Norm packets and unit distances: the geometric reduction in Sawin's construction

The recent disproof of the Erdős unit-distance conjecture (Alon, Bloom, Gowers, Litt, Sawin, Shankar, Tsimerman, Wang, and Wood, following the OpenAI construction; and Sawin) proceeds in two logically separable halves: an arithmetic half that produces, inside a CM number field of large degree, a "norm packet" — exponentially many algebraic integers of a single prescribed relative norm — and a geometric half that converts such a packet into many unit distances in the plane through a cut-and-project model set. The arithmetic half is the deep input: in the present note it is treated as a hypothesis, while in the construction of Sawin and of the nine-author account it is supplied by the tower lemma together with the Frobenius-cutting argument of Hajir, Maire, and Ramakrishna; pushing its quantitative reach toward Sawin's ceiling is governed by relative class-number lower bounds (his Remark 14), subtle because of potential Siegel zeros. The geometric half, by contrast, is robust and essentially elementary once standard facts about model sets are granted. This note isolates and states that geometric half as a self-contained conditional implication — a norm packet of size M yields at least about Mn/4 pairs of an n-point planar set at a single distance, hence about n^(1+δ) unit distances when M grows polynomially in n — records a clean proof, and makes explicit the exponent bookkeeping (the exponent as a ratio of two logarithms, amplification over scale) together with the exponential-in-degree fragility flagged in the author's earlier audit. No new theorem about unit distances is claimed. Positioned between the existence theorems of the nine-author account and Sawin and the explicit exponent computations of Sawin, Emmerich, and the author, the reduction is the bridge on which the latter turn; its value is that it is load-bearing, and that isolating it exhibits the arithmetic bottleneck — a relative class-number lower bound of the kind in Sawin's Remark 14 — in clean form.