In particular, I’ve figured out a way to specify sentiment/interests efficiently and combine it reliably over federation, and the data structures required to do that.
I’ve also provided some ideas for sensible defaults (automatic selection of instances, and accounting for instance load), with incremental enhancements to specificity for more advanced users ^.^, as well as a general search mechanism that can be derived from this - though for efficiency, it might be worth trying to develop some sort of probabalistic reverse index to avoid a linear scan, if we’re talking about discovering entities like users or groups where there may be very large numbers.
I hope that if people are interested they will boost the post onto Mastodon, which afaik is where the devs and ActivityPub standards people are, and try and get the ball rolling, because my focus is elsewhere right now, and the social aspects of developing things like this are much more difficult for me than the algorithmic and architectural parts ;3