Currently, shakespeare just ignores the whole relay side of Nostr and has this afterthought retarded implementation. It just uses a handfull of popular big relays. Now here are just my assumptions: It uses those relays in hopes it covers most stuff out there; It forces you to select between them probably because doing anything else is too complicated. I.e. the point of the relay selection is only there as a 'well it should all be on this relay, but in the offchance it is not, you can try these others as well. Its terrible. We can give it some slack because the whole vibecoding thing is novel, and getting the AI to create apps that handle various event kinds is innitially more important, but it is only ( the proverbial) half of the story.

You're approaching it from the wrong side. Of course if you frame it like that relays become a burden, my entire note was just to say you should try to frame the fact that relays exist and they're different from each other as a good thing. I don't know what is the best way to solve that, but it has to be solved, you can't just ignore the relays and hide them entirely. Well, you _can_ do that actually, but if you do it will just mean you're condemning Nostr to be a centralized garbage fake protocol, and at that point why waste your time in it anyway?