Detecting spam just from HTTP headers

I'm going to do the thing that makes it seem like I'm a suckup, but I'm unaffiliated with YC:

PG mentions this in his articles about Bayesian Spam filtering. Seems pretty compelling.