Oddbean new post about login | logout You mean it calculated global trust scores?
Local trust scores. It computed a N x N matrix T, where t_ij is a number between 0 and 1 that represents how much person i trusts person j.
How big is N?
I did a nice asymptotic analysis here: https://github.com/pippellia-btc/The-Problem-of-Spam/tree/main/v0.2 N ~ 100k The neat part is that multiplying two Boolean matrices cost O(N^2) instead of O(N^3) (or O(N^2.8) with strassen)