The d'Alembert Inevitability Theorem

strengths

• Clean conceptual reduction: symmetry + quadratic degree bound on P forces P(u,v) = 2u + 2v + c·uv, which after F → F∘exp is the textbook d'Alembert equation with fully classified continuous solutions.
• The degree-≥3 exclusion (Thm 3.1), if rigorously established with a sharp non-cancellation hypothesis, is a genuinely new rigidity statement complementing classical Aczél-type quadratic results.
• The cost-function pipeline (F ≥ 0, convexity ⇒ hyperbolic branch with c > 0) is a clean way to break the trichotomy without ad hoc input.
• The multivariate corollary — collapse to a single log-linear direction for c ≠ 0, general quadratic form for c = 0, and exclusion of nontrivial coordinate-wise separable costs — is a strong and falsifiable structural statement.
• If correct, the paper supplies the functional-equations lemma that downstream foundational programs (including the Pith canon) currently postulate as Level-1 input.

major comments

1. Theorem 3.1 (degree-mismatch exclusion). The proviso 'provided the leading term does not cancel' is load-bearing and the abstract leaves it informal. Symmetric polynomials of degree ≥3 in two variables have a multi-dimensional space of leading homogeneous parts. Please state explicitly (a) which degree is meant — total degree in (u,v), partial degree, or degree on the diagonal u = v; (b) what 'cancel' means after substitution u = F(x), v = F(y) — does it mean the leading symmetric form vanishes identically, vanishes on the image of (F,F), or vanishes only on the diagonal? — and (c) at least one explicit example of a degree-3 symmetric P whose leading term does cancel and which therefore escapes the exclusion, to demonstrate the hypothesis is sharp rather than vacuous. Without this, the scope of Thm 3.1 cannot be assessed.
2. Theorem 3.3 (uniqueness of P up to degree 2). Make explicit which inputs are essential: F(1) = 0 presumably kills the constant term in P; symmetry plus the diagonal identity F(x²) = P(F(x),F(x)) = 4F(x) + c·F(x)² presumably pins coefficients; and nonconstancy is needed to avoid the degenerate case P(u,v) = u + v with F ≡ 0. Please also state whether continuity is essential or whether measurability suffices via the standard d'Alembert argument, and add a short remark contrasting the result with the Aczél/Kuczma classification of d'Alembert-type equations so the reader can locate the contribution within the literature.
3. Calibration to c = 2 and the 'canonical' cost. The phrase 'unit log-curvature calibration selects the canonical value c = 2' should be sharpened. The two-parameter family F_α(x) = (1/c)(x^α + x^{-α} − 2) admits rescalings of F and of the argument that trade c against α, so c = 2 with α = 1 looks like a normalization rather than a derivation unless an extra invariance (e.g. F''(1) = 1 in log coordinates *together with* a fixed ambient scale on x) is imposed. Please state the calibration condition as a precise equation, and justify why it is canonical rather than conventional. This matters because the Pith canon and other downstream programs treat F(x) = ½(x + 1/x) − 1 as uniquely determined.
   • all constants from phisupports Downstream canon uses precisely the c = 2, α = 1 cost; making the calibration step rigorous closes the loop with the canon's Level-1 input.
4. Multivariate extension on ℝ_{>0}^n. The n-variable composition law should be written as a numbered equation early in the section: it is genuinely ambiguous what equation is being assumed (componentwise pairwise relations? a single n-ary law?). The asymmetry between c ≠ 0 (collapse to a single log-linear direction) and c = 0 (arbitrary quadratic form Σ a_{ij} log x_i log x_j) is striking and should be commented on explicitly, with at least one explicit non-collapsing example at c = 0 to show the dichotomy is genuine. The compatibility/coherence conditions used to glue per-pair d'Alembert rigidities into the n-variable conclusion should be stated.
5. Convexity hypothesis. Specify whether 'convexity' means convexity of F on ℝ_{>0} or convexity of G = F ∘ exp on ℝ. These are not equivalent, and only one of them yields the cosh-branch selection cleanly. Stating the precise convexity hypothesis is important because it is the step that breaks the c ≠ 0 trichotomy and rules out the trigonometric branch.

minor comments

• Title. 'Inevitability Theorem' is dramatic phrasing for what is mathematically a rigidity/classification theorem in the Aczél tradition. A more conventional title ('rigidity' or 'classification') would also help functional-equations readers find the paper.
• Abstract. State the domain of P explicitly. 'Polynomial in (u,v) over ℝ' is presumably intended, but a reader could read 'symmetric polynomial combiner' as allowing real-analytic combiners; the distinction matters for Thm 3.1.
• Theorem 3.3. Spell out the role of F(1) = 0 (kills affine constant in P) and clarify whether the result is invariant under F ↦ F + const after appropriate rescaling of P.
• Notation. If F is sometimes treated through F ∘ exp on ℝ, introduce a distinct symbol (e.g. G(s) = F(e^s)) and use it consistently; otherwise the proofs become harder to follow.
• References. Cite the standard references on the d'Alembert equation (Aczél, *Lectures on Functional Equations*; Kannappan, *Functional Equations and Inequalities with Applications*) so the reader can locate the classical hyperbolic/trigonometric branch theorem invoked in the reduction.
• Multivariate section. State the n-variable composition law as a numbered equation at the start of the section to remove the ambiguity left by the abstract.
• Connection to downstream programs. A short remark connecting the canonical cost F(x) = ½(x + 1/x) − 1 to the Pith canon's use of this cost as a Level-1 input would help orient readers coming from the Recognition-Science literature.

where the referees disagreed

• Canon match strength

  Referee A: Partial / tangential: the canon assumes the cost J(x) = ½(x+1/x) − 1 as Level-1 input, and this paper purports to derive it.

  Referee B: None: treats the paper as pure functional equations with no canon overlap.

  synthesizer: Side with Opus. The canon does not prove the paper's theorem, but the paper's headline output is exactly the canon's Level-1 cost primitive. That is a real, if non-circular, alignment, best classified as 'partial'.

• Final recommendation

  Referee A: Uncertain (low confidence) due to four substantive concerns about hypotheses and calibration.

  Referee B: Accept (moderate confidence); no major comments.

  synthesizer: Major revision. Opus's concerns about Thm 3.1's non-cancellation proviso, Thm 3.3's exact hypotheses, the c = 2 calibration step, and the n-variable formulation are concrete and need to be addressed in the body before acceptance. Grok's confidence that 'this looks like a standard classification theorem' is reasonable but does not engage these specific load-bearing hypotheses, so cannot dominate.

• Whether the c = 2 selection is a derivation or a normalization

  Referee A: Calls it a normalization unless an extra invariance is supplied; demands a precise calibration equation.

  Referee B: Accepts the log-curvature calibration as derivational without comment.

  synthesizer: Side with Opus. The two-parameter family (c, α) genuinely admits rescalings that trade the parameters; the paper must state explicitly which independent scale fixes α before c = 2 can be called canonical.

• Need for an explicit non-cancelling example for the degree-≥3 exclusion

  Referee A: Required; without it the proviso may be vacuous.

  Referee B: Did not raise the issue.

  synthesizer: Side with Opus. Producing one explicit symmetric degree-3 P with cancelling leading term that admits a nonconstant continuous solution would establish that the exclusion is sharp; absent that, the strength of Thm 3.1 is unclear.

how each referee voted

Opus voted uncertain / low confidence, raising four substantive concerns about the precise hypotheses behind Theorems 3.1 and 3.3, the calibration that selects c = 2, and the multivariate extension. Grok voted accept / moderate confidence, treating the paper as a clean classification theorem and listing no major comments. The synthesizer sides closer to Opus on the technical concerns — in particular the load-bearing "leading term does not cancel" proviso in Thm 3.1 and the calibration-vs-derivation question for c = 2 — while agreeing with Grok that the overall mathematical content is substantive and well-motivated. Net call: major revision, low confidence, since the abstract alone does not let the synthesizer verify how the load-bearing hypotheses are formalized.