I have been thinking about this problem recently and believe the solution may be a new fediverse protocol/service that provides:
- Federated Emergent Topic Taxonomies
That is, a model of the relationships (e.g., is the same as, is a type of, is related to, etc) between different communities (/groups/services/instances, etc.) that emerges from the way that users/servers interact with them, that different servers can maintain independently and merge or split by consensus if they choose. Then other services (like Lemmy instances or clients) can tap into this information to provide solutions to problems like the one you describe (e.g., a feed of all the photography communities, regardless of which instance they're on).
I think there are several big conceptual and technical challenges to implementing this. I'm keen to discuss them.
Does anyone know where I would go to discuss this with the people who care, have struggled with developing new fediverse protocols and/or are best positioned to spot the flaws and possiblities in the idea? So far I see mostly w3c working groups taking behind closed doors.
How do you sort the content without votes? How do you pick out the good stuff from the spam?