-inversion mechanism and the verified finite orbit dimension; see the attempt note.]

This is NOT an unconditional theorem. It is explicitly scoped to the strict unitary, single-static-patch, isometry-orbit, common-state route. Two routes remain genuinely open: the non-unitary dS/CFT pseudo-entropy route (arXiv:2511.07915), and Jacobson's non-algebraic small-ball continuum (which already works, but is non-holographic). INFERENCE, high

The load-bearing citation is verified. Chen–Xu, arXiv:2511.00622, "An algebra for covariant observers in de Sitter space" (Bin Chen, Jie Xu) is the published execution of exactly this route — an averaged modular crossed product over static patches, the $L^2(SO(1,d))$ QRF, all $SO(1,d)$ constraints — and it reaches entropy-matching and a (second-order) quantum first law but derives no local Einstein equation / $\delta G_{\mu\nu}$, landing precisely where the cardinality argument predicts. ESTABLISHED that the paper exists and is so characterized

---

Program A status

OBSTRUCTION SHARPENED / one target promoted toward no-go — NOT advanced toward geometry-generation. INFERENCE, high

Program A (geometry-from-algebra) made zero progress toward generating geometry at $\Lambda>0$ and lost its single highest-leverage open theory target. Specifically:

• HYP-dS-NOMIN promoted from iteration-3's [INFERENCE-conjecture + one un-foreclosed escape] to a scoped near-no-go (new registry item HYP-dS-NOMIN-R3):

  There is no derivation of the local tensor equation $\delta G_{\mu\nu} = 8\pi G\,\delta T_{\mu\nu}$ from a dS static-patch generalized-entropy first law $\delta S_{\rm gen} = \delta\langle K\rangle$ via the FGHMVR mechanism, even after enlarging to the full $SO(d+1,1)$ family, for three independent reasons: (i) finite-dimensional orbit $\Rightarrow$ $\le 10$ scalar Killing-charge constraints, cannot invert to the infinite-dim local field; (ii) the family is geometric in a common state only at Bunch–Davies, where the probing perturbation $\delta\Psi$ is null; (iii) the non-trivial first-law content is second-order, where $\delta^2 S_{\rm gen} < 0$ renders it a concave GSL inequality, not an invertible equality. INFERENCE, high; near-no-go for the scoped route

• A reusable generalization extracted (new registry item HYP-dS-CARDINALITY): Einstein-from-first-law requires an infinite-dimensional local region family probing every bulk point (in AdS: an RT surface through every bulk point in every orientation, Radon-invertible to $\delta G_{\mu\nu}(x)$). Isometry orbits of a maximally-symmetric background are finite-dimensional and yield only global Killing charges. Therefore local dS dynamics from entanglement requires an independently infinite local family (Jacobson's small geodesic balls — one at every bulk point in every frame), which the isometry orbit structurally lacks. This converts a soft "finiteness worry" into a hard dimension count, and is the most substantive product of iteration 4. INFERENCE, high

The escape routes that survive — dS/CFT pseudo-entropy with complex central charge $c = i\,3R_{\rm dS}/2G_N$ (2511.07915); Jacobson's continuum realized inside a single CLPW Type II$_1$ algebra (importing the local continuum without a boundary) — are exactly the iteration-3-listed verdict-movers, now isolated as the only remaining ways Program A could move from encode to generate at $\Lambda>0$.

Status verb: the diagnosis is reinforced and the live frontier is narrower, not closer to a positive result.

---

Merge status

UNCHANGED in kind, SHARPENED in detail. INFERENCE, medium

The A$+$B merge remains "narrow success / deep failure": the Born-form $+$ thermal/entropic sublayer merges non-circularly on a shared $(\mathcal M, \omega)$; the Hilbert-space-FRAMEWORK sublayer (GPT composition / local-tomography) does not. Iteration 4 did not convert the partial merge into a full merge, nor into a clean no-go — but it settled OP-48 / OP-GPT-TYPE3 as a sharp DISJUNCTION with a precise dividing line, replacing iteration-3's vaguer "candidate-no-go / missing-theorem gap":

• (a) NO-GO for exact GPT local tomography / tensor composition on a single Type III$_1$ factor (no factorization, no density matrix). ESTABLISHED
• (b) State-/collar-dependent YES for nested regions $\mathcal O_1 \subset \mathcal O_2$ under the split property — an exact tensor factorization $\mathcal H \cong \mathcal H_1 \otimes \mathcal H_2$ with local density matrices and statistical independence does exist, defeating the would-be hard no-go as a no-go. ESTABLISHED
• (c) But it is NOT a clean operational reconstruction, for two live-web-verified reasons:
  1. The canonical intermediate Type I factor $\mathcal F = \mathcal R(K) \vee J\,\mathcal R(K)\,J$ is made canonical only by the vacuum $\Omega$'s modular conjugation $J$ (infinitely many non-canonical factors otherwise) — re-importing the BW/CHM geometric reference state Program A was meant to generate (new registry item HYP-SPLIT-GEOM: the split-property rescue relocates the circularity to the factorization stage, it does not escape it).
  2. It requires a nonzero collar and fails for touching/adjacent regions, so the operationally natural "compose adjacent subsystems" has no exact tensor structure.

The split-property facts (canonical factor fixed by $J$ relative to the vacuum; nontrivial relative commutant) are web-verified. Confirmed (via the literature track) that the 2026 Born-form rigidity result arXiv:2602.09056 and its finite-dim predecessor (both Torres Alegre) treat only the single-system probability RULE and do not fix composition — so the merge's deeper sublayer stays unmerged.

Net: a full merge still requires solving (c) — a state-/collar-free composition surrogate on a Type III$_1$ factor — which remains OPEN. (See the forthcoming attack note 2026-06-08-iter4-gpt-type3-resolution.md.)

---

Literature-refresh impact

ONE genuinely claim-moving result; ONE citation defect to scrub; the rest re-confirmation. ESTABLISHED Every citation independently checked on the live web verifies exactly; NO fabrication anywhere — there are no fabricated IDs to scrub, only one misattribution and assorted residue.

Claim-moving (action = update): Muon $g-2$

arXiv:2505.21476 (Muon $g-2$ Theory Initiative WP2025), verified via WebSearch against the Theory-Initiative and CERN Courier pages, switches the leading-order HVP input to a lattice-QCD average, shifting the SM prediction to

$$a_\mu^{\rm SM} = 116\,592\,033(62)\times 10^{-11},$$

so that the difference with the experimental world average is $\Delta a_\mu = 38(63)\times 10^{-11}$ — no tension. The historic $\sim 4$–$5\sigma$ muon $g-2$ anomaly is thereby largely resolved into SM-consistency (an internal lattice-vs-$R$-ratio / CMD-3 tension persists). Any wiki page still listing muon $g-2$ as a live BSM anomaly is now STALE and must be updated. [VERIFIED]

Scope note: muon $g-2$ is peripheral to the central wager — it touches the Standard-Model-precision core, not geometry-from-algebra — so it does not move the headline verdict, only a domain page. INFERENCE, high

Citation to scrub (defect, not fabrication)

arXiv:2308.14797 is REAL but was MISATTRIBUTED in the iteration-3 synthesis note as "Forste et al." Verified on the live arXiv abstract: it is "Modular Hamiltonian for de Sitter diamonds," single-authored by Markus B. Fröb, JHEP 12 (2023) 074. Iteration 4 correctly flags and fixes this (new item GAP-FROEB-CITATION); a cleanup task to propagate the fix has been spawned. [VERIFIED]

Three previously-flagged IDs cleared (unverified → [web-confirmed])

All three were genuinely flagged with prudence; the papers are genuine. The one required edit is the inner-product-not-linearity trim on 2601.13012. [VERIFIED]

Non-moving refreshes (action = add / none)

• Newer collapse-model bounds (2401.04665, 2411.17588, 2512.02838): tighten but do not close CSL/DP space; KA-4 / O-4 conclusion unchanged.
• First Lorentzian-signature asymptotic-safety fixed-point evidence (D'Angelo, arXiv:2310.20603, PRD 109 066012): incremental, still CONTESTED, no distinguishing test.
• DESI $w(z)$ still DR2; the $2.8$–$4.2\sigma$ hint stands; DR3 forthcoming — no claim moves.
• Island-beyond-AdS and KLS / crossed-product follow-ups already current.

---

How far from a conclusion

Two DIFFERENT distances, and they must not be conflated. INFERENCE, high

Analytical saturation — CLOSE (≈ 1–3 more focused iterations)

The analytical conclusion is already largely in hand and stable across iterations $2 \to 3 \to 4$:

1. "Is a universal theory possible?" is answered in disambiguated form:
   • (A) a consistent unified GR$+$QFT framework is POSSIBLE at high confidence (AdS/CFT is an existence proof; the 17 catalogued clashes are domain-mismatches, not contradictions);
   • (B) a final substrate is possible but not demonstrably so (a Wilsonian tower-with-no-bottom is live);
   • (C) a parameter-fixing complete theory is least likely.
2. The one central wager is named precisely: causal/algebraic/entanglement structure precedes metric geometry, true of our $\Lambda>0$ universe.
3. Why it is not-yet-physics is established and has survived every attack across three iterations: it encodes but does not generate geometry, and it has no distinguishing near-term test.
4. Iteration 4 shrank the remaining analytical to-do list by closing the single highest-leverage open theory target (OP-41 via its only un-foreclosed route).

What remains is a finite, named list:

• settle OP-48 / OP-GPT-TYPE3 point (c) — the state-/collar-free composition surrogate (full merge vs clean no-go);
• OP-44 / OP-CP1 — the nonperturbative fate of the crossed product once gravity is dynamical (whether "Type III$_1$ for any $\Lambda$" can even be a fixed core);
• the OP-46 clause-(iii) construction — recover Killing/horizon as the OUTPUT of a genuinely non-symmetric, non-CKV modular flow (the only move that turns encode into generate).

The OP-46 escape is itself now near-vacuous under the standard geometric readout (new registry item HYP-CKV-VACUITY). Two chained theorems, web-verified, force this: (a) Borchers/Wiesbrock — half-sidedness $\Leftrightarrow$ positivity, so the $ax+b$ group is generated from the relocated positivity premise, not from neutral data; (b) Caminiti et al. (arXiv:2502.02633) — recovered boost generators are conformal Killing vectors, which presuppose a symmetric background. The vacuity is conditional on the standard readout — a non-CKV readout (modular Berry / Connes spectral distance) is the genuinely live, currently-unrealized target. After these are settled, marginal structural insight goes to near-zero and the program becomes a watch-list. [INFERENCE, high on the classification; medium that the vacuity is a true no-go rather than a limitation of the standard readout]

Object-level confirmed theory — NOT FORECASTABLE

This is a stronger obstruction than mere distance. It is gated by two things the program cannot schedule:

• (i) the $\sim 15$-order-of-magnitude empirical desert between accessible energies ($\sim$ TeV) and $M_{\rm Pl} \approx 1.22\times 10^{19}$ GeV;
• (ii) more bindingly, the central wager currently has NO distinguishing experimental test even in principle. Every live channel tests the program's logical GATE (is gravity non-classical? — BMV/QGEM, gated by matter-wave/optomechanics mass scale, $\sim 5$–$15$ yr), the status of its fixed CORES (is linearity/Lorentz exact? — objective-collapse bounds decisive $\sim 2035$; LIV/EP nulls), or its correct TARGET (eternal dS? — DESI DR3+/Euclid $w(z)$, $2$–$7$ yr). None tests geometry-from-algebra itself.

So even optimistic enabling-technology timelines would open the gate, not confirm the theory. The object-level conclusion lacks a defined experimental decision procedure — it is non-forecastable, not merely far. INFERENCE, high

---

Recommendation

HYBRID. [confidence: high]

Iteration 4 mostly sharpened obstructions and confirmed convergence — say so plainly. It produced one genuinely new high-value NEGATIVE result (the dS cardinality no-go), settled one open problem as a sharp disjunction (OP-48), promoted one open obstruction toward a near-no-go (the clause-(iii) vacuity), found ZERO new positive structure, and stabilized the PARTIAL / not-yet-physics verdict for the third consecutive iteration. The new-insight trend is DIMINISHING — each iteration's marginal yield is increasingly a sharper obstruction or a self-correction, not new positive structure or a new degree of freedom in the problem.

Recommended allocation for the remaining $\sim 1$–$3$ iterations:

1. ATTACK the three named, genuinely-decidable analytical problems. OP-48 point (c) and OP-44/OP-CP1 each saturate a gap either way; the OP-46 non-CKV-readout construction (modular Berry / Connes spectral distance on a non-symmetric state, yielding a Lorentzian metric) is the only move that would change the headline verdict — note that Connes reconstruction gives only a Riemannian manifold (new registry item GAP-NCG-LORENTZIAN-SIGNATURE, web-verified).
2. DO the two cheap documentation passes: the spawned propagation/citation-scrub task, and the muon $g-2$ staleness update.
3. SIMULTANEOUSLY stand up lightweight tracking for the four empirical channels (BMV $m^2$-vs-$m^3$; collapse bounds $\sim 2035$; $w(z)$ DESI DR3+/Euclid; LIV/EP), since those — not further analysis — gate the object-level verdict.

Pure continue-attacking risks diminishing returns; pure pivot-to-tracking prematurely abandons 2–3 genuinely decidable named problems.

Accept all new registry items — each is web-grounded and correctly tagged: HYP-dS-NOMIN-R3, HYP-dS-CARDINALITY, HYP-CKV-VACUITY, the OP-GPT-TYPE3 refinement, HYP-SPLIT-GEOM, GAP-FROEB-CITATION, GAP-LL-ENERGY-POSITIVITY-R2, GAP-NCG-LORENTZIAN-SIGNATURE.

---

Next targets

1. Attack OP-48 point (c). Exhibit (or prove impossible) a state-independent, collar-free composition surrogate that singles out QM from split-composition axioms on a Type III$_1$ factor. A constructive yes converts the partial A$+$B merge into a full merge; a proven no is a clean no-go fixing the boundary of the analytical question. Either answer saturates a named gap. OPEN
2. Resolve OP-44 / OP-CP1. The nonperturbative fate of the crossed-product Type II structure once gravity is dynamical — does a von Neumann-algebraic subregion description survive beyond all-orders-in-$1/N$, or does Giddings' algebraic spacetime disruption (arXiv:2505.22708 / 2510.24833) replace it? Decides whether "Type III$_1$ for any $\Lambda$" can be a fixed core at all. OPEN/CONTESTED
3. Attempt the OP-46 clause-(iii) escape under a NON-CKV readout. Recover a Killing/boost structure $+$ horizon as the OUTPUT of an $(\text{algebra}, \text{state})$ pair whose modular flow is genuinely non-symmetric (non-Bunch–Davies, non-Hadamard), using modular-Berry information-metric or Connes spectral distance, crucially yielding a Lorentzian (not merely Riemannian) metric. This is the only move that would turn "encode" into "generate" and change the headline verdict. OPEN
4. Attempt the two un-foreclosed dS routes that survive the cardinality no-go: (a) the non-unitary dS/CFT pseudo-entropy route (2511.07915) accepted as physics despite a complex central charge, or (b) Jacobson's infinite small-ball family realized inside a single CLPW Type II$_1$ algebra (importing the local continuum without a boundary). Either would move Program A from encodes toward generates at $\Lambda>0$. OPEN
5. Documentation/hygiene (cheap, required). Update any wiki page treating muon $g-2$ as a live BSM anomaly to SM-consistent per WP2025 (arXiv:2505.21476); propagate the spawned iteration-3 red-team fixes (heuristic-not-program relabel; Caves–Fuchs–Manne–Renes spelling; Fröb-not-Forste byline) into the iteration-2 synthesis note; promote 2603.06211 / 2601.18856 / 2601.13012 to web-confirmed, trimming the "/linearity" rider on 2601.13012.
6. Stand up lightweight experiment-tracking for the four empirical channels (BMV $m^2/m^3$ $\sim$2030s–40s; objective-collapse decisive $\sim 2035$; $w(z)$ DESI DR3+/Euclid 2–7 yr; LIV/EP nulls), since the object-level verdict is gated by these, not by further analysis — and none currently tests the geometry-from-algebra wager directly.

---

See also

Iteration-4 notes

• 2026-06-08-iter4-dS-firstlaw-SO-family-attempt.md — the executed $SO(d+1,1)$-family escape; the cardinality no-go (HYP-dS-NOMIN-R3) and its three failure counts.
• 2026-06-08-iter4-gpt-type3-resolution.md — OP-48 resolved as a sharp disjunction (Type III$_1$ no-go vs split-property qualified yes); HYP-SPLIT-GEOM. [companion attack note]
• 2026-06-08-iter4-modular-circularity.md — the OP-46 Bisognano–Wichmann/CHM circularity test; HYP-CKV-VACUITY; GAP-NCG-LORENTZIAN-SIGNATURE. [companion attack note]

Prior synthesis and registries

• 2026-06-08-iter3-synthesis.md — the iteration-3 synthesis this one updates (and whose "Forste et al." mis-citation is corrected to Fröb).
• 2026-06-08-unified-program-synthesis.md — the iteration-2 synthesis (target of the pending propagation/citation-scrub task).
• 2026-06-08-algebraic-background-independence.md — the encode-not-generate / Type III$_1$ background note.
• ../FINDINGS.md — consolidated findings register.
• ../ROADMAP.md — iteration roadmap and open-problem queue.
• ../OPEN_PROBLEMS.md — OP-41, OP-44/OP-CP1, OP-46, OP-48/OP-GPT-TYPE3.
• ../HYPOTHESES.md — H0–H7; HYP-dS-NOMIN-R3, HYP-dS-CARDINALITY, HYP-CKV-VACUITY, HYP-SPLIT-GEOM.