From a UX perspective, decentralization makes everything worse, and those issues get worse with scale, unfortunately.

For example, you might reply to a post on one instance, but everyone ends up missing your reply. Was your reply just not worth reading? Or was it a problem with federation? Both can happen. If there's only one giant instance the behavior is clear and ordering makes sense, but with lots of smaller instances you can have ordering issues, propagation delays, and the like.

If its not simple enough to use for someone with a 6th grade reading/comprehension level then it will not be widespread (I think that's about 50% of adults in US). I think de-centralized as a concept fails that test. Widespread meaning 100's of millions of users aka people's grandmothers are on it.

If a free service requires fairly involved explanation of how it works, its not going to catch on. So the problems you are asking about don't need solving because the nature of a de-centralized social network prohibits widespread adoption. Previous flavors of decentralized internet forums never gained widespread usage either and the people that administered them typically didn't want them to.

IRC scaled well for its time but it didn't have today's population. Also remember that we could download every movie and game for free at one point.

Likely decentralized networks might end up like web3 or torrents.

You're also likely to get partially centralized communities again like 4chan or the more extreme 8chan, or something fairly sustainable like Something Awful.

Incentives are everything. Eventually, one of these platforms will get big enough where advertisers are interested in the data. Right now, platforms like Farcaster are completely open (you can access user data from dune analytics). I think this is the wrong approach. First, user data should be private. Second, user data should be monetizable. The latter is how all trad social media platforms have scaled to billions in revenue. The former is why so many users hate them for it. IMO, the right solution for decentralized social is one that leverages zk protocols to protect users data, while enabling a compute to data solution for advertisers where BOTH platform owners AND users are compensated for their data.

Mastodon would already gain so much if they integrated fedifetcher[0] as an option for all instance admins. Make it opt-in, add a note that this component isn't maintained by Mastodon (the legal entity that used to be a German non-profit). Most of the annoyances, especially on small instances, would go away.

[0] https://github.com/nanos/FediFetcher

Meta is embracing ActivityPub, lending credence to the protocol

Bluesky has a more interesting compositional design (ATProto), seeing more ex-X users there every day

Both can be monitored by creating a similar UX for users, running large data nodes so most dont notice the difference, and then injecting ads in their interface or charging a monthly fee

Both open the door to increased competition, because switching cost should be lower (in theory), which is what we really need.

I definitely think decentralized architecture is the better way to build social media.

AFAICT, Mastodon seems to scale quite well. Not without issues but they seem to find ways to overcome the biggest hurdles.

Social networks scale because of the social component, not the network component. Beyond commodity technology, technology mostly gets in the way.

It's hard to beat "friends chat" for decentralization. And it's hard to beat Facebook for centralization.

Good luck.

Why should it scale? People act as if everything is scalable and should scale. And if it doesn't scale something is wrong. This is just absurd mindlessness and has nothing to do with solving an actual (social or otherwise) problem.

It happens more because their skills are more aligned with adding and supporting new racks at the data center, or advertisors/corps are frantically searching to create more artificial digital Real Estate on top of which to put up digital billboards and capture attention.

So you have to be clear about what problem you are trying to solve and for whom.

Also check out the UN report on the Attention Economy. It provides lot of analysis of the space and food for thought on how future social should be designed.