Its not clear to me that the issues I’m identifying can be resolved with a new release. Id love to have the discussion, but it goes deeper than just discovery.
I think the best example of how to ‘lemmy’ properly, or in a way that doesn’t create ‘wasted votes’ (in the gerrymandering sense) of content, is the startrek lemmy. The focus on a niche topic and own it entirely. Theres no point in having lemmy subcommunities based on startrek because the startrek lemmy is so great and makes such great content.
Contrast this with lemmy.ml, lemmy.world, most of the other big instances. Mostly redundant sublemmys. Because of this, the quantity of content is very thin, and because of this, good content is less likely to gather momentum. A way to group sublemmys across instances could work, almost like a kind of sub-federation (like, on lemmy.world, maybe lemmy.ml/c/memes is federated into lemmy.world/c/memes. any content posted to one is ‘considered’ posted to the other). It seems relatively straightforward (not a trivial lift, but reasonably possible) to merge the two RSS feeds.
The same issue exists with discovery. Things are too diffuse in lemmys current network structure.
I’ve gone through some graduate level coursework on network mathematics, and I work with networks in a very different context, but I don’t exactly have the math skill to write out a proof on this. This paper outlines the basic concept, but to extend it to social media, we basically get much much better content in the form of submissions, comments and discussions, with super-connectivity. Its kind of fundamental to social media graphs and there are some clear barriers in lemmys design that prevent it.