Panel review: Tao’s uncertainty principle

Reviewed edition: Terence Tao, An uncertainty principle for cyclic groups of prime order, arXiv math/0308286v6, 22 July 2004. This comparison concerns this arXiv edition, not a separately inspected journal typesetting.

All eight independent responses are available (four models, two prompt profiles). This report is separate from reading.md; the reading packet contains no panel verdicts.

Substantiated local defects

All eight completed responses identify both of the following issues. Independent checks made before reading the responses support both; these checks have AI-source-adjudication origin, not human-grade origin.

1. Missing sign, Lemma minors, source line 258. For n=2 and frequencies 0 and 1, D=z₂−z₁, while the chosen Vandermonde factor is z₁−z₂, so P=−1. The Euler derivative at (1,1) is +1, contradicting the printed unsigned factorial identity. The general missing multiplier is (−1)^{n(n−1)/2}. It does not affect the nondivisibility modulo p needed for the lemma.

2. Max instead of min, Cauchy–Davenport proof, source line 316. The last displayed conclusion uses max(|A|+|B|−1,p), although the theorem statement and preceding derivation give min. For p=3 and A=B={0}, the printed max bound wrongly demands 1≥3. Replacing max with min restores the intended conclusion. No new mathematical idea is required. The incorrect max bound occurs only in the concluding sentence and is not used later in this manuscript. The following higher-dimensional remark invokes the earlier uncertainty theorem, not this sumset bound.

Compressed argument and external scope

The generic-linear-combination step can be expanded using finitely many proper complex hyperplanes: fix A′=A, let B′ range over subsets of B of size p+1−|A|, and choose the equality-case functions. Their spatial supports all equal A, and their frequency supports cover B. Each required coordinate is therefore a nonzero linear functional of the coefficients; a generic combination avoids their finitely many zero hyperplanes. This supplies the omitted routine explanation.

Several reports also comment on convolution normalization, conjugation of Fourier minors, abstract wording and the attributed higher-dimensional extension. Normalization and conjugation preserve the support/nonvanishing conclusions. The abstract's sparse-polynomial wording is ambiguous; the body explicitly evaluates at complex pth roots of unity. The attributed Meshulam extension is unproved in this manuscript and is not used in its principal argument; its external source was not inspected. These observations should not be merged with demonstrated errors in the main theorem.

Returned assessments

Every completed report judged the central argument sound after the local corrections. None reported a critical defect. This records the panel's assessments plus the limited source checks above, not an independent certification of every theorem.

Conservative token accounting for all eight requests: $1.781770, before Gemini promotional credits. No request remains pending. These are usage estimates, not invoices.

What this comparison establishes

The panel can identify small genuine defects in this human-authored source without mistaking them for a refutation of the central theorem. The available responses do not establish a unique Fable advantage on these two defects. One familiar manuscript cannot rank authorship categories or research-proof auditing models. The final Sol response confirms the two local defects and the distinction between the main proofs and the unverified attributed extension. Human quality grades remain pending.