Increase the encoded dimension of the public [[80,24,8]] 2BGA code by retaining a smaller stabilizer subgroup, while keeping maximum check weight 10. The parent was contributed by @FarLab in PR #3773.
The selected subgroup keeps 24 original X rows and 24 original Z rows. Their zero-based row masks are X = 877032374220 and Z = 879599236095. Both selected sector ranks are 24, giving k = 32. The displayed basis has maximum check weight 10, maximum qubit degree 7, 48 checks and total incidence 480.
The submission tool searches for logical witnesses and verifies the generated JSON. Its sector distance claims are witness-backed upper bounds. The retained exact-distance research check excludes all nonzero centralizer words through weight 5: 3,241 syndromes for errors through weight 2 are distinct, and all 82,160 triples avoid that set. Genuine weight-6 quotient witnesses establish exact distance 6 separately.
The fixed-parent degree-6 census leaves 100 admissible binary selections, all with sector ranks 20 and 20. None spans the required 24 plus 24 ranks; the selected degree-7 presentation avoids that obstruction.
Recorded workflow configuration: GPT-6 Astra + GPT-6.1 Sol. The official qldpc submit run used 20,000 RIS trials per sector with seed 0, followed by its independent refutation check.
Take the parent's 80-24-8 JSON at fa4b0de3bef2b078ffa41860af8d6ee1633931a8. Its Git blob is 43dae9d893a9f2b5c37480d6df4dbc2e10ac423f. Preserve qubit labels and original row order. Keep X row i when bit i of 877032374220 is 1, and Z row i when bit i of 879599236095 is 1, starting at bit 0. This directly reproduces the submitted matrices.