As you might have noticed, /everyone is drowning in spam. It was always a bit bad, but now it's practically unbearable.
While /everyone is unusable right now, we're not giving up on it, nor on the "laissez faire" policy of including everyone. At least not quite yet.
That's why I've been building an artificial intelligence for recognizing spam over the last months that is now in an early test phase. May I introduce: Gargamel!
With your help (your spam reports), a bit of python, a hacked dspam, a bit of scikit.learn cobbled together with a few feature extraction regexes, Gargamel is slowly coming to life.
And one of those signs of life you might have experienced today: a funky little mistake made him report each and every post as spam. Doesn't matter that I built him with the express purpose to generate as little false positives as possible. He took no prisoners. Read More »