PN-001Settled
../panels/PN-001-CASE-08-A.mdVerdictUnmet precondition, not missing geometry — OS positivity needs a named half-space (only conjugation is in hand), and the finite-field flip suggests positivity is a parameter-indexed threshold, not an attribute of the space; the four accounts stay unrankable and the fourth smuggles in the case's own verdict.
Seatsledger, framing, hypothesis, lateral
Bound firedfalse
Type. derivation / physics
Presenting object. A space carrying a scaling flow, offered as the home for a Lefschetz-trace argument. The trace formula is in hand and verified: one local integral per place, Atiyah–Bott shape ∫|u|^{1/2}G(log|u|)/|1−u|_v d*u. At finite places the identity |1−u|_p = max(1,|u|_p) for |u|_p ≠ 1 makes it exact rather than asymptotic — verified with no failures at p = 2,3,5,7, giving weight exactly p^{−k/2} with no (1−p^{−k}) correction. The archimedean term matches the digamma form up to one universal constant, γ + log 2π, across six test functions with spread 0.002.
Friction. Everything a Lefschetz argument needs is present except the sign, and the sign is not a missing lemma — it is a missing object. Over a finite field the analogous statement is a signature computation and it works: on a surface with h = f₁ + f₂ the intersection form has signature (1,3) or (2,2), flipping exactly when |a| exceeds 2√q, giving an ample class with negative-definite complement. The space in question has a pairing and a trace with real-valued dimensions, but no ample class and no hard-Lefschetz sl₂. This cannot be repaired algebraically: over ℂ all nondegenerate symmetric forms are equivalent, signatures exist only over ℝ, and a real form is not determined by its complexification.
A second instance of the same structure has an eighty-year literature. A path integral's existence is a positivity statement — no Lorentzian measure exists, the Euclidean measure exists exactly under reflection positivity, and Wick rotation is the choice of a real form. Same shape: an algebraic object with everything present except a sign, a positivity condition not derivable from the algebra, and an import from a real structure that makes the object exist.
Background. Measured: the finite-place weights (p = 2,3,5,7, exact, no correction term); the archimedean constant γ + log 2π across six test functions, spread 0.002; the signature flip at |a| = 2√q, with the orthogonal complement's top eigenvalue −0.033 at a = 5 and +0.101 at a = 6.
Stated: over ℂ all nondegenerate symmetric forms are equivalent; signatures exist only over ℝ; an ample cone is a cone in a real vector space; a real form is not determined by its complexification.
Stated, from the second instance: Osterwalder–Schrader reconstruction succeeds under reflection positivity; the import is checkable independently of what it is used to prove; constructive field theory succeeded in low dimensions and not in four. A reflection is known to exist here — conjugation — with no positivity condition attached to it.
Four accounts are on the table: the missing geometry is a Riemann–Roch over the field with one element; it is a foliated space with positivity from leafwise Hodge theory; the ample class exists on the span of the correspondences rather than on the space; or no import is possible because any geometry rich enough already contains the conclusion.
Must predict. What plays the role of the half-space. Reflection positivity is positivity of the pairing a reflection induces on a half-space, not the existence of the reflection; the reflection here already exists and the positivity does not. A satisfactory answer must name the half-space and say what the induced pairing is — or say why no such decomposition exists, which would be evidence for the fourth account.
Target form. behaviour — Whether the presenting space admits a half-space decomposition on which conjugation induces a positive pairing — naming the half-space and the pairing, or showing no such decomposition exists.. Bound: not fired.
Bound note (added by hand — the v4_1 export dropped it; see differential/README.md): Bound not added — target is a behaviour question with no threshold in it — outside the number-kind taxonomy, Bound not applicable.
Ledger (Ledger)
TAKE
Two different epistemic grades are being treated as one register. The finite-place term is pinned by identity: |1−u|_p = max(1,|u|_p) for |u|_p ≠ 1 is an ultrametric fact, true by definition of the p-adic absolute value, so "exact with no correction" is not a discovery requiring four-prime confirmation — it is forced for every prime, tested or not. That line item is closed.
The archimedean constant γ+log2π is a different grade of object. It is currently pinned by only one computation (numerical match to six test functions, spread 0.002). No second, independent derivation is reported — e.g., recovering γ+log2π from the same digamma asymptotics by a route that doesn't already assume the trace formula's normalization. Until a second method reproduces that constant, it is a measured coincidence, not an identity.
The signature side does have a genuine two-way check: intersection-lattice signature ((1,3)/(2,2), fixed by the rank-4 structure) and the Weil-bound crossing at |a|=2√q are independent computations of the same flip, and they agree. That is what a resolved sign looks like in the ledger. The presenting object has no second column at all for its sign — nothing to check the missing positivity against — which is exactly the ledger's way of saying the "half-space" is currently unnamed, not merely unverified.
PREDICTION
Extending the finite-place check to p=11 and p=13 will continue to give the identity exactly (weight p^(−k/2), zero correction term) — because it follows from |1−u|_p=max(1,|u|_p), not from the sample of primes tested.
ASSUMES
That the "exactness" claim is meant as a genuine per-prime identity rather than an empirical regularity awaiting exception; the case's phrasing ("verified with no failures at p=2,3,5,7") is compatible with treating it as still-open, so I assume the stronger identity reading.
CHANGES MY MIND
A single prime p where |1−u|_p ≠ max(1,|u|_p) for |u|_p≠1, or where the local integral picks up a nonzero (1−p^{−k}) term — that would mean the "identity" was never an identity, only a coincidence in the tested range, and would collapse this axis's main distinction.
Framing (Framing)
TAKE
The presenting object is named as "a space carrying a scaling flow" — a fixed geometric object that either does or doesn't have an ample class. But the only piece of evidence offered for how positivity actually behaves (the finite-field analogue) shows a signature that flips with a parameter: (1,3) versus (2,2) depending on whether |a| exceeds 2√q. That is not a structural invariant of a space, it is a property of a point in a family indexed by a — here, a Frobenius trace, whose archimedean analogue is a spectral/eigenvalue parameter (the analogue of a real part on a critical line). Asking "does the space have an ample class / a positive real form" treats a parameter-dependent condition as if it were a fixed structural fact about a single object. That's a category change hiding inside the framing: the sign is not a missing datum attached to "the space," it is a threshold condition on a family that the space merely indexes. The four accounts on the table (Riemann-Roch over F₁, foliated leafwise Hodge theory, span-of-correspondences, no-import-possible) are all still asking "does the space have X" rather than "for which values of the spectral parameter does the induced pairing on the half-space become positive." The object the case should be about is that threshold — the analogue of 2√q — not "the space."
PREDICTION
If the analogy is taken literally, the positivity condition will fail to hold uniformly and instead flip sign at a specific value of the archimedean spectral parameter (real part = 1/2 in the trace-formula normalization), not hold or fail globally for "the space" as a whole.
ASSUMES
That the archimedean local term's variable (the digamma-form parameter) is the correct continuum analogue of the Frobenius trace a in the finite-field case.
CHANGES MY MIND
A demonstration that the positivity condition, once found, is constant across the entire relevant parameter range (no threshold, no flip) — i.e., that it really is a fixed attribute of the space and not a parameter-indexed condition.
Hypothesis (Hypothesis)
TOOL
Lefschetz trace/fixed-point formula (Atiyah–Bott shape) applied to the scaling-flow space
REQUIRES
- A compact (or suitably finite-dimensional, cohomologically bounded) space with well-defined Lefschetz numbers
- A cohomology theory with hard Lefschetz / a polarization (ample class + Hodge–Riemann bilinear relations) supplying signs on each graded piece, so that the "trace" can be interpreted as an alternating sum with definite-sign contributions
- The fixed-point data (local integrals) to converge to and equal global topological invariants, which requires the positivity/sign structure to assign each term to the correct graded piece with correct multiplicity
SATISFIED
- Existence of trace formula, local integral shape: yes — measured and verified at finite and archimedean places
- Hard-Lefschetz / polarization structure: no — case states explicitly "no ample class and no hard-Lefschetz sl₂"
- Positivity assigning correct sign to each term: no — this is exactly the "missing object," not a missing lemma
TOOL
Signature invariance under complexification (used to argue the sign cannot be recovered algebraically)
REQUIRES
- The claim rests on: nondegenerate symmetric bilinear forms over ℂ are classified only by rank (no finer invariant)
- Real forms of a given complex form are not unique / not determined by complexification alone
SATISFIED
- "All nondegenerate symmetric forms over ℂ are equivalent": yes — standard linear algebra, correctly invoked, no hidden assumption
- "A real form is not determined by its complexification": yes as a general fact, but its relevance here requires that the presenting space's pairing is genuinely a complexification awaiting a real structure — this is asserted, not derived, and is flagged in the ledger as [assumed, LOAD-BEARING]. Not independently checked against the actual algebraic structure of the scaling-flow space.
TOOL
Reflection positivity / Osterwalder–Schrader reconstruction (analogy import)
REQUIRES
- A reflection operator on the space
- Positivity of the pairing induced by that reflection on a half-space decomposition (not mere existence of the reflection)
- The half-space and the induced pairing must be explicitly identifiable to make the positivity claim checkable independently
SATISFIED
- Reflection exists (conjugation): yes — stated directly
- Half-space identified: no — case explicitly says a satisfactory answer "must name the half-space," and no candidate half-space or induced pairing is given for any of the four accounts
- Positivity of induced pairing: not checkable from the case — the entire point of the friction section is that this is unresolved
TAKE
Every tool that is actually verified in this case (trace formula, finite-place exactness, archimedean constant) verifies a statement about magnitudes, not about signs. The one tool whose preconditions would supply the sign — hard Lefschetz / polarization — is explicitly absent, and the case is honest about that. The analogy to reflection positivity is doing real rhetorical work but is not itself a tool applied to this space: no half-space has been named, no induced pairing computed, so OS reconstruction is not "invoked and checked," it is invoked as a hoped-for template. The finite-field analogue (signature flip at |a|=2√q) is the strongest piece of evidence in the case, but its use as a "faithful analogue" of the infinite-place object is asserted, not argued — nothing shown here demonstrates that the archimedean/adelic space carries a lattice structure homologous to the surface's rank-4 intersection form. So the four accounts on the table are not competing solutions to a solved problem; they are four different guesses at what precondition-satisfying structure would need to exist, and none of them has yet named its half-space, its ample class, or its foliation with an actual verified positivity computation. The honest state is: zero of the needed tools have satisfied preconditions yet.
PREDICTION
If a half-space decomposition is eventually named, the induced pairing will fail positivity at exactly the same threshold analogous to |a| = 2√q — i.e., some parameter analogue will show a sign flip, not uniform positivity, meaning the "import" will only be partial (positive on a sub-cone, not the whole space).
ASSUMES
That the finite-field analogy's sign-flip behavior is a real structural constraint on any correct answer, not an artifact of that specific surface's rank-4 lattice — the case does not establish this generalizes.
CHANGES MY MIND
An explicit construction of a half-space with a computed, everywhere-positive induced pairing (checked, per the ledger's own stated criterion, independently of the Lefschetz conclusion it would support) would satisfy the missing precondition and retract this take.
Lateral (Lateral)
TAKE
The case already names the shared theorem-candidate correctly, and my job is to check whether the cross-reference to Osterwalder–Schrander (OS) is doing real work or just decoration. It's decoration as stated, because the case never names the half-space or the induced pairing for the presenting object — it only asserts the analogy holds. OS reconstruction is not "there exists a reflection plus a vague positivity requirement"; it is a specific theorem: positivity of the inner product ⟨θF, F⟩ on the forward half-space of Euclidean time, where θ is time reflection and F ranges over observables supported on that half. The theorem gives a Hilbert space and a self-adjoint Hamiltonian. If the presenting space's conjugation is to play θ, someone must exhibit (a) the "time" coordinate being reflected, (b) the "forward half-space" subalgebra, and (c) the sesquilinear form ⟨θf, f⟩ built from the existing pairing, then show positivity is equivalent to something already measured (e.g., the exact finite-place weights, or the archimedean constant). None of that is done here. The four accounts are not on equal footing with OS-as-analogy until one of them produces this triple. Absent that, "same structure" is a resemblance, not a shared theorem, and should be labeled as such.
PREDICTION
If the half-space and induced pairing are ever exhibited, the positivity condition will fail exactly at the signature-flip locus |a| = 2√q's real-place analogue — i.e., the induced form will have a negative eigenvalue whose sign flips as the archimedean parameter crosses a value within 5% of the constant γ + log 2π already measured, not at an unrelated value.
ASSUMES
That "half-space" must mean a genuine self-adjoint-generating subalgebra decomposition (Haag-duality-type split), not merely a fixed-point set of conjugation.
CHANGES MY MIND
An explicit construction naming the half-space subalgebra and showing ⟨θf,f⟩ ≥ 0 reduces to (or is equivalent to) one of the already-verified identities (the exact finite-place weight or the digamma constant) — that would upgrade the analogy to a shared theorem.
VERDICT
Weight falls with Hypothesis and Lateral, jointly: the tool the disputed claim depends on has an unmet precondition, and it is unmet in exactly the place the case says the answer must live. Osterwalder–Schrader is not "reflection exists plus positivity hoped for"; it is positivity of ⟨θF,F⟩ on a named forward half-space subalgebra. Conjugation supplies θ and nothing else. Until the half-space and the induced sesquilinear form are exhibited, the OS analogy is a resemblance, not a shared theorem, and none of the four accounts is a candidate solution — they are four names for the same unnamed object. This outranks the four-way mechanism dispute, which cannot be adjudicated before the precondition is met. Ledger's contribution is real but narrower: it closes one flagged assumption (finite-place exactness is an ultrametric identity, forced for every prime, not an empirical regularity) and correctly downgrades γ + log 2π to a single-method measurement. Framing's reframing is the strongest live rival and is tested below.
DISAGREEMENT
Two conflicts, one of them with the case's own framing.
Primary: the case presents the missing positivity as a missing object attached to a fixed space — "the space has no ample class" — and then enumerates four geometries that might supply it. Framing denies the category: in the only worked instance offered, the signature is not an invariant of a surface but a threshold condition on a family, flipping at |a| = 2√q. If that is faithful, the correct question is "for which values of the spectral parameter is the induced pairing positive," and all four accounts are malformed because all four ask "does the space have X."
Secondary, and outranked: which of the four geometries (F₁ Riemann–Roch, foliated leafwise Hodge, ample class on the correspondence span, no-import-possible) supplies the sign. This is a mechanism dispute downstream of a precondition nobody has met, and it cannot be scored now.
The case's fourth account is also not symmetric with the other three: "no import possible" is the conclusion the case invites if the half-space cannot be named, so the enumeration smuggles in its own verdict condition.
FORCED
Forced by structure:
- |1−u|_p = max(1,|u|_p) for |u|_p ≠ 1, hence weight exactly p^{−k/2} with no (1−p^{−k}) correction. This is the ultrametric triangle equality, true per prime by definition. The "no failures at p = 2,3,5,7" framing is misleading — the count of four primes is decorative, not evidential. The audit's "unclassifiable: absence of correction term" is misfiled; a route exists and Ledger named it.
- The exponent 1/2 in |u|^{1/2} and in p^{−k/2}: fixed by half-density scaling normalisation, not fitted.
- 2√q: the Weil bound, forced by weight-one scaling.
- Signature entries (1,3)/(2,2): forced by the rank-4 intersection lattice on a product surface with h = f₁ + f₂.
- Over ℂ all nondegenerate symmetric forms of equal rank are equivalent; signatures exist only over ℝ. Correctly invoked, no hidden premise.
Fitted, inherited, or artefact:
- γ + log 2π as universal: inherited from one measurement across six test functions, spread 0.002. Universality is asserted from a sample, not derived. Single-column entry in the ledger.
- Eigenvalues −0.033 (a = 5) and +0.101 (a = 6): measured values of one specific form; a = 5, 6 chosen to straddle 2√q. They confirm the flip, they do not establish the flip's generality.
- "The presenting pairing is a complexification awaiting a real structure": asserted. The general fact (real forms not determined by complexification) is forced; its applicability here is not.
- Exhaustiveness and mutual exclusivity of the four accounts: fitted enumeration, shown nowhere.
- Eighty-year literature: rhetoric, no structural role.
TESTS
RESIDUE
If test 2 returns no conjugation-invariant half-space decomposition, the case does not close — it converts. The sharper case is then: state precisely what fails. Is there no "time" direction to reflect (the scaling flow is one-parameter and its orbits are already the reflected coordinate, so θ acts within rather than across), or is there a splitting with no positive form on it? These are different negative results. The first says the OS template was never applicable and the eighty-year analogy was load-bearing rhetoric; the second says positivity is genuinely unavailable and supports account 4. The case currently treats both as "no import possible" and they are not the same claim.
If test 3 returns a flip, the residue is sharper still: name the threshold and ask what it is the Weil bound of. That is a better object than any of the four on the table.
No tests run. Panel output recorded at House 1 as a smoke test of the tool.
Tests 1-6 ordered and unrun.
Ledger
TAKE
No tests were run, so nothing in my ledger changes — I defend it as filed. The House correctly used my finding (finite-place exactness is a forced ultrametric identity, not a four-prime sample) to sharpen the audit's misfiling, and correctly kept my downgrade of γ+log2π (single-method measurement, not yet a two-column check). Test 1 is the direct continuation of my axis: it asks for a second independent derivation of that constant, which is precisely "compute it two ways" — the thing I press on. Until that test runs, the correct bookkeeping status of γ+log2π is unchanged: pinned by exactly one computation, universality asserted from a sample of six, not established. I'll note one thing the House did not flag from my seat: test 2 (exhibit the half-space) is not on my axis and I take no position on its likely outcome — that's Hypothesis/Framing territory, structure not identity. My axis has purchase only on which quantities are forced-by-definition versus fitted-from-sample, and on that ledger nothing here moves without new computation.
PREDICTION
When test 1 is actually run — deriving γ+log2π from digamma asymptotics by a route independent of the trace formula's normalisation — it will reproduce the constant only to a residual larger than 0.002 (i.e., it will not close tighter than the existing six-test-function spread), because the six test functions likely share a common normalisation convention with the asymptotic route rather than being truly independent checks.
ASSUMES
That "independent derivation" in test 1 can actually be constructed without smuggling in the same normalisation choice that produced the original match — this is not guaranteed and the case doesn't specify a candidate second route.
CHANGES MY MIND
If test 1 is run and reproduces γ+log2π exactly (residual at or below floating-point/measurement precision, not just within 0.002) via a genuinely normalisation-independent route, that promotes the constant from single-column measurement to a two-way-computed identity, and I withdraw the downgrade.
Framing
TAKE
No tests ran, so nothing empirical has moved. I defend the take unchanged and sharpen it against House's synthesis. House folds my point into test 3 as one possible outcome ("Lateral's sharper version… is a bonus, not the discriminant") — that undersells it. My claim isn't a hypothesis to be adjudicated alongside the OS-half-space question; it's prior to it. Test 2 asks "does a conjugation-invariant half-space exist." But the only reason we believe positivity is the kind of thing that could live on a half-space at all is the finite-field picture — and that picture is explicitly parameter-indexed (flips at |a| = 2√q). If the archimedean analogue inherits that structure, then "exhibit the half-space and its induced form" is itself mis-scoped: there may be a family of half-spaces, or a half-space whose induced form's sign depends on where in the spectral parameter you sit. House's test 2, run alone, could report "no invariant splitting" and everyone would take that as evidence for account 4 (no import possible), when the real diagnosis is that they searched for a parameter-independent object. I want test 2 gated: don't run it without simultaneously running test 3's sweep, because a null result on 2 without 3 is uninterpretable — it can't distinguish "no half-space exists" from "the half-space exists but only supports positivity in part of the range."
PREDICTION
If test 2 is run first in isolation and returns null, and only later paired with test 3, test 3 will reveal that a half-space decomposition does exist but its induced form is positive only for spectral parameter values on one side of a threshold — i.e., test 2 alone will produce a false negative for "no import possible."
ASSUMES
That the panel would otherwise sequence test 2 to completion before running test 3, rather than jointly.
CHANGES MY MIND
Test 2 and test 3 run jointly, with test 3 showing the induced form's sign is constant (no threshold) across the full swept range — that would confirm House's framing (positivity as a fixed attribute) over mine.
Hypothesis
TAKE
No tests were run — the LAB REPORT is empty. My round-1 take stands unmodified, and I decline to soften it toward the accounts on the table. The House's VERDICT and test list confirm rather than challenge my axis: Test 2 (exhibit the candidate half-space, compute ⟨θf,f⟩ on a finite-dimensional truncation) is precisely the missing-precondition check I flagged for the OS-analogy tool, and it is still unrun. Nothing in the House's synthesis discharges any of the three unmet preconditions I listed: hard-Lefschetz/polarization is still absent, the "complexification awaiting a real structure" premise is still asserted not derived (Test 4 targets exactly this and is unrun), and the half-space/induced-pairing is still unnamed (Test 2, unrun). I'll note one refinement worth crediting to Ledger/Forced: the finite-place exactness is now correctly reclassified as structurally forced (ultrametric identity) rather than an empirically-verified regularity — that retires one item from my "not checkable" column into "yes, satisfied," but it was never the load-bearing precondition in dispute. The load-bearing precondition — positivity of an induced pairing on a named half-space — remains exactly as unmet as at Round 1. A result cannot be said to have moved closer to being established by the mere act of ordering tests; only running them moves it. Zero tools have satisfied preconditions. That sentence from Round 1 is still the accurate summary.
PREDICTION
When Test 2 is actually run, it will fail to produce a conjugation-invariant half-space decomposition at all — not merely fail positivity — consistent with the "first" branch of RESIDUE (the scaling flow's orbits already contain the reflected coordinate, so θ acts within rather than across, and no genuine half-space splitting exists to test for positivity).
ASSUMES
That a one-parameter scaling flow, absent any stated second (time-like) direction, has no natural transverse coordinate for a reflection to split — this is a structural guess about the object, not stated in the case, and is exactly what Test 4/Test 2 would settle.
CHANGES MY MIND
Any of: (a) Test 2 actually run and returning a named half-space with a computed nonnegative form, or (b) Test 4 showing the pairing is a genuine complexification with a determinate real structure to descend from. Absent either, my take does not move.
Lateral
TAKE
No lab work happened — the tests I named (specifically Test 2, "exhibit the candidate half-space") were ordered, not run, so my round-1 take is unchanged in substance but its status shifts: it's now the unmet precondition confirmed as unmet, not just flagged. I want to sharpen one thing House's verdict got right and I should have stressed harder: the four-account enumeration is not damaged by absence of data, it's damaged by structure. All four accounts presuppose a decomposition (half-space, subalgebra, spectral region) exists to be identified; the exercise of naming which geometry supplies positivity cannot even start before Test 2 returns something. So the correct move right now is not "which account looks likely" but "downgrade every account to unranked until Test 2 runs." I'll also flag that Test 1 (independent re-derivation of γ + log 2π) is cheap and should have been run before Test 2 — it's a pure computation, no construction required — and its absence is a missed opportunity, not a limitation of the domain. My axis has purchase here precisely because the whole dispute is "is this the same theorem" and the answer is still: unverified, because the object the theorem needs has not been exhibited, only gestured at via the finite-field analogy.
PREDICTION
Same as before, now conditional on Test 2 actually running: if a half-space decomposition is exhibited, the induced form's sign will flip within 5% of γ + log 2π on the archimedean parameter sweep (Test 3), not at an unrelated value. Unchanged since no data arrived to update it.
ASSUMES
That "no tests run" means the epistemic status of the OS-analogy claim is exactly where it was at round 1 — i.e., silence is not evidence either way. Also assumes Test 1 and Test 2 are independent enough that running one doesn't presuppose the other's result.
CHANGES MY MIND
Test 2 actually returning a named subalgebra and computed ⟨θf,f⟩ — positive, negative, or absent — is the only thing that moves me; a smoke test confirming the panel ran is not that.
VERDICT
Nothing was run, so the verdict cannot move on evidence — but it moves on sequencing, and that is Framing's point, correctly made. The unmet precondition still outranks everything: OS reconstruction requires positivity of ⟨θF,F⟩ on a named forward half-space subalgebra, and only θ (conjugation) is in hand. That precondition is load-bearing for the claim actually in dispute — that some geometry supplies the sign — so it outranks the four-way mechanism question, which remains unrankable. But Framing has shown my Round-1 ordering was defective. Test 2 run alone admits a false negative: "no invariant splitting found" would be read as support for account 4 when it may only mean the search was for a parameter-independent object the finite-field analogue says does not exist. Tests 2 and 3 are one test and are now gated together. Lateral is also right that Test 1 is cheap and should have gone first; that was an ordering error, not a domain limit.
DISAGREEMENT
Primary, unchanged and still with the case's own framing: the case asks "does the space have an ample class / a positive real form," treating positivity as a fixed attribute of a named object. The only worked instance offered — the finite-field signature flip at |a| = 2√q — shows positivity as a threshold condition on a parameter-indexed family. If that analogue is faithful, all four accounts are malformed, because all four ask "does the space have X" rather than "on which side of which threshold."
New this round, and procedural rather than substantive: Hypothesis predicts Test 2 returns no conjugation-invariant splitting at all (θ acts along the scaling orbits, not across them). Framing predicts that exact result, obtained in isolation, is a false negative. These two are not rival hypotheses about the world; they are rival readings of the same anticipated null. That is decisive for sequencing and I side with Framing: an uninterpretable null is not a result.
Outranked: which of the four geometries supplies the sign. Still unscoreable.
FORCED
Forced:
- |1−u|_p = max(1,|u|_p) for |u|_p ≠ 1, hence weight exactly p^{−k/2}, no (1−p^{−k}) correction. Ultrametric triangle equality, per prime, by definition. The four-prime count is decorative. Premise 1 is retired, not tested.
- Exponent 1/2 in |u|^{1/2} and p^{−k/2}: half-density scaling normalisation.
- |a| = 2√q: Weil bound, weight-one scaling.
- Signatures (1,3)/(2,2): rank-4 intersection lattice on a product surface with h = f₁ + f₂.
- Over ℂ, nondegenerate symmetric forms of equal rank are equivalent; signatures exist only over ℝ.
- OS is positivity of ⟨θF,F⟩ on a half-space, not existence of θ. Lateral stated the theorem correctly; premise 11 is retired on the statement of the theorem, not on its applicability here.
Fitted, inherited, or artefact:
- γ + log 2π as universal: one measurement, six test functions, spread 0.002. Single column.
- Eigenvalues −0.033, +0.101 at a = 5, 6: measured values of one form, at points chosen to straddle 2√q. They confirm a flip; they do not generalise it.
- "The presenting pairing is a complexification awaiting a real structure": asserted. The general fact is forced; its applicability is not.
- Exhaustiveness and mutual exclusivity of the four accounts: fitted enumeration. Account 4 is additionally not symmetric — it is the case's own verdict condition wearing a hypothesis's coat.
- Eighty-year literature: rhetoric.
TESTS
RESIDUE
Unchanged and now sharper by one fork. If Test 2 returns no splitting at any parameter value, the case converts rather than closes, and the conversion must distinguish two negative results the case currently conflates: there is no direction transverse to the scaling orbits for θ to reflect across, versus there is a splitting but no positive form on it. The first retires the OS analogy as load-bearing rhetoric; the second is substantive support for account 4. If Test 2 returns a flip, the residue is the threshold itself: name it, and ask what it is the Weil bound of. That is a better object than any of the four on the table.
BACKPROP
open.
Write back an edge: Test 2 and Test 3 of Round 1 are not independent tests and must never be run in isolation; a null on the existence of a decomposition is uninterpretable without the parameter sweep that would distinguish "absent" from "present but restricted." Record this as a general controller fault — when a hypothesis says the target is parameter-indexed, existence tests must be run parameterised or not at all.
Write back a correction: my Round-1 ordering put a construction (Test 2) ahead of a pure computation (Test 1). Cheapest-decisive-first was violated; Lateral caught it.
Write back a probe: premise 6 — that a Lefschetz-trace argument needs a positivity input at all — was flagged load-bearing and went untested for two rounds while the whole panel argued about where positivity would come from. It is now Test 5.
Next case: the threshold object. Take the finite-field flip at |a| = 2√q as the presenting object in its own right and ask what conserved quantity the crossing is the saturation of, independent of any surface. If that case resolves, it tells this one whether it is looking for a geometry or for a bound.
generated 2026-10-25the lab index