Target cell: weight-8 × unrestricted. The paper's headline code: a weight-8 BB-type code whose distance strictly exceeds the BB benchmark's 12 at the same length n = 144 — the board's weight-8 cell had no entry at this (n, k) with d >= 15 (all smaller-n codes with d >= 15 carry checks of weight 12-16, so they live outside this cell). Lu, Yang and Guo report it in Table 1, entry 9, with a certified lower bound d >= 15. This entry is a reproduction of that published code, not a new search.
No search. Generator polynomials transcribed from the paper's Table 1 and rebuilt with the repo kit: QC(A, B) on Z_8 x Z_9 with
(exponents written x^a y^b). The rebuild reproduced the paper's parameters: n = 144, k = 6 (recomputed by the verifier's rank arithmetic), max check weight 8. Notably the paper's subgroup-coset design rule (their Theorem 3.13) explains the high distance: neither generator is divisible by (1+x) or (1+y), so none of the cheap weight-l or weight-m coset logicals exist.
weight <= 14 and found no logical (certified d >= 15); the scan is translation-anchored for completeness (their Theorem 4.5) and cross-validated against Magma (their Remark 4.4).
side and weight-15 on the Z side, so the witness-backed claim is d <= 15. The validation gate (verify/validate_candidate.py) ran an 8000-trial fresh-seed RIS refutation: no lighter logical found; the gate reports the code as board-advancing in the weight-8 x unrestricted cell.
upper_bound, d <= 15. Source lower bound d >= 15(exhaustive, published) plus this witness pins the true distance at 15, but the d= tier on this board requires the server certifier, so the entry ships as an upper bound.
Not applicable to a reproduction. As with the paper's other Table 1 entries, most of its weight-8 census is dominated by existing board entries in this cell; this code and [[144,16,10]] (submitted separately, one code per PR) are the two that advance the board.
Human-authored source paper (Lu, Yang, Guo, arXiv:2609.06572); authorship of the code belongs to them — this entry is seeded as a baseline reproduction. Reproduction harness (by @MathysRennela): this repository's kit — research/kit/bb.py (build_bb), research/kit/css.py (k recomputation), research/kit/submit.py and the CLI ./qldpc submit (witness search, verification), verify/validate_candidate.py (gate).
Rebuild (H_X, H_Z) with the kit on the torus Z_8 x Z_9 (n = 2*8*9 = 144):
from bb import build_bb
HX, HZ = build_bb(8, 9,
A_terms=[(0,0),(4,2),(0,4),(0,7)],
B_terms=[(0,0),(2,6),(5,7),(7,8)])
Then ./qldpc submit <npz> --authors "Liangdong Lu" "Ruipan Yang" "Guanmin Guo" --anonymous --family bivariate-bicycle. The construction string in the JSON carries the same polynomials. Source: arXiv:2609.06572 (https://arxiv.org/abs/2609.06572), Table 1, entry 9.