I claim no expertise here ... so take this with plenty of salt. I also don't know how much of this is specific to the protocol itself or is just the way bluesky have decided to build things.

I see two interesting and nice things here:

• Users and their follows or social graph are portable across the protocol
• The architecture (again, not sure how much of this is a protocol thing) has different levels of centrality or decentralisation for different parts of the system. So you don't have to pick an instance just to create an account but can instead pick moderation policies and feeds when you want to. The issue is that underlying everything is a big giant server that's collecting data and spitting it all out as a firehose. There's only one right now (BlueSky's) but the code is open and they say that others can start one too (however onerous that would be). The upside is that all the things downstream from the giant server can rely on it and instead make apps, feeds, execute moderation etc ... which could be a nicer experience for both devs and users.

In the end, my impression of it is that they're building more of a framework and ecosystem for others to build social media within. ActivityPub by comparison is much more of a playground of ideas and tools that people can make and host whatever they want with it. So more truly decentralised but also, IMO, puts more weight on the developers and the users to make the ecosystem happen and work well. For instance, we could have more portable user accounts on the fediverse, but we don't (yet), because that'd have to be built and then implemented by all the platforms.

Once I see another Big Giant Central server running in some sort of sustainable or functioning way, then I personally think it'll have a lot of promise. Before then, however, a number of developers might get interested in developing in that ecosystem because of how it might allow them to make the thing they're interested in and not worry about other things.

As for how ATProto and ActivityPub can and should relate to each other in the future ... they're the only two decently sized projects really having a good shot at this decentralised thing (though there a few web3.0 things out there AFAIU, eg farcaster) ... and I think they're better off being "friends" rather than "enemies" for that reason.

If my impression of their differences is accurate, they'll have different strengths going forward. IMO, ActivityPub will be more of smaller community thing. If all of the neuroscientists want to create a network of forums and blogs, that they're in control of, around the world that all talk to each other but without being connected to all of the other social media, then the fediverse and its platforms will be the ecosystem to use. If neuroscientists want to talk to the rest of the world but still have ownship and control over their data and maybe their platform or feed or moderation, then AT-Proto will be the place to go. Bridges between the two would complement the flexibility here.

i have a layman's understanding of AT Proto, but it seems to compartmentalize between different parts of the service. Front ends, databases, and backends can be hosted separately and amalgamate into one, in the abstract.

Practically, however, AT Proto allows account portability, wherein users can swap what instance they use as a frontend on a whim, even if their home instance is down. Usernames are domains instance of username@domain, that are verified by the DNS. But AT Proto seems a lot less flexible than ActivityPub. We'll have to see when federation is live, but I'm not sure it really suits anything beyond (micro)blogging.