I recently decided to enable Curated Onboarding on pixelfed.social and a majority of the applications are suspicious and most likely spam.
I'm brainstorming a few ways we can better handle this without requiring admins to work so hard.
We can use simple keyword detections a la bayesian analysis to automate this to a degree, but that requires some training first.
While it's unfortunate to have a spam problem, this does indicate some level of success, so I'll take that 😎