Oddbean new post about login | logout So I’m guessing the way bitvora works is I run my own bitvora WoT relay, use only that one relay, turn on global feed in any client and I see content from my follows + their follows- is that right? Kinda a cool idea. We could do that, but with lots more options.
Yep it can work for global or follows only. I follow about 1500 people and my network is about 55,000 people after the graph populates, so pretty sure I'm catching almost everything on one relay
This would be an alternative to the NIP-51 list export. Have a simple control panel with a dual slider for the Grapevine WoT score, to set the max value and min value. See only content from users in that range. 🤔
Dammit david!! Not even built the first two use cases and already someone comes along and suggests another!! Good thing we have @cloud fodder on our team with a whole bunch of relay.tools to play with!!! 👀
Haha yes the possibilities are endless! But they will all use the core library we’re working on right now! 🤓
is that library FOSS?
Yup. The algo is similar in some respects to PageRank, but with a few very important changes. It’s implemented right now at grapevine.my but it’s embedded in react, so we’ve just started the process of rewriting the algo as a standalone js library. Might do other languages after that, haven’t decided yet.
Here’s a geeky overview of the grapevine algo where I draw comparisons to PageRank and make the case for the differences between PageRank and “GrapeRank.” Specifically, why the introduction of a nonlinear term is unavoidable. https://habla.news/u/david@bitcoinpark.com/1725006462512
This looks pretty good but I am currently unable to read the mathematics part.
Honestly … I needed a personal walkthrough to grasp it for myself … but that’s prolly just because I’m dense. Maybe we can do a live stream presentation of the GrapeRank mechanics at some point … @straycat ?
I’d be happy to do that. Might work best to have someone else as host who can ask questions as I attempt to explain how the math works and why it is what it is. Right now I’m writing up the idea of the “grapevine worldview” which lets you visualize the sources of information (follows, mutes, lists, notes, etc) and how they are processed in multiple stages by the grapevine to provide answers to whatever questions we are interested in. That may help to clarify why the grapevine WoT score needs to function as a *weight*. By weight, I mean: how *loud* is someone’s voice on some given topic? We want to screen out the bots, but we don’t want to make it a popularity contest, and the math is chosen so it strikes the right balance.
