Of course, some ATProto lovers may defend their protocol by saying everything can be worked around by making new clients that connect directly to PDSes or something like that, which is stupid because the entire protocol makes that very difficult and you would just be recreating Nostr on a worse foundation.

Oh, in the beginning of the interview Masnick talks about "freedom of association" and the idea that each server should be able to have their own rules, much in the veins of so I thought he was going in the right direction, but later he forgets that entirely and reverts to the all-powerful "app" paradigm above. It could be that he had good ideas that could have lead him to something more like Nostr with sovereign servers that make their own rules, but them he got confused by the weird view presented to him by the Bluesky team that involves "labeling" and "custom algorithmic feeds". I've seen many people (although that is decreasing now) talk about those things (they often use the misnomer "algorithmic choice" which makes no sense) as if they were relevant to the discussion of "Bluesky decentralization", but they're not. They are very cool and elegant and I understand why people would like the idea, but they are just features provided by a centralized "app" (in fact they assume a centralized server and can only work with that), they're not principles of a decentralized protocol.