Oddbean new post about login | logout The grapevine algo would ignore those 10000 npubs, because their grapevine WoT score is zero by default.
I have followers and I, as the wannabe spammer, would follow some of the generated ones. Would that endow them with some score?
Everyone’s score is zero by default unless there is at least one path of follows from me to the user in question. So the wannabe spammer would probably also have a score of zero, which means it doesn’t matter (from my point of view) who the wannabe spammer is following. Hope that addresses your question
How do you know you aren't following wannabe spammers. They could be posting a couple of witty Satoshi quotes to gain everyone's trust while behind the scenes scripting 10,000's of npubs, some of which they follow to endow them with that transitive point score.
That’s always a possibility. If it becomes a problem, like if you see your extended WoT network grow by 20k bc someone 4 hops away from you followed 20k spammers, you just unfollow and mute that pubkey. Or maybe one or more others in your network find the problem before you do and unfollow / mute the bad actor, and the problem is solved before you’re even affected.
In that scenario, if one user many hops away from you follows a spammer, that spammer’s grapevine WoT score will be nonzero but still pretty low. It takes more than just one follow to push your grapevine WoT score up high enough to approach the max score of 1.
That's very interesting. How do I see what someone 4 hops away from me does? Is this somehow visually presented as a graph that I can click through? And if my network suddenly grows by 20k npubs maybe someone 4 hops away just followed people from Paraguay and now I suddenly have all the nice people from Paraguay in my extended network.
Eventually I want those kinds of visualization tools to exist. I tried building them a while back using visjs and the graph bogged down too much with more than about 500 pubkeys. For now, I’m focused on my exposition and understanding of the algorithm itself. And on rebuilding the PoC on the back end instead of all front end.
Also, the reason we’re using follows and mutes to build the baseline grapevine WoT score is that that’s the best available source of raw data. But the grapevine is designed so that you can pull in multiple sources of data into a single score’s calculation. So if bad actors start to weasel their way into the networks, and people care enough to want to weed them out, then we can start to employ NIP-58 badges or NIP-32 labels or NIP-51 lists to help us distinguish the good vs bad actors. And those datasets can be worked easily into the score. In this way, you neutralize the strategies of the bad actors as they implement them. My expectation is that you will be able to modify your grapevine more easily and more quickly than the bots can implement their attacks. Which will hopefully make it less profitable to create bot farms in the first place.
Sounds like you haven't thought this through.