hmm did you read the entire thread? comments like «The problem is that _AP itself_ is hard to implement correctly and nontrivially in a way that scales», «I was surprised at how difficult I found the spec(s) to read compared to theoretically more complex ones I've read in the past,» «something went direly wrong in the standards realm sometime around the OAuth2 RFC, which says verbatim that following the spec as it stands will produce implementations that don't interoperate. ActivityPub has the same nature and when we complained about it last year we got a bunch of dweebs telling us that the activitypub spec is fine because it has optional protocol annexes that "all the real implementations" follow»…
and basically all the comments. including yours.
I wasn't specifically endorsing Bluesky, but it would not be surprising if a professional, clean slate protocol based on the "learnings" of the past would be simpler to implement and just all round better. I think the use of DIDs opens up to many interesting options, none of which need to use a blockchain.
It's a shame about ActivityPub, I think Linked Data is a really important tech that hasn't been picked up enough, but from what I hear it's a really half hearted implementation at best.