straycat
david

I’m waiting for @npub176p…vgup and @npub1wf4…dgh9 to hammer out the details of their WoT DVM. If they can convince a few devs to ingest scores using their proposal, that will definitely motivate me to offer personalized PageRank, personalized GrapeRank and personalized degree-of-separation scores (and maybe some other scores) via DVM. I already have some endpoints built out so I’ll hopefully be able to employ the DVM format relatively quickly.
I encourage discussion of their WoT DVM PR, which is here:
https://github.com/nostr-protocol/data-vending-machines/pull/38

2025-02-14 19:40:11

- reply

If you go to grapevine-brainstorm.vercel.app you can calculate personalized PageRank and personalized GrapeRank scores and export them in a format (NIP-85) that is easily accessible by clients. The about section tells you a little about it.
NIP-85 is new. I don’t think anyone clients are using it yet. But hopefully they will soon!
Work in progress:
1. personalized grapevine WoT relay that calculates those scores for you, stores them using NIP-85, and uses them to filter data that you keep on your relay. It will use strfry and neo4j, the most mature FOSS graph database that I am aware of.
2. Contextual GrapeRank scores, so your trust network can tell you who are the world’s most trustworthy sources on any given topic.

2025-02-13 16:12:01

- reply

👀

2025-02-12 17:02:33

- reply

We have all grown accustomed to the idea of global reputation scores. But I think we will see — are already seeing in nostr — a change in focus and mindset from global to personalized WoT scores. Your npub doesn’t have a personalized WoT score until you specify the observer npub. You have to ask: what is your npub’s score, as calculated from some other npub’s perspective.
Ultimately, the access to sophisticated, contextual, provable, personalized WoT scores is going to unlock value in ways that we’ve never experienced or perhaps even imagined.

2025-02-12 13:35:08

- reply

GM PV ☕️🌨️

2025-02-11 16:42:27

- reply

He’s followed by 178 pubkeys in my extended follows network but muted by 149 😂
3 hops away from me by follows.
His personalized GrapeRank score is zero because of the mutes.
Personalized PageRank score is way too high than it ought to be, but that’s because pPR does not accomodate mutes.
GrapeRank ftw 🍇 🔥
https://i.nostr.build/XVLB2WSy9pVnbWtv.png

2025-02-10 01:14:00

- reply

Ding ding ding ding!! 😂

2025-02-09 20:23:27

- reply

Just got mine
looking forward to reading it https://i.nostr.build/kTg4cqh9sKmof874.jpg

2025-02-09 17:27:18

- reply

@npub10np…tl5h bring you in because this is how we could do it for relay tools

2025-02-09 17:25:13

- reply

CREATE INDEX event_id IF NOT EXISTS FOR (a:ApocNostrEvent) ON (a.event_id)
CREATE INDEX pubkey IF NOT EXISTS FOR (a:ApocNostrUser) ON (a.pubkey)
Like that?

2025-02-09 17:23:50

- reply

Here’s the script I’m running now which took about 45 min:
```
#! /bin/bash
sudo /home/ubuntu/strfry/strfry export --since 0 >> /home/ubuntu/neo4jExport.json
sudo mv /home/ubuntu/neo4jExport.json /var/lib/neo4j/import/neo4jExport.json
cypher-shell -u XXXX -p XXXX -a bolt://localhost:7687 -f /home/ubuntu/apocCypherCommand
sudo rm /var/lib/neo4j/import/neo4jExport.json
```
And here’s apocCypherCommand:
```
CALL apoc.periodic.iterate(
"CALL apoc.load.json('file:///neo4jExport.json') YIELD value AS event",
"
MERGE (u:ApocNostrUser {pubkey:event.pubkey})
ON CREATE SET u.proceses=['initialization'], u.foo='bar'
MERGE (e:ApocNostrEvent {event_id:event.id})
ON CREATE SET e.kind=event.kind, e.created_at=event.created_at, e.author=event.pubkey, e.processes=['initialiation'], e.foo='bar'
WITH u, e
MATCH (e2:ApocNostrEvent) WHERE e2.author=e.author AND e2.kind=1 MERGE (u)-[r:AUTHOR {foo:'bar'}]->(e2)
",
{ batchSize:1000, parallel:true}
)
;
```
Some of the things like the edges I’m adding now just to see if I have the syntax right.
I’ll probably replace strfry export with strfry scan so I can use filters

2025-02-09 01:34:42

- reply

The Starks are clearly portrayed as the protagonists from the first episode, and Jaime is an enemy to the Starks. We even watch him be mean to Jon just for fun. So we the audience know we’re expected to hate Jaime.
Later in the show we the audience love Brienne for lots of reasons. So when she comes to respect and even love him, we know we’re expected to do the same.
One character’s judgement of / reaction to another character. Very powerful narrative device.

2025-02-08 05:37:29

just managed to export strfry data as json, then import json into neo4j using apoc

2025-02-06 20:08:35

For any devs who are interested in an easy way to implement personalized WoT scores in you app, you should consider NIP-85 Trusted Assertions.
https://github.com/vitorpamplona/nips/blob/user-summaries/85.md
https://github.com/nostr-protocol/nips/pull/1534 #note1mul…znu4

2025-02-06 19:39:55

@npub1aka…l9dl maybe you should try this out #note1fx4…hzxg

2025-02-06 19:38:46

- reply

Because I’m always on the lookout for anyone whose thought patterns might be converging with mine.
I have a server that maintains a graph in neo4j of ~ 200k pubkeys connected by follows and mutes, and I use the neo4j graph data science library to run centrality algorithms such as personalized PageRank. My server also runs custom coded GrapeRank, which is nonlinear (unlike PageRank which is linear), but unlike PageRank is designed to yield trust scores that are contextual and that can be used as the weights when calculating weighted averages and weighted sums.
For now, I calculate GrapeRank on just a single, very generic context. Next steps include expanding the graph with more node types (:NostrEvent, :NostrRelay … maybe someday :PubkyUser?) and using the data as input for GrapeRank scores in multiple contexts. May or may not implement GrapeRank using neo4j GDS’s Pregel.

2025-02-06 00:02:52

- reply

That doesn’t completely answer the question

2025-02-05 23:57:36

- reply

Let me see if I understand the demo in your video. You requested a list of Odell’s followers and you ranked them according to personalized PageRank using franzap’s pubkey as the source node (source = the specific node in the graph from which pPR calculation is initiated). The results came back very quickly. Is this because you precomputed franzap’s personalized PageRank? For me that takes about 15 seconds because the graph is very large, approximately 170,000 pubkeys extending out 7 or 8 hops and involving millions of follows. Or did you restrict the graph to be only Odell’s followers? Which would have fewer nodes, fewer edges and calculate much faster.

2025-02-05 23:16:21

- reply

Which pubky engineer is working with neo4j and semantic social graph? @npub17lg…p267

2025-02-05 22:38:37

- reply

Suppose Alice is the logged in user, and you have a list of authors A_I that you wish to sort by rank. Step 1: find Alice’s 10040 note, which tells the client where to find her kind 3038x notes. (If not available, use the platform’s default 10040.) Step 2: for each author A_i, fetch one kind 30382 and extract the score using the “rank” tag. Use that score to sort.

2025-02-05 22:31:32

- reply

I just took this discussion to the above thread and I included links to the NIP-85 Trusted Assertions spec and discussion threads. I think it will be valuable to consider specific use cases and ask whether it makes more sense for clients to access WoT scores via Trusted Assertions versus DVMs (or a combo of both).

2025-02-05 22:14:17

- reply

Takes a minute or two to publish them all.
Nostr.band scores are global, unless I am mistaken. My service and Vertex offer personalized scores, which to me is one of the holy grails of WoT.

2025-02-05 22:04:17

- reply

Maybe we should discuss specific use cases. One where NIP-85 is better, one where DVM is better.

2025-02-05 21:54:12

- reply

Vertex has GrapeRank? 👀 Surely you don’t calculate it … (I’d love it if you did!)

2025-02-05 21:49:30

- reply

I found export of data via NIP-85 to be surprisingly fast and easy. I thought it would take longer to publish that many notes. You can test it at https://grapevine-brainstorm.vercel.app/. Just make sure you don’t publish the kind 3038x notes indiscriminately - relays will probably hate you.

2025-02-05 21:13:33

- reply

I can def see that. I’m planning to look into the DVM spec and possibly using it as a mechanism to access some of the grapevine server’s endpoints.
I’ve been thinking about how clients could put NIP-85 to best use. Perhaps: if you can locate 10040 for the end user, use those scores. If not, each client could have its own default set of scores. Personalized WoT scores if they’re available; if not, default to “global” scores, where global means from the perspective of the client dev. This would allow us to bootstrap the WoT service from scratch, since most users will not have generated NIP-85 scores until it catches on.

2025-02-02 00:32:51

- reply

Yeah it probably could! I’ve been wanting to dig into DVMs so maybe your spec will be the place for me to start. Definitely let me know when you have something ready for review and discussion. 🔥

2025-02-01 20:56:15

- reply

Do you remember when big blockers used to spam the mempool in a failed attempt to prove small blocks bad?
I ‘member.

2025-02-01 20:11:24

- reply

Regarding NIP-85: zapstore is an app, currently android or desktop. And you’re planning on calculating personalized PageRank. So I presume the calculations happen on the user’s app, cached locally, and can be updated whenever you want. My zapstore will scour the universe for kind 3 notes and use them as the raw data for PageRank calculations.
But that would mean don’t need NIP-85 *OR* DVMs. So maybe I’m not understanding your plan. Who is going to be calculating PageRank?

2025-02-01 20:01:04

- reply

Regarding the WoT on zapstore, follows who follow: iiuc you want the intersection of two sets, Alice’s follows and Bob’s followers. And you want to calculate that quickly. Keeping track of follows is simple: all you need is the most up to date kind 3 note. Keeping track of followers and producing that list within the fraction of a second is the challenge. When I click on followers on damus, my app scours the known universe to compile the list from scratch. Takes on the order of 10s of seconds, too long for your purposes.
So am I correct to assume that zapstore maintains a cache of pubkeys and their follows? In sql perhaps? With a column for follows and another column for followers? I tried that for a while and discovered that keeping it current was a pita. Why? Keeping follows current is simple, but doing it for followers is a pita.
My solution for tracking followers is to use a graph database. Why? Because it can retrieve not only follows but also followers very quickly. To test this, I built a few api endpoints to produce follows, followers, mutes, muters, etc on demand. All very fast. And I have one endpoint currently active which would produce the WoT score exactly as you describe, very quickly, if I’ve understood you correctly. You provide the pubkeys for Alice and Bob and you get back either a number or, if desired, the entire array of pubkeys. I can give you the endpoint if you’d like to play with it, maybe check to see if your code and my code produce the same results.
So that doesn’t use NIP-85. But it does use neo4j, which I believe is very powerful.

2025-02-01 17:05:06

- reply

So maybe we’re doing basically the same thing, except I’m using NIP-85 and you’re using DVMs. To me, DVMs seem too complicated. But in all fairness I’ve never tried building with DVMs.
Having calculated the WoT scores, it didn’t take me long at all to make a button to format and publish the data as kind 30382 notes, plus one kind 10040 note. That’s all that NIP-85 requires.
I suppose DVMs will be useful if I want to make money as a third party WoT service provider.
But if I want to calculate WoT scores myself, using my own personalized WoT relay, is the complexity of a DVM really necessary?

2025-02-01 16:46:38

- reply

Do you think that every nostr client that wants to use PageRank should calculate it from scratch?

2025-02-01 16:16:50

- reply

*precalculation and storage

2025-02-01 16:15:54

- reply

Here’s one way it could play out: either third party services or personalized WoT relays create a big picture snapshot, with generic-ish trust scores for a huge number of pubkeys. This will require lots of raw data and lots of compute. So these are the scores that are in greatest need for recalculation and storage.
Next, we have trust scores that are highly contextual. The opposite of generic-ish. Like who are the experts in some narrow topic. These scores require less raw data, less compute and involve fewer pubkeys. So perhaps specialized apps could handle these calculations. The generic-ish trust scores could be accessed and used as filters so that the raw data used for contextual scores is not polluted by spam.
So two ends of a spectrum: generic trust scores calculated ahead of time with results made accessible for lots of purposes; versus highly contextual, specific trust scores that could be handled on the fly.

2025-02-01 16:08:06

- reply

It’s not gonna be one size fits all. I look forward to seeing the DVM specs!
It is a lot of data. But the question is this: do we calculate WoT scores on the fly just when we need them? Or ahead of time? Right now, we’re trying to do all of them on the fly. I think that’s great for some use cases, but it places severe limitations on what we can accomplish. If we want the notion of WoT to mature, some calculations are simply gonna have to be done ahead of time.
Per NIP-85: there is one event which holds multiple personalized WoT scores (I export GrapeRank, PageRank, & hops, with more scores to be added later) per pubkey. So that’s one event with multiple WoT scores, all concerning Alice, from the perspective of Bob. (“Personalized” means from the perspective of Bob, as opposed to “global” which really just means from the perspective of a big centralized entity like Google.)
We can have third party services to generate and store these events. Right now my app exports these events to my relay. But my vision is personalized WoT relays where Bob will store all of the events that are calculated from his perspective. NIP-85 provides a way for Bob to tell clients where to look for any given WoT score (kind 10040 notes), which can be a personalized WoT relay, a third party service, wherever.
So I see this as basic division of labor. Let each app do what it does best without also necessarily having to provide a complete WoT solution from scratch all by itself.

2025-02-01 15:26:06

- reply

You mean you want to find new people to follow? For now you can go to my app and stratify users in your extended Follows Network by their GrapeRank or their PageRank score. Filter out the ones you already follow (set minimum hops = 2), check out the high scores and decide if you want to follow them. People like @npub1clk…pup9 have told me they particularly want to look at their followers and maybe follow some of them back, but they don’t want to wade through thousands of bots.
NIP-85 allows these scores to be exported so other clients can use them for that purpose plus lots of other purposes.
https://grapevine-brainstorm.vercel.app/

2025-01-31 23:03:02

- reply

fyi I’m going to be issuing kind 10040 notes soon as per NIP-85.

2025-01-31 17:32:45

- reply

Haha yup me too

2025-01-31 17:25:06

- reply

I just looked up pubky on X. Looks like John is speaking at El Salvador today. And I see a short clip from John talking about pubky saying he’s learning from bluesky and nostr (what precisely, idk) and that pubky is being tested internally.
But no mention of any actual users. Or who are his customers and why they will want to use his products. Admittedly it was a short clip. But my impression is too much talking and not enough listening.

2025-01-31 16:58:03

- reply

I’ve listened to one or two of these podcasts with John. I like a lot of his ideas, I think he correctly identifies a lot of problems and correctly identifies many of the tools we need to solve those problems. But what exactly he wants to do with those tools was always too vague.
Maybe his vision is coming more into focus. Looks like the products are getting fleshed out. But is anyone actually using pubky?

2025-01-31 16:50:01

- reply

I like that pubky has a semantic social graph and uses neo4j to power it. But my fear would be that pubky is falling into the trap of trying to build a cathedral, making the requisite underlying structure like their DHT more rigid than it has to be, which is a major critique in my mind of projects like IPFS and bluesky.
I’m guessing pubky users truly control their private keys and their identities, which is true for IPFS but untrue for bluesky which (I think?) holds users’ private keys.
The biggest strength of nostr (in addition to the fact we hold our private keys) is that nostr is resisting the cathedral trap, better than any other protocol that anyone is actually using.
But the nostr ecosystem lacks a semantic social graph powered by a mature graph database like neo4j. This really needs to happen.

2025-01-31 03:34:54

- reply

Why golang? Remind me which relay you’re using - one of the ones in rust?

2025-01-31 03:22:42

- reply

Have you thought yet about exporting your WoT scores so they can be imported by other clients? I just yesterday made a first pass at exporting GrapeRank, PageRank and degree-of-separation scores as kind 30382 events (one event per pubkey, almost 90k in all!) following @npub1gcx…nj5z ‘s proposed NIP-85, Trusted Assertions. See my comments starting yesterday at the end of the PR discussion: https://github.com/nostr-protocol/nips/pull/1534 . @npub1xdt…ntxy has also issued Trusted Assertions. Not sure who else. Would be cool to have more!

2025-01-31 02:45:36

- reply

I keep flirting with the idea of hiring a team that specializes in neo4j ETL pipelines to bang it out. Like graphable.ai maybe. Know anything about them?

2025-01-31 02:43:19

- reply

So are we working on (or soon to be working on) pretty much the same problem: LMDB-to-neo4j sync? You’ve started in rust and moving to golang. I’m most familiar with nodejs but learning some C++ as I dig into strfry. So we’re working in different languages. But it would be nice to build a common set of tools that could be readily adapted by any relay dev who wants to work with neo4j.

2025-01-31 01:00:24

- reply

I’ve never heard of RabbitMQ; learning about messaging queues now

2025-01-30 23:51:24

- reply

Help me understand what you mean @npub10mt…27hf by a persisted queue.
Here’s what I’m thinking: you have a sql database with a column for lmdb key or event id and a col to indicate whether the event has been imported successfully into neo4j.
Or maybe I’m way off base and there’s no need for sql (which would be a THIRD database including LMDB and neo4j)

2025-01-30 23:13:41

- reply

relays, golang 👉🏻 cc @npub10np…tl5h

2025-01-30 22:46:33

- reply

Thank you! There are some techniques here that I’ve heard of but never used and I need to learn them! ⚒️🤓

2025-01-29 21:50:09

- reply

https://i.nostr.build/xln1wViHYcpuS74k.jpg

2025-01-29 13:35:15

- reply

GM Shawn! ☕️

2025-01-29 05:44:29

- reply

btw if you look at your extended Follows Network, there are about 180k pubkeys and about 8k follow you:
https://grapevine-brainstorm.vercel.app/#/profile?npub=npub18yvpnchj7yaepjk8yz2pn66hfmmup505aqvx0lpyc3aree0g5fyq8clpz3
(You may notice your follows count is wrong — that’s bc my neo4j follow numbers are currently 4 or 5 days behind bc of a bug that i just fixed. Hopefully will catch back up soon.)

2025-01-29 05:37:11

- reply

How’s this one? https://i.nostr.build/s5uItkCCpQc84Bfo.jpg

2025-01-29 04:24:33

- reply

In a way I’m currently doing the same thing as you: start with my npub and then expand out. Except I’m using follows rather than engagement, my reason being that it’s easier to implement (at least that’s how it seems to me). Eventually GrapeRank will ingest data from all sources and calculate a wide variety of scores, including engagement data for an engagement score.
The first step in GrapeRank is *interpretation* of data into a format ready to be digested by my algorithm, and it sounds like that’s exactly what you’re doing as your first step, e.g. you *interpret* a reaction in one way, a zap in another way (they “mean” the same if I understand correctly, but carry different weights).

2025-01-29 04:03:41

- reply

This is why we need personal WoT relays: so we can all set weights according to our individual values, preferences, beliefs, etc.
Of course, not everyone will want to muck around with the details. Which means we’ll need default values. But no set of default values will make everyone happy, which means they’ll need to be configurable for those who wish to do so.

2025-01-29 03:42:52

- reply

👍🏻
Do you put the relay and neo4j on the same server or separate? I’d think the same server might theoretically be more efficient for communicating gigantic volumes of data (or perhaps websockets are good enough for even huge volumes of data?) but for smaller volumes like what you’re doing I suppose separate servers would make more sense.

2025-01-29 02:49:31

- reply

Are you using strfry plugins? Seems to me that would be a useful way to close the loop: put relational metadata into neo4j, use that to update preferences regarding what content you want to keep and what to discard, use a combination of REQs and strfry plugins to implement those updated preferences, feed new data into neo4j, and the cycle is complete.

2025-01-28 21:34:03

- reply

Ah, cool. Alas, I am on iPhone!

2025-01-28 20:47:15

- reply

What’s your plan for what you’re building? Any particular hurdles you’re facing?

2025-01-28 18:20:46

- reply

I spy nostr + neo4j! 🔥👏🏻💪🏻🔥

2025-01-27 03:55:32

- reply

Do you have in mind a mechanism to flag dishonest / scammers / bad actors? NIP-56 reports or NIP-32 labels would be a possible place to start.
I suppose it would be simple enough to calculate the number of successfully completed projects score, something similar to what eBay does, but do a GrapeRank-weighted sum rather than a simple sum to screen out fake projects.
This reminds me that I’ve been wanting to export WoT scores using @npub1gcx…nj5z ‘s NIP which you could use to import GrapeRank scores for the above calculation.

2025-01-24 14:53:55

- reply

Awesome 💪🏻
Almost half a billion total events in your db, with almost a million new events coming in per day! Those are the stats I was looking for. We’ll need a neo4j pipeline that can handle those quantities at a bare minimum. 🛠️

2025-01-23 21:46:05

Stats.nostr.band api appears broken - js console indicates a json syntax error
(idk if this is new or not)
@npub1xdt…ntxy

2025-01-22 19:09:26

- reply

I’m getting all zero scores now

2025-01-22 18:34:37

- reply

The desire to procrastinate can be a powerful motivator
If you learn to harness it you can get a lot done
Like mace windu harnessing the dark side to serve the light 😂

2025-01-22 18:14:10

- reply

Awesome progress 🔥
I few bugs but it looks like it’s calculating GR scores for one of my test accounts 👍🏻
Really looking forward to exporting the list to nostr and using it on other clients! But having trouble - console says no relays accept the list. I wonder if it’s rejecting bc of the additional data in the p tags? Bc I CAN get the list published if I change filter settings so the list is empty.

2025-01-21 15:18:41

- reply

Are you saying it’s a popularity score when you’re basing it follows but not when you’re basing it on zaps? I’m not convinced.
I’m trying to get you to face what is apparently a difficult question, and I feel like I’m pulling teeth. Why should we care about the PageRank score in any of its variations? What does PageRank measure? It’s not enough to regurgitate the equation. You could point out abstractly “it’s useful and that’s all you need to know” but that’s avoiding the question. Boil it down to its essence in a way that is substantive, meaningful and that makes sense to a pleb. I say it measures popularity. You say it measures … what?

2025-01-21 04:39:13

21 minutes left to free Ross
Unless “day 1” means within 24 hours of inauguration

2025-01-21 04:00:49

- reply

In high school we called it toxic hell and I can’t stop calling it that in my head

2025-01-21 00:19:23

Would you use a measuring stick without knowning what it measures? Would you use a tool without knowing what it does?
Empirically, PageRank is a powerful tool. It’s great at solving a very real problem: spam. With PageRank, you calculate a score for various sources of content (URLs, pubkeys) and rank them from top to bottom. Spam tends to sink to the bottom, leaving non-spam at the top.
Although PageRank is imperfect — not immune to manipulation — I enthusiastically support its application to nostr. Surprising we haven’t seen more of it already. Nostr.build started using a variation of PageRank long ago. @npub10r8…t2p8 will use it soon.
Here’s the question: how does PageRank work? what does the PageRank score mean? What does it measure? The answer is simple: POPULARITY. Which explains its utility: spammy content sources tend to sink to the bottom of popularity contests.
But we need to beware the corrosive attributes of putting popularity on a pedestal. The last thing we want to build with nostr is another influencer/attention economy that values popularity above all. I advocate the use of PageRank, but I also advocate that we should look beyond at how we can do better.
If PageRank measures popularity, I propose we make a concerted effort to ask: if not popularity, what SHOULD we be measuring? What do we WANT to measure? And for what specific purposes do we want to use these measures?
Here are my answers: we should measure context-specific Value. Merit. Skill. Expertise. Quality. Things that reflect fundamentally intrinsic attributes of people, places, things. Objective truth, as judged from trusted perspectives. And the scores we come up with should be suitable for use as weights for weighted averages and weighted sums.
We should start with PageRank, use it, play with it, all the while asking: how can we modify it so that it measures value and merit over popularity? How can we make it contextual? And are our scores suitable — better yet, optimized — to be used as weights?
GrapeRank is what I come up with when I follow this thought process to its logical end. But I really don’t care what anyone calls it, or if my thought processes are followed to a tee.
Let’s just make sure we don’t blindly reproduce the mistakes of legacy fiat social media. Truth >>> popularity. We’re building freedom tech, goddam it.

2025-01-20 19:28:26

- reply

GrapeRank is designed to use whatever data is available, for whatever context you wish to calculate. You can use multiple data sources simultaneously for one context. If the data quality is low, you can weight it accordingly, bc GrapeRank explicitly keeps track of confidence which is a factor used in calculating weights.
Low quality, highly ambiguous doesn’t mean useless. For many potential contexts, high quality data may be sparse, but low quality data is abundant. I’d put follows and mutes in the “low quality but abundant” category.
For some contexts, high quality, unambiguous does exist. For other contexts, people would be willing to issue high quality data if devs built the tools to do so and if the data could be put to good use. GrapeRank is one such good use.
High quality, low ambiguity can be useful even if it’s not highly abundant. I don’t need every pleb to issue an opinion on topic X if I’m only going to listen to the experts and if there’s only a small handful of experts.
Suppose I want my grapevine to curate a list of NIPs. I might want one list curated by users and another list curated by nostr devs. GrapeRank can (in principle - I haven’t yet coded this up) curate a list of nostr devs right now using NIP-51 lists, filtered by Real Nostr Users. Is there currently a way for devs to indicate their preferences or their approval of NIPs? I don’t know — but if there were, GrapeRank could synthesize their preferences into a curated list, which would be useful even if we only had a handful of devs contributing to it. And if people started to rely on this system, it would incentivize more devs to use it to contribute their preferences, thus increasing the utility of the list. Positive feedback loop.

2025-01-20 18:17:24

- reply

Yes, you could do personalized PageRank, set an arbitrary cutoff, weight everyone above the cutoff equally, and ignore everyone below the cutoff. Crude but easy.
Here’s what I want us to be thinking about: using scores as ranking vs using scores as a weight. These are two very different purposes. Ranking means placing items in order. Using scores as a weight means you’re tallying votes or calculating weighted averages and you want to decide how much weight to give to each pubkey.
How much weight you want to give to each pubkey will depend on the context. Sometimes, I may want to give each “verified real nostr user” a weight of 1, and a weight of 0 to everyone else. In this case you could calculate pPR (personalized PageRank), select and arbitrary cutoff as discussed above. A step function, in other words. Of course the problem here is that the cutoff is arbitrary. What do you do with the 100k pubkeys who have a handful of high quality followers? They’re *probably* not bots, but you’re not 100 pct sure? They’re in-between. GrapeRank allows you to scale up the weight as your level of confidence increases that the pubkey in question is not a bot. You could do that with PageRank except replace the step function with some sort of curve. In which case, you’re halfway to reinventing GrapeRank.
What about the more complicated case where you want a context specific weight which is proportional to someone’s skill? Not set to unity for each normal user? Not clear to me how pPR could accomplish this. Unless you modify it and probably end up inventing GrapeRank.
I should reiterate: my point is not to say that PageRank is worthless. I’ve said many times I’m excited to see it being put to use, like at @npub10r8…t2p8. I think it will add value to the nostr ecosystem.
I’m ALSO looking beyond PageRank, bc I believe we need to optimize it for freedom technology, not just for Google. Measures of popularity are useful but sometimes we’d prefer to measure quality and merit, not popularity.

2025-01-20 17:23:10

- reply

Given the stated assumptions about numbers of followers, if we’re comparing baseline personalize GR and personalized PR scores (not using a contextualized GR score):
Personalized PageRank tells you that Kardashian’s opinion is substantially more valuable than Einstein’s because she’s an influencooor; GrapeRank tells you they are roughly equal because they are both real nostr users as opposed to bots. GR will give MARGINALLY more weight to KK, but this is only because AE’s limited follower count means we are MARGINALLY more confident that she’s not a spambot, compared to AE.
Hopefully that makes sense.

2025-01-20 11:53:29

- reply

Do you have any experience using graph data science pregel to write a custom centrality algo? I’m thinking of doing that for GrapeRank at some point

2025-01-20 02:38:05

- reply

Feel free to chime in with thoughts and ideas! Gonna take lots of work to make personalized nostr knowledge graphs a thing; the more relevant talent and experience we can apply to the task, the better 🧠

2025-01-20 02:23:53

could it? … nahhhhhh 😂 https://i.nostr.build/5beBPoUch0jqGF4a.jpg

2025-01-20 01:58:02

- reply

Do you have any experience with LMDB? Most nostr relays use LMDB (either exclusively, like strfry, or as one of several options like khatru) so a set of performant ETL pipeline tools from LMDB to neo4j would be valuable to the nostr ecosystem

2025-01-19 23:15:29

The neo4j graph database query language, cypher, can add nodes one at a time using a command like:
MERGE (n:NostrEvent {id:”abc123”})
But for a large number of nostr events, it’s much more performant to create a csv file, one row per event, and load them all at once with a command like:
LOAD CSV FROM 'file:////neo4jExport.csv' AS row MERGE (n:NostrEvent {id: row[0]})
In addition to the event id property, other properties can be included, e.g.:
pubkey: row[1], created_at: row[2], kind: row[3]

2025-01-19 20:35:29

- reply

Yup. PageRank was magic compared to what came before it, but it can definitely be gamed!
People will do what they can to game GrapeRank. But the contextual nature of GrapeRank makes it more agile than PageRank. My expectation is that in many cases, the effort it takes for the community to counter and neutralize an exploit will be less than the expense incurred by a bad actor engaged in the exploit. Exploits will become less profitable and therefore less of a burden on the system.

2025-01-19 19:55:45

- reply

What is the trade volume for $TRUMP?
I could create my own memecoin, issue 100 trillion coins to myself, sell one coin for 1 penny each to a few frens and celebrate that the memecoin market cap and my net worth is a trillion dollars.
Problem is that my coin is thinly traded, so the “market cap” of one trillion is basically meaningless.

2025-01-19 19:44:39

Ok, back up now #note1p8r…3ly2

2025-01-19 19:40:28

- reply

Absolutely! The Grapevine is designed specifically to reflect the values and beliefs of YOU, the end user.
If you trust zaps more than follows, or vice versa, then adjust the relevant weights accordingly. The community will need a developer to write the script to ingest follows and zaps into the GrapeRank system, but once that’s done, the average user (who is not necessarily a skilled developer) will have the power to customize his or her Grapevine according to his or her values and beliefs.

2025-01-19 19:26:04

Looks like my GrapeRank server crashed again. 😝
Gonna do a small upgrade and restart, will take a few mins

2025-01-19 19:22:25

- reply

Interpretation means take whatever data is available to you and translate it into a ratings format that is ready to be digested by the GrapeRank algo.
For example: if Alice follows Bob, I INTERPRET the follow AS IF she had issued a rating in the GrapeRank format. Which she didn’t, of course, but that’s why I call it “interpretation.”
The format requires 5 fields:
- context (string)
- rater (string)
- ratee (string)
- rating (number)
- confidence (number between 0 and 1)
At my site right now, every follow and every mute is INTERPRETED as a rating, issued by one pubkey to another pubkey. The context is something like: Real Nostr User. The rating field is a 1 or 0 for follow or mute, respectively. The confidence is 0.03 or 0.5 for a follow or a mute, respectively.
The final GR Real Nostr User influence score is a number that is suitable to be a weight in a weighted average (eg, to calculate ratings at Yelpstr if such an app were to exist). It is a number between 0 and 1, where 1 means “verified Real Nostr User.”
If we were to use PageRank to calculate average scores of businesses at Yelpstr, the opinions of the K Kardashians of the world would dominate. If we were to use the GrapeRank “Real Nostr User” score, the opinions of K Kardashian and A Einstein would carry roughly equal weight.

2025-01-19 19:01:02

- reply

If your goal is to put pubkeys in order, PR and GR can both serve that purpose in a meaningful way. May as well do the one that is easier to calculate.
If your goal is to do anything that requires a pubkey-specific WEIGHT, such as tallying votes or doing a weighted average, GrapeRank is better than PageRank because PR is inherently a popularity contest, and most of us don’t want weight to be proportional to popularity. Most of us want weights to be proportional to merit or quality, in the relevant context.
GrapeRank is inherently designed to create contextual scores. One context might be designed to distinguish real users from bots, ie a binary outcome (this is the GR score currently available on my site). Another context (I have not yet implemented on my site) might be designed to reflect a pubkey’s expertise in some topic on a scale from min to max. Either of these scores are suitable to be used as the pubkey’s weight in the relevant context.
It’s not clear to me how to do this with classical PageRank. Unless you modify it, of course.

2025-01-19 18:47:22

- reply

A high variance does not imply GR is not useful. I’d argue the opposite: it implies the GR score conveys information that is not conveyed by the DoS score alone.
A low variance (tight correlation) between the PR score and the DoS score would imply we may as well just scrap PR and use the DoS score which is easier to calculate. Although I don’t think the correlation is tight enough to suggest we should do that.

2025-01-19 18:41:37

- reply

ty!
Kept crashing almost immediately after opening it.
I suspect it was due to a malformed kind 3 note with empty content field (content: “” ), which also crashed my personal desktop client. Crash kept happening even after I fixed and updated the note. Maybe the malformed note got stuck in damus cache? Only way to fix it was to delete and reinstall.

2025-01-19 18:37:59

- reply

Correct. I am not claiming GR to be “too good to be true.”
I would counter that PageRank is likewise not “too good to be true.” It’s famous but that doesn’t mean we should blindly accept it as the only tool in our arsenal.

2025-01-19 18:33:46

- reply

My critique of PageRank is not that it uses follows, as I stated a few posts earlier in this thread.
The ratio of K Kardashian’s PageRank score divided by A Einstein’s PageRank score (the example from my prior post) is very high. I suspect it’s unbounded as Einstein’s follower count remains fixed and Kim’s score continues to grow. This is true of global as well as personalized PR.

2025-01-19 18:27:01

Can someone remind me the script to reinstall zaps to damus? I had a to delete and reinstall the app bc of a bug.

2025-01-19 18:25:42

- reply

The average GR score decreases as hops increases, as it does with PageRank. Distance plays a role with both.

2025-01-19 18:18:24

- reply

They’re bounded after you normalize it, sure. That doesn’t change the fact that it’s effectively a popularity contest. It’s the relative scores that matter, and ratios don’t change with normalization.
Suppose A. Einstein is a pleb with 20 or 50 or 100 high quality followers. Enough to prove his profile is probably not a bot. K. Kardashian is an influencer whose follower list is 10k or 100k or 1M and continues to grow by leaps and bounds.
What is the ratio of Kardashian’s PageRank score divided by Einstein’s PageRank score? It’s really high, that’s what it is.
PageRank is a great measure of popularity, which is a valid thing to measure and has its uses, but a poor measure of contextual merit or qualification.

2025-01-19 17:30:13

- reply

How’d you boost the ranking of the 170 websites?

2025-01-18 23:42:43

- reply

I agree with you regarding global ranking. This is why I calculate personalized PageRank, not global PageRank. My understanding is that @nprofile…rp63 and @nprofile…mus0 are doing the same, calculating personalized PageRank.
GrapeRank is also personalized, not global. And one of the GrapeRank parameters allows you to deemphasize rankings based on number of hops away.
If you truly want to completely ignore anyone more than 2 hops away, this is why my site also calcuates your Follows Network. For any given user, you can see whether that user is 2, 3, 4, 5, however many hops away.
So that's 3 scores: PageRank, GrapeRank, and number of hops away. None are global scores; they're all personalized, meaning they're relative to you.