Gregory's Server
Fediverse Spam is a real problem!
Most fediverse implementations cannot handle it without reports or manual intervention.
There is one #fediverse platform that can though, and the local AI algorithm that powers it gets better every time a potential spam post is detected, and the admin trains it by deciding "spam" or "not spam"
And it's about to get even more powerful, we're working to release the nextgen Autospam that can do much more, with much less intervention.
@dansup Good old Bayes, I remember reading the spamassassin docs decades ago and thinking it was like magic. Do you have any idea how well this scales?
For anyone reading the toot and not the blog post, and being immediately put off by "AI", this is not AI: it is a proven mix between maths and technology and bears almost no resemblance to the damage that today's AI is doing.