... but any replacement you build will, in practice, have to include a single centralized "relay" that aggregates all content. Since that's a lot of content, it has to be run by a big, easily found, easily pressured organization. And everybody "porting their accounts across" means a flag day that's going to be almost impossible to organize in practice. It'd effectively be just as much work as switching to an entirely new protocol.

Maybe you could theoretically have an AT "app view" that takes data from multiple relays, but nothing in the implementation does anything to support that, and as far as I know nothing in the protocol does anything to help it discover the relays... which in practice means that even if you extend the app views to use multiple relays, there will never be more than a handful of relays with meaningful reach.

The AT protocol is at best a really crappy excuse for decentralization. And frankly a pretty poor example of open source too, given the usability and organization of the code they release.

Compare with, say, Nostr, which is actually decently decentralized... but, in not-unrelated news, suffers from massive content discovery problems. Or compare with Briar, which is even more decentralized but has both discovery and scaling problems. Or for that matter Usenet.