I think his idea was to reduce visibility of harmful content (like exclude it from the algorithm, or maybe just penalize it heavily) without fully banning it.

I suppose downvotes are also a solution (up to a point - doesn't work unless there is a strong enough consensus).

But even if these don't work it doesn't mean censorship is the only solution. Just because we don't have a good idea doesn't mean nobody will.

Radication and disinformation are serious problems and some censorship is better than nothing - but still seems like there has to be a better way.