Let's say that for news we need the "economic web", meaning no more "newspapers", but independent reporters with their own blogs and followers who micro-pay for individual articles by downloading and truly spreading them from a human to another instead of "many humans toward a network hub". This would solves the censorship problem, as you have thousands of scattered sources that are impossible to control or corrupt all at once, and it generates pluralism where the background noise is mitigated by personal scoring, like the WoT of GNUPG back in the day or Nostr now. The FLOSS foundation guarantees neutrality, while distributed sources like Radicle ensure it can't be censored.
But... for this to succeed, you need LOADS of participants; otherwise, the small amount of compensation collected isn't enough to live on or even maintain as a side hustle. It still works to some extent as long as people doing other things have their say in an interesting way, but it doesn't take off. To get a lot of people, you need to attract a lot of people.
Increasing censorship in recent times has made people migrate from Reddit and Discord to other things, but honestly, the alternatives out there are a bit of a mess. Personally, I set up Matrix for family and friends, only because XMPP doesn't seem to attract anyone, and both Matrix and XMPP are largely a pain to self-host properly if you want to include audio/video calls. The "fragmentation" of other tools is total. To attract people, you need a single, slick go get -able, cargo build -able, pip -able (and so on) application that does pretty much everything without a ton of dependencies. That way, someone discovers it, it's easy, they come for one or two features and discover others, providing enough mass to kickstart the spread. The Fediverse model does not offer that so far, Nostr is only a little bit better, ZeroNet is dead, ...
It seems that recent/young developers can't grasp this, so caught up as they are in what they do at work, the "zero barrier to entry" of living on someone else's servers, which hamstrings every FLOSS project. Creating countless separate applications useful for selling services in a commercial model, but it's a recipe for failure in FLOSS. No idea to integrate client and server in a single app to solve even if DHT and alike are there since decades...
The mind is one, so the application must be one and integrated to cover the bulk of needs in a single environment. Emacs understood this a long time ago, Smalltalk workstations even earlier; today, it seems most people still can't wrap their heads around it...