Often you can sort of moderate it yourself based on what you choose to see. Anyone can host a racist site on the internet, but people aren't really turned off from the entire internet by it. Similarly, if there were a racist subreddit, right now that feels weird because reddit is actively hosting it and choosing to allow it. But if there were some kind of decentralized reddit and people made a racist sub on that, I don't think people would view it the same way.

As for scammers, imagine something like twitter. If I follow my friends and celebrities I like, scammers will be relegated to the comments section of their tweets (I wouldn't see their actual tweets because I wouldn't follow a scammer). Then you could use some kind of web of trust system to filter the comments - perhaps I would only request comments for people I'm following, or people followed by people I'm following, etc.

You compare a pull to a push there. Sure on a service where you pull the information, you wouldn't pull what you don't want. Internet is a good example of that, your example of twitter-like is also pull.

It fail when it's a push instead. That would be anything that have suggestion, any public conversation, like a forum, or twitter, even private conversation in some case, like an email service, etc...

For sure you won't pull what you don't want, that's by definition, but you moderation is needed for the push system, where you don't control everything that goes to you.

Most service are push too, we want things to be suggested, we don't always know exactly what we want, we don't know even if it exist.

Given that people are moving from the regular web to web based platforms I don’t think this is true.

Ah, the fallacy of the killfile and transitive reputation. Sorry, but that doesn’t work and has been shown over and over again to not work for large scale communities.