in the thread, someone links PG's filtering method, which claims 99.5% effectiveness? that would still be getting me 20+ spam emails a day at that rate

further, if you are trying to block, lets say, crypto scams, on a platform which allows strong positive discussion about crypto, which allows people to talk about stuff they are selling for crypto, etc etc etc, you easily start losing points to differentiate

the defender needs to classify every message on the site, in a way that allows detecting spam well after classification, while maintaining over a 99.8% rate these days, while aiming for a 0% false positive rate

the attacker just needs to type random messages at their keyboard and use reused passwords / buy client id/secrets from shitty devs, to get access to verified accounts