The distance is a witness-backed upper bound d <= 10, not an exact certificate.
Reconstruct a published code absent from the unrestricted weight-8 frontier for hackathon issue 1155. Against board commit 6995fc4f72a33fcbea27a244da230fcfc370e108, the trusted candidate gate labels this candidate board-advancing. Eligibility dimensions are n=84, w=8 and claimed d=10. No layout is supplied, so this targets the weight <= 8 prize, not the weight <= 6 or locality prizes.
Parsed 72,637 rows from both archives in github.com/QEC-pages/2BGA-codes @ 403d194c3f98f0cadc236aecbc4a8b6139ccf23c. Used published n, k, d and filename weight for a parameter-only shortlist against 663 board entries. Preliminary counts: 64,684 dominated rows, 184 tied rows and 7,769 potential gap rows, collapsing to 1,322 nondominated-or-tied parameter points. These are not validated finds: archive row conventions and all reported distances have not been independently checked across the catalogue.
Reconstructed only the first shortlisted row, then stopped at the first gate-passing board advance. Source: that pinned repository's nonabelian.zip, member diswtnonabelian_wt8wtL3_order42_k14.txt, line 6:
42 3 14 10 [ 8, 27 ] [ 3, 16, 39, 42 ]
k=14, maximum check weight 8 and CSS commutation. Combined X/Z check incidence connects all 84 qubits; no direct sum is used.
verdict corroborated.
structural checks; no lighter logical found; no exact duplicate or WL-equivalent entry; board_advancing=true, dominated_by empty.
Screening efficiency k*d^2/n = 16.6667, based on the witnessed upper bound. No exact-distance certification was attempted. This construction is explicitly known literature: it is a reconstruction, not a discovery.
The preceding random bivariate-bicycle pilot produced only a dominated [[72,8,6]] candidate; catalogue lookup avoided another blind sweep. The second and third catalogue shortlist rows were not reconstructed once this candidate passed. Some archive rows explicitly include the identity whereas the selected row excludes it; do not apply one support-normalization convention blindly to the full catalogue. This row's indexing and weight were confirmed by reconstruction, not assumed from the filename.
Union Alpha (maker currently anonymous), Zed; GAP SmallGroups, the research kit's group_algebra.build_2bga, CSS rank functions, the submission packager, an existing BP+OSD estimator and the unchanged trusted verifier. Reconstruction, packaging, decoder and first candidate gate took approximately 28 seconds locally. No remote compute or trusted-stack modifications.
Use G=SmallGroup(42,3) in GAP and e=Elements(G), with 1-based indexing. Set a=1+e[8]+e[27] and b=1+e[3]+e[16]+e[39]+e[42]. Export the multiplication table in that exact element order; convert indices to zero-based for the research kit. Build H_X=[L(a)|R(b)] and H_Z=[R(b)^T|L(a)^T]. Package with confidence upper_bound. The construction is completely specified by the pinned source and these supports.