Oddbean new post about | logout
 Another reason the Influence Score is better than the WoT Score is that the WoT Score COMPLETELY IGNORES anyone more than 2 hops away on your social graph.

Looking at the data below, there are almost 13 thousand users users who are between 3 and 5 hops away from me. Their Influence Scores range from just over zero to 0.3433 (max score is 1), but their WoT scores are ALL ZERO.

If you want to discover content from outside of your bubble chamber, use the Influence Score, not the WoT Score. https://i.nostr.build/Vwn2G.jpg https://i.nostr.build/KGjlO.jpg  
 Where is this from? something you’re building ? 
 nm found the link..

https://brainstorm.ninja 🫡 
 Am I doing anything wrong? I wanted to stress-test with my 2487 follows.

https://m.primal.net/IqSF.png  
 Hmm, maybe hasn’t downloaded your follow list yet … go to your profile page and see if that’s the case 
 Looks like the connection to purplepages and wikifreedia websockets are not going anywhere.

Over a minute after page load:
https://m.primal.net/IqSZ.png  
 Are you able to see any data on nostrapedia? Topics, categories, authors? 
 All 0.

https://m.primal.net/IqTF.png  
 Hmm. I wonder: if you logout does it download any wiki data? 
 Nothing. I might be IP blocked by purplepag.es. 
 Hmmm

https://m.primal.net/IqTv.png  
 That is odd. The whole demo app works for me 
 Yes! Arbitrary 2-hop limit is ridiculous on its face. 

nostr:nevent1qqsyg9f0zkkjgqm9nnv8lhrv7vj857s9dcelehr8k7vvjqse7qzlxvgpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsyg89yuk7j99axqt4t3pehz8xjkdy8jwjveyrruync50fc7v6z6ss9upsgqqqqqqszvaywt 
 All of these are useless. 
 (this reply is funny if you notice that I don't really follow anyone on this key) 
 David's plan is aligned with your comments here: nostr:nevent1qqstph7e5cdwd2dq68sp46qka9seuym2szfnmn8drwvhl7ddg4qmahgppemhxue69uhkummn9ekx7mp0qgs8wakr9493685t725kh2ltgwke4hs400fk8kymslak8eh3g42c3zqrqsqqqqqp6jqrpr

It isn't about mutual follows or prioritizing the npub itself in a crude way. The trust graph is context-dependent (as you rightly point out is the best way to go), subjective, and follow-agnostic:  https://github.com/nostr-protocol/nips/pull/1208

 
 Yup, I like NIP-77, the heart and soul of which is 3 fields: context, score, and confidence.

And when it comes to follows and other proxy indicators of trust, I like the idea of interpreting those data points as if they were NIP-77 formatted. And giving them a low confidence, since proxy indicators do not necessarily equal trust. But sometimes it’s the only data we have, so we use it. 
 I skimmed the thread Vinney linked but I'm not understanding at all. What are these fields or the points of them? Why not just 1 field, which is either a priority ranking or a weight for each npub in your list, allowing the rest of the functionality to rely on well used indicators like follows, reports, likes, upvotes/downvotes? 
 For one thing, you lose the context then. I may trust you highly for your programming opinions but not for your political opinions. If you lose context then you're back to the crude world of "this npub 100% good, that npub 100% bad", which is not a nuanced social model. 
 That was the first thing I addressed tho - you do context by changing the rankings

They could be weights but rankings are easy to click and drag when changing, or a UI could represent weights as rankings with advanced settings to see the numbers for best of both worlds

Ideally something like a wiki client would remember your list for each wiki category, for example, so if you have wildly different lists for different topics you don't have to change them every time or switch npubs to switch lists  
 This would also allow cool features like a hypothetical "trusts all npubs in existence" slot in the list that you can click and drag to the top to override all reports and see everything unfiltered  
 Oh, I see after giving it more thought how a context field would be useful for adding another degree of separation into the "web"

I rate npub x at 90% for politics, they rate npub y at 90% for politics, npub y gets their opinions included without being in my list - doesn't work for different contexts without the "for politics" field

But that's still only 2 fields - why would we want 3? 🤔 I might have to read the thread more later today 
 After reading the thread a bit more, I can now see reasons for 3 fields: genre context (code vs politics), action context (edit vs delete), and trust score/ranking.

I still can't see the reason for the 3 fields to be context, score, and confidence. Still reading. 
 Have read about half the thread and it all further solidifies my opinion that the ideas being discussed in that thread are over-complicating something that can be much simpler.

If anyone wants to give a direct explanation of what I'm missing, I'm open to it.

Otherwise, it seems to me the best path to functionality right now would start with adding 1 field, then maybe a second and third could be added later. 
 I see those three fields as a goal although that doesn’t mean we have to graduate from what we have now to using all 3 instantaneously. Doing things in baby steps for the sake of simplicity is the strategy I’m following right now.