Skip to main content


Spam is one of the biggest issues @write_as and other @writefreely instances face -- especially in the age of LLMs.

I've wanted to build a spam classification system for a while, and I'm finally getting to it this week. It'll be made especially for #WriteFreely, and plug into it to help with moderation, both manual and automatic.

Will share more details soon!

in reply to Matt Baer

This is an extremely interesting initiative! I'd like to know more.

I'm also thinking there might he huge potential in fighting spam/malicious/bot accounts by pooling & analyzing what different instances are seeing.

in reply to Matt Baer

I had to manage the same problem in HedgeDoc and ended up using a naive Bayes filter, which seems pretty effective so far.
⇧