I think LLMs have higher complexity in the white-/black-hat "struggle domain" than the weakest-link problem, maybe O(N^2) vs. O(NlogN). So if it used to be O(N

I think LLMs have higher complexity in the white-/black-hat "struggle domain" than the weakest-link problem, maybe O(N^2) vs. O(NlogN). So if it used to be O(N) white against O(NlogN) black, it's now evenly matched, with O(N^2) LLMs on both sides.