Answers are grounded in this paper and cite their sources.
v2 (18 Nov 2026, refereed revision): Corrects the Hessian normalisation (at the hexagonal point ρ the eigenvalues are +7.711 V₀, at τ = i V = 0.1958 V₀ with eigenvalues −2.216 and +10.012 V₀; the minimum-at-ρ / saddle-at-i verdict is unchanged), states the scale-flip rule as a known Poisson-summation result with prior art (Cohn–Kumar–Schürmann, Montgomery, Sarnak–Strömbergsson), and narrows the flow claim to the tested starts. License (from this version): CC BY-NC-SA 4.0 under the AEGIS Community License — free for individuals, scientists and students; any business or for-profit use needs the AEGIS Commercial License (https://github.com/dannyeldridge54/AEGIS/blob/master/licensing/COMMERCIAL_LICENSE.md). Built by one independent researcher with the help of AI. SERAPIS-E is a modular plateau inflation model whose potential is built from the Eisenstein series E(τ,2) of a torus shape τ. This note asks why its vacuum sits at the hexagonal shape τ=ρ=e2πi/3, and whether that choice survives changes of scale and dimension. Vacuum. The potential is monotone in E(τ,2), so it inherits E's critical points: the hexagon is an isotropic minimum (Hessian [+3.562, +3.562] V0); the square τ=i is a de Sitter saddle, unstable only along the axion ([−1.279, +5.779] V0). A τ=i vacuum needs extra physics. Scale. For the Gaussian lattice energy at fixed covolume the hexagon wins at every scale, and a torus started at any shape flows to it. Dimension: a flip rule from Poisson duality. The best lattice changes with scale if and only if it is not its own dual, and then it swaps with its dual exactly at the self-dual scale. Tested for d=2–8: hexagon, D4 and E8 never flip; FCC/BCC, D5/D5*, E6/E6* and E7/E7* swap. The E6 and E7 outcomes were predicted before they were computed. String lattices. Toroidal compactifications are even self-dual Narain lattices and T-duality is the same Poisson map, so they never flip. The c=2 Narain lattice with the largest spectral gap (SU(3) point; Angelinos–Chakraborty–Dymarsky, arXiv:2206.14825) has the hexagonal shape of the SERAPIS-E vacuum but sits at the string scale, while SERAPIS-E needs Im S·Im ρ ≈ 780: a match on shape only, recorded as such. No new prediction for the sky; the frozen predictions and the audited point (n_s = 0.97439, r = 4.82e-4) are unchanged. Open question: is there a dynamical mechanism that drives compactifications to maximum-gap points? Methods. Exact lattice enumeration (Fincke–Pohst, |v|²≤12, converged to 3e-11), Epstein zeta sums, hyperbolic gradient flow, Hessians from the model's own code. Scripts and outputs: github.com/dannyeldridge54/aegis (field_tests/results, notes D1276–D1283). Calculations were done with an AI assistant (Claude, Anthropic) under the pre-registration workflow of doi:10.5281/zenodo.24291026.
No takes yet. Share an insight, caveat, or question.
Danny Eldridge (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: