It's not the Community NIP, it's just clusters of npubs and the main npubs they all associate with. Or something like that.
Like. Actual communities.
Discussion
That was my understanding just glancing at it. Attempt to surface the actual (apparently 500 of them) communities. Visualization forthcoming, could be interesting
I don't understand, but I don't understand much. It's fine lol
Oh! I didn't realize!
Though if no graphs are shown then isn't the data kind of meaningless? Like, what were the clustering parameters?
I'm used to dot arrays. So, no idea how this works.
I'm assuming he used distance to collapse clusters of dots and then pulled out most-common follows or something.
I have no idea how this works either. It looks like OP had his own idea of how the communities should be and combined anything that didn't fit that idea into a few raw numbers.