When it comes to spam, I find that there's a

lot of quantity substituting for a lack of quality. There's a lot of patterns to be found in the messages they emit. Which makes sense, because.. well, spamming the same messages over and over, right. On the defensive side (in a Telegram context, although in my daily life it also extends to e.g. email), I think it's more of an issue with lack of "programmatical" features in common bots. Sure you could filter for message content, but what about display name? What about bio contents? What about a URL string? What about common behavior of a given spammer? There's so much untapped potential.

Dear bot developers, if you're looking for a project to make — make an anti-spam bot. And as you combat spam by hand, implement every pattern you can find in that bot of yours.

Telegram not so long ago launched its own anti spam feature that as far as i understand also checks things like bio etc

And ofcourse it sucks as always I'm the victim 😒

As far I know, it checks only messages and premium status, I've seen so many spammers with premium, and they are unaffected by aggressive antispam.

I'm sure it checks for more because i saw it deleting "Hi" messages of people who had clear spam in their bio etc

Hm, nice.

With this I presume you're referring to this feature called "aggressive anti-spam"? I believe it's a step in the right direction, but it was so poorly executed. In theory, it should be possible to churn through all the billions of messages that were emitted on Telegram after all the years of its existence, and figure out what's spam and what's not. Messages that were deleted by chat admins instead of the poster themselves would be a good start, although even that still puts a lot of responsibility on the chat admins. I don't think that AI is necessarily the only answer to this though. Spammers are plentiful, but they are generally to programmers, what carjackers who hotwire cars are to car mechanics. In other words, filthy skids. The vast majority of them just use others' code, and send a torrent of the same junk over and over again. That's a pattern, which proper programmers could look out for. Even if they employ all their (not so fancy) evasion mechanics, the patterns would still persist. I believe that this is how a management bot that fully taps into Telegram's potential could tackle this once and for all.