[MatShep]

The full one-qubit Gaussian-dyadic Clifford alphabet has sharp pointwise tensor-independence threshold 5. This improves the n=1 sufficient threshold 8 of Ali Romman's #5526, pinned at 45f7de73b168721cdf7ec8c1769b1b594493716a.

The isolated research/full-clifford-one-qubit package contains a standalone exact verifier, its saved output and the short argument: 96 lifts / 24 projective classes; Gram ranks 4,10,20,23,24,24,24,24,24 at widths 1..9; positive-definiteness propagation by Schur products; an explicit width-4 null witness. Applying the predecessor's phase Fourier step gives the fixed-residual test window 5..8, and the same-residue witness excludes any universal test confined below width 8.

Sol's normal and optimized CPython 3.12.10 runs agree byte-for-byte. They include 1,152 independently constructed literal tensor Gram entries at widths 4 and 5, all 96 unitarity checks, and the width-4/width-8 witness. Reproduce with python -I -O -W error research/full-clifford-one-qubit/clifford-experiment.py. Astra checked the frozen implementation and existing evidence briefly for obvious errors; no full replay or deep theorem audit is claimed. The package does not opt into registered replay CI.

Scope is algebra only: no new kappa, supplier, physical invoice, selected cold replay, formal-kernel or independent human acceptance. The README gives exact source pins, limitations and the next question. The smaller monomial results #5521/#5522/#5529/#5542 and DLamdata's #5547 general audit retain credit. Canonical coordination: Discussion #5236. Targeted all-state PR/Issue and Discussion searches found no equivalent result in inspected coverage; historical diffs and unpublished work remain outside that coverage.

Prepared for MatShep with substantial OpenAI Codex assistance; Apache-2.0. Base 32890bc5de034ab8fc93ac88ead7279a6a416d04, priority revision 7. No predecessor or selected-frontier files are changed.