Imagine you're packing a suitcase with a strict weight limit. You can fit any single item easily — each weighs well under the cap. But you need a complete outfit: shirt, pants, shoes, belt, and they all have to fit simultaneously. The heaviest single item isn't the bottleneck; the heaviest complete coordinated set is. That's the core tension in this paper, translated from fashion to fault-tolerant quantum computing. The authors call the weight of that heaviest coordinated set the "frame width." The committed claim: for high-rate quantum error-correcting codes, the minimum weight of individual logical operators (distance) is the wrong metric for resource planning. What matters is the minimum weight at which a full regular canonical logical basis can be assembled — and this paper introduces frame width f as that parameter, proves structural obstructions that force it above distance, and builds explicit codes where f stays close to d. This is a genuine conceptual addition to the QEC parameter landscape, not just a new construction. The constructions are concrete and large. A [[1088, 128, 22]] lifted-product code achieves f = 22 (matching distance exactly). A [[1024, 256, 24]] pair-partition code reaches f ≤ 27 with check weight w = 10. At w = 11, the authors demonstrate a strict separation: d = 24 but f = 25 (proved optimal), showing these parameters are genuinely independent. These are not toy examples — they operate at scales directly relevant to near-term fault-tolerant architectures, and the explicit logical bases are provided, not merely existence-proved. The structural obstruction result is the theoretical spine. For CSS codes over odd-order groups built from fully populated binary monomial checks, full regularity forces distance two — a hard no-go. For power-of-two order groups, a sufficient condition for full regularity is identified (full row rank after collapsing group elements to 1). This cleanly separates the design space into tractable and intractable regimes and explains why certain natural constructions fail. On the validation front, the distance certifications use a new method combining symmetry reduction with stabilizer-based pruning for exhaustive search. The authors certify exact distances for codes above 1,000 qubits — a nontrivial computational feat in itself. The method is described in enough detail to reproduce, though no independent replication exists yet. The comparison ladder is implicit: prior work on lifted-product and fiber-bundle codes reported logical bases several times heavier than distance, and the point is that this paper's constructions close that gap. The milestone question is where this gets practical. Current fault-tolerant proposals assume logical operations compiled from whatever canonical basis is available, and compilation overhead scales with basis weight. If frame width can be driven to match distance as codes scale to thousands of logical qubits, the compilation overhead drops substantially. The gap between f and d at these scales — zero for the LP code, at most 3 for the PP code — is the number to track. The obvious experiment not run: applying these design principles to bivariate bicycle codes or other families currently favored by hardware groups (IBM, Google). The authors stick to lifted-product and pair-partition families. The honest read is (a) and (c) combined — the algebraic machinery developed here (group-valued codes, binary monomial checks) may not port cleanly to every family, and extending it is likely the next paper.
