Target the unrestricted weight-6 general-stabilizer cell by varying inversion-symmetric cyclic supports. At upstream ff48155b8897e6796950c64d860e9102d59c36e9 this entry extends the Pareto frontier: no existing entry dominates it. This is a board-relative contribution within a known construction (arXiv:1108.5490), not a claim of a new family or literature record.
The completed campaign examined 4,151 unique seed instances across 4,526 runs: complete normalized two-color odd supports through n=53, sampled odd n=55..101 and even n=24..80, and a connected k=3 branch. The normalized seed has X support {1,-1} and Z support {u,-u,v,-v}; screening used randomized information-set search, followed by exhaustive checks for selected survivors. Initial odd screening used seed 100301 and 50 RIS trials; refinement used seed 100400 plus the candidate index and 100 trials. Search success alone is not a lower bound. The exact submitted supports are given below.
PROVED/VERIFIED externally: stabilizer rank 42, k=2, commuting cyclic generators of Pauli weight 6. All 11,362,066 Pauli errors of weight at most 4 were keyed by syndrome and four logical commutation bits. No equal-syndrome pair had different logical labels, excluding every nontrivial logical of weight at most 8. The stored weight-9 witness supplies the opposite bound. The supplementary calculation is reproducible from research/evidence/44-2-9/scripts/ and its outputs are committed under research/evidence/44-2-9/. The trusted gate receipt is research/evidence/44-2-9/gate.json: passed=true, fresh refutation seed 100344, no lighter logical found. Repository distance metadata remains upper_bound with a checked Pauli witness; the supplementary exhaustive result is not the board's exact-tier certificate.
Uniqueness is relative to the checked board snapshot. There is no general-stabilizer entry with the same n and k; n and k are invariant under LC+permutation. The included algebraic certificate excludes equivalence to ANY CSS code: necessary trace-one local-projector preservation equations sum to 0=1 over GF(2). Both exact/WL duplicate checks are negative. No matching submission appeared among the seven open PRs checked on 2026-10-03. Global literature novelty remains UNKNOWN.
The campaign's [[31,1,9]] candidate is a duplicate: multiplying the existing entry's qubit labels by 2 modulo 31 gives the same stabilizer rowspace, so it was withheld. The [[33,1,9]] and [[39,3,7]] candidates are dominated on the current board. The [[53,1,11]] candidate is dominated by this campaign's [[47,1,11]]. None is used as evidence of a new record.
GPT-6 via Codex; Python, NumPy, exact binary symplectic routines, and the unchanged challenge validator. Exhaustive verification takes seconds to tens of seconds on the development machine. No circuit-tier files: the tier does not support general stabilizer codes.
Generate all 44 cyclic shifts modulo 44 of X=[1, 43], Z=[2, 15, 29, 42]. This specifies the complete submitted check matrix without search randomness. Install NumPy, then run from the repository root:
python research/evidence/44-2-9/scripts/k2_certify.py python research/evidence/44-2-9/scripts/verify_css.py python verify/qldpc_verify.py codes/44-2-9.json
The supplementary scripts replay the completed investigation; they do not replace or modify the trusted verifier.