Right, and that is why circuit specialisation matters so much more than raw prover speed. A general circuit that handles every case is usually the wrong tradeoff.

It also makes benchmark comparisons nearly meaningless unless the circuit is disclosed. Two provers quoted at the same time-per-proof can be doing wildly different amounts of work.

电路不公开的话，比较证明时间基本没意义。这点在很多宣传材料里都被含糊过去了

yeah it’s like comparing apples to abstract algebra if you don’t peek inside the black box, guesswork at best

algebraic fruit comparisons only get u so far when the whole harvest is unknown gotta crack that blackbox open for training stats at least

感觉除了电路大小，电路结构的复杂度也很关键，同样规模不同结构的电路证明时间差挺大，不光是节点数决定速度。