pippellia
Pip the social graph guy

pippellia.com

I demystify the social graph so you can learn faster and use this powerful tool more effectively.

Public Key

npub176p7sup477k5738qhxx0hk2n0cty2k5je5uvalzvkvwmw4tltmeqw7vgup

NIP-05 Address

pippellia@verified-nostr.com

NIP-57 Address

whitebat1@primal.net

Profile Code

nprofile1qqs0dqlgwq6l0t20gnstnr8mm9fhu9j9t2fv6wxwl3xtx8dh24l4auspzemhxue69uhhyetvv9ujumn0wd68ytnzv9hxgqg3waehxw309ahx7um5wgerztnrdaks24vx7p

Publishing to

relay.nostr.band

nostr21.com

relay.damus.io

Author Public Key

npub176p7sup477k5738qhxx0hk2n0cty2k5je5uvalzvkvwmw4tltmeqw7vgup

Show more details

Last Notes

2024-10-02 08:11:40

- reply

I visited Munich quite a few times the last year.
Amazing city, super green, really good quality of life, not that more expensive than north Italy with the exception of houses

2024-09-18 11:57:56

- reply

what do you think about Pow endorsements?
https://pippellia.com/pippellia/Social+Graph/Navigating+the+social+graph#PoW+endorsement
tldr;
- Alice pays 10$ to a miner
- Miner mines a note that reference Alice's npub
- Alice gets her checkmark so she can defend her audience from impersonators

2024-09-16 14:14:50

I'm not sure how to continue my programming journey, so I'm asking here:
I've only ever used Python, and I love it, except when it's slow (which is *always* when you're interested in writing very efficient graph algorithms like I am).
I like simple, clean code, I've mostly never used objects or classes, they're not particularly useful for what I'm doing.
So, after some research, I am thinking about learning C or Go. The thing is, C is much more widely used and supported, but I don't want to become a hardcore developer who has to manage memory and shit. I want to be a mathematician who can write good code, that's all.
Also, I like that Go makes concurrency easier, at least that's what I've read, which is something I'll need to use a LOT because of the particular algos I have in mind.
What do you guys recommend? cc @npub1fjq…leku

2024-09-14 08:47:16

GM 🚂
https://m.primal.net/KpXv.jpg

2024-09-03 09:43:04

- reply

fedimit is an open source project. You are talking about Fedi the app developed by Fedi the company.
Fedimit solves many interesting problems related to consensus, that are absolutely non trivial to solve, that's why I think it's silly to try to recreate them.

2024-09-03 08:34:49

- reply

fedimit exists too. I don't see any reason to hack something into becoming a consensus machine when there is already a battle-tested implementation.

2024-09-03 07:57:43

- reply

not a shitcoin, it's really fun, perhaps even more fun than regular chess because it avoid opening theory that kinda ruined chess

2024-09-03 07:50:33

- reply

What's the difference with Fedimint?

2024-09-02 12:12:39

- reply

I don't understand this part
> On the other hand, if the user waits until the last moment8, the cost could be nearly zero, at the risk of missing the expiration time.
How can it be zero? The ASP will take that 0.23% upfront as a fee for the service of providing liquidity.

2024-08-31 23:10:16

- reply

lovely. I would appreciate a behind the scene, to understand what I had in mind when composing it

2024-08-30 07:09:22

- reply

ever tried @npub10r8…t2p8 ?

2024-08-29 14:17:08

- reply

Next conf we drink even more!

2024-08-29 07:52:45

- reply

Creating an identity on nostr is basically free. So imagine an attacker creates a whole new network of nibs following each other, exactly replicating the structure of the real network of npubs, same structure.
The two graphs are isomorfic, so any graph analysis will yield the same result. Hence, if you don't have a pov (e.g brand new npubs that don't follow anyone) you can't distinguish between the two, hence my focus on PoW for this case.

2024-08-29 07:49:07

- reply

It's difficult to game it tough. An attacker can pay a miner 10$ to get his PoW checkmark, however, unless he's followed or interact with many other real an trusted people, he'll lose that initial weight in algos like Trustrank.

2024-08-27 09:06:35

- reply

it's not rocket science
https://m.primal.net/KPPC.jpg

2024-08-27 08:57:58

- reply

Yeah, in some situations Dave is the bad guy, but u are right that more generally it's Eve

2024-08-27 08:05:29

Dave, being the real bad guy, is free to this day. There is a lesson here
https://m.primal.net/KPLn.jpg

2024-08-26 10:57:32

and parties are superior
#nevent1q…pw4l

2024-08-20 18:10:49

- reply

> If you could solve name-squatting would you find it useful?
yes, but my guess is that it's unsolvable. Why? because it requires someone/something to know that "Apple" it's one npub and not all others, so it has to have knowledge of meatspace, of what "Apple Inc" the company listed on the stock exchange is. Such a protocol would have to have knowledge outside of its boundaries, and in some cases should rely on trusted third parties like in the example with stock exchange, which become trusted.
Also, the use case you mention can be solved by adding the QR code of the npub next to the name, or by looking at the set of npubs next to the platform/that is interviewing the creator (I prefer the first)

2024-08-19 11:56:06

- reply

cc @nprofile…pyug

2024-08-14 16:48:24

- reply

I see. Don't you think devs are judged by the code they write in a similar fashion?
Maybe the big demand for Devs account for it

2024-08-14 16:37:08

- reply

I get that rewards are fewer, but why stakes are higher?

2024-08-12 18:17:44

- reply

What if I take an ASIC, make it mine, and people can pay 10$ to me to have these PoW keys or even better these PoW endorsements (so you can use the same old key), in order to get a "golden badge" displayed next to their name on clients, as a form of anti-impersonation.
Like, I would pay for it, and other would as well I think.
And if the market for anti-impersonation service gets big enough, it can actually work at scale.

2024-08-12 17:46:27

- reply

okay, so just for fun. I thought you did it as an anti-impersonation mechanism

2024-08-12 17:27:00

- reply

oh, thanks for sharing this interesting story, and sharing the code I used.
However, I was more interested in why did you mind that key instead of using a random one?

2024-08-12 15:28:51

- reply

genuine question:
why did you mine your npub so that it ends with 'mleku'?

2024-08-12 15:03:48

- reply

> turning elliptic curve signatures into a new inflating monetary asset
this is not what I am saying; What I am saying is mining over a nostr event.
the event:
{ target: npub1xxxxxxxxxxxxxxxxxxxxxxxx
nonce: 59h998543...
}
If you sha256 this event, the result starts with a certain number of zeros.

2024-08-12 15:01:17

- reply

> That sounds like Damus leaving sats on the table.
idk, maybe. Not every service is directly monetizable I think, even in a bitcoinized world. Plenty of examples from non-bitcoin product and companies.

2024-08-12 14:35:43

- reply

wo wo calm down cowboi lmao.
What I have in mind is to use
npub_pow / total_pow.
So, if a miner heavily attacks this system, it will increase the denominator. Unlike shitcoins that need to mantain global consensus and are "defended" by the total pow online at that moment, this considers the total pow over the whole history.
So it's a strict improvement.
Then by re-using SHA256 one can use ASICS for Bitcoin (I think) and offer these PoW endorsements I talked about here https://pippellia.com/pippellia/Social+Graph/Navigating+the+social+graph#PoW%20endorsement.
So, you mine blocks with the ASIC, but when the price people are willing to pay for PoW endorsement is slightly higher, you go and mine that. This should result in having basically the same price for hashrate.

2024-08-12 14:26:48

- reply

yes, they would, but not every service is *directly* monetizable.

2024-08-12 13:48:06

- reply

more generally, you can use someone else's curation as your own only to the extent you trust the curator. PoW has the benefits you don't have to trust anyone, which is I argue an important property if you take the PoV of service providers like relay operators.

2024-08-12 11:53:09

- reply

yes, and in some contexts this is another tool in the wot toolbox.
However, any payment or zap is signal *if and only if* one is absolutely sure the sender and recipient aren't colluding or aren't the same person.
So, if you already trust Fran as moderator, then you can assign a non-zero level of trust to anyone who:
1) has posted on Fran community
2) hasn't been removed
However, this system is a bit fragile when you try to use it as an anti-spam measure for relays for example.
Why? because one moderator (like Fran) becoming malicious is enough for open up a DDoS attack vector.

2024-08-12 10:25:55

- reply

Generally agree. My take on how I use it quite effectively with my work:
1) List generator. I am (and I would argue most humans are) kinda bad in making long lists of things, so I ask the LLMs to generate it for me like:
"rewrite this phrase 20 times".
"Give me 20 synonyms for this word"
"Give me a list of 10 inequalities I can use here to prove this result"
2) Do simple things I don't want/don't have time to learn. For example I wanted to make a simple redirect from pippellia.com to the Obsidian link. Don't know shit about the web, so I asked and it did the job.
3) Replace Stack overflow / looking at the documentation for beginners questions like
"What's the pythonic way to write this"
"how can I get the position of an element in a list"

2024-08-12 10:12:08

- reply

all of the things you mentioned requires time and social interaction, which defeats the purpose of having a disposable key that you can use only once and then throw away.
If we are talking about long lasting identities, then yes these approaches are good solutions.

2024-08-11 19:38:09

- reply

lol

2024-08-11 19:37:48

Pineapple on pizza is disgusting.
Mushrooms on pizza are acceptable.
(I am Italian so you can either agree with me or being wrong)
#nevent1q…6q3t

2024-08-11 19:29:51

- reply

+1 on this question

2024-08-11 19:21:04

- reply

well, if you are a customer, you can and should ask for features you need.
If clients don't have sound business models, and so it's not clear who are the customers, well that's unfortunate but it's not the users' fault.

2024-08-11 16:32:34

- reply

which action?

2024-08-11 14:58:22

- reply

doesn't add anything. The thief can post something outrageous with your nsec and 2s later can steal the money.

2024-08-11 11:42:44

I've watched the Fedi annoucement video and it's cool and well made and inspiring.
But still I have no clue about how Fedi the company is going to make money.
Maybe they'll use their earned reputation to build a Bitcoin bank? No idea, but I've always wondered how companies like that can go years without any monetization avenue. Do you have more clues than I have?

2024-08-10 11:50:34

- reply

sounds really promising. Do it bro

2024-08-10 11:37:34

- reply

why did you copy plan B tweet?

2024-08-09 19:40:55

- reply

Ops, I re-read it. PoW keys / endorsements solve the reputation part I believe

2024-08-09 19:11:36

- reply

what do mean exactly with reputation?

2024-08-05 23:01:28

- reply

a) It's not volatility, it's long term trend into nothingness
b) why creating a new money supply for paying for storage? money creation is tight to ethics.
c) I haven't come up with a solution (yet), just pointed out that this one is stupid
d) If Bitcoin were just a store of value, that alone would make it the most valuable asset because the inability to save IS the problem of this century

2024-08-05 07:00:36

- reply

Permanent data until this token goes to zero
https://image.nostr.build/21da0ce3147979324bdb007e03f5859010e69efd835dec1f8946c0419bf33fa2.jpg

2024-08-04 15:02:06

- reply

It's like using a tank to move a piece of furniture in your living room

2024-08-04 13:14:45

- reply

Blockchain to solve spam? Sounds stupid

2024-08-03 17:50:50

- reply

yes, go for the nip

2024-08-01 18:12:58

- reply

and some of these metric can be improved with training, so are we going to create a category for those that don't train? lmao

2024-08-01 16:08:35

- reply

#nevent1q…ja3e

2024-08-01 16:07:28

- reply

great idea, why using common sense when there is *clearly* a way to measure the impact of every body measure like:
- weight
- muscle mass
- arms' length (which is important in boxing for instance)
- hormones levels
- density of the skin...
Who is more likely to win in a boxing match, someone who is 2kg heavier or someone whose arms are 5cm longer?
You think the solution is making 10k categories? lmao

2024-07-31 11:06:18

- reply

hey, would you like to beta test some recommendation systems?
I'll give you 2 lists of npubs (let's say 20 npubs you don't follow each).
After getting your feedback, I'll tell you how these are derived

2024-07-28 19:00:48

- reply

Nostr and TBD's Web5 have a lot in common. Loosely speaking, DIDs are like npubs and DWN nodes are like Relays.
DIDs vs npubs
- DIDs have the benefits that the Identity itself contains some metadata about the entity, so you could have multiple keys, and endpoints for contacting the entity (endpoints can be your own DWN, which is like a private relay)
- DIDs allow for key rotations, which means they are more secure
- DIDs are more complex
DWN vs Relays
- DWNs are mostly designed for private things, data is encrypted with your DID's key.
- DWNs seem quite complex for me, you can do a lot of stuff with these, like running Apps like you can on Umbrel OS or other personal server OS. So I think about a DWN as Personal server OS + Relay.
So, my take is that I don't know, different tradeoffs that can be summarised in a couple of points:
- more complexity
- more features (are these important?)
- users are required to have a DWN, either self hosted or rented. (Google Cloud is currently helping bootstrapping that network)
- less network effect

2024-07-27 15:23:17

Absolutely beautiful 😍
#nevent1q…s4te

2024-07-26 15:16:18

Nostr can't win without solving spam, discovery and trust.
How? I don't have the complete answer, but I’ve got a new tool for you:
Subrank – Pagerank reimagined for the post-truth world 🧵
https://m.primal.net/JfFb.png

2024-07-26 09:07:19

- reply

I had a similar argument to a dude saying we "need to explain users about Relays and how Nostr works".
It's obviously good if your users know about the technical details, but that can't be necessary.
Like, wtf are we talking about, we don't know how 99% of the tech we use daily works, why would Nostr be different.

2024-07-08 18:50:49

- reply

Yes. Both are protocols, both started fairly decentralised.

2024-07-08 18:48:07

- reply

why?
Doesn't your mom use email?

2024-07-08 18:18:52

- reply

does your mom knows how email works?

2024-07-02 20:09:20

- reply

thanks for the encouragement

2024-07-02 19:42:54

- reply

AHAHHAHAHAH @npub1t3g…upxq and I Will play this in Riga

2024-07-02 19:39:23

- reply

social graph stuff?

2024-07-02 18:28:42

- reply

Thanks for the interest but no, too high capital requirement for that kind of hardware project.
Also, operating from EU is a nightmare

2024-07-02 18:18:30

- reply

yes I am.

2024-07-02 14:17:32

- reply

You got this mate.
Maybe consider asking for someone else to look for it. Sometimes a new look is all that's needed.

2024-07-02 12:38:02

Embrace this
#note1yzp…3azw

2024-07-02 11:40:23

- reply

While checking the route twice, I saw a guy in an ambulance and I thought that there are always people who are worse off.

2024-07-02 11:37:35

GM. yesterday night I lost my physical wallet or someone stole it from me.
Yeah, all documents, cash and cards are gone

2024-07-02 11:29:20

- reply

no, on Lightning you can unilaterally exit, but there are fees to consider

2024-07-02 09:36:15

- reply

I bet my morning was worse than yours.
My physical wallet was stolen/lost 🤣

2024-07-02 09:15:58

- reply

numbers wayyy to low. Nature will heal when that rookie numbers reach +90%

2024-06-27 11:55:12

Ratings bigger than 1 can lead to non-convergence, but even if all are normalized, it can still diverge.

2024-06-27 11:27:09

- reply

As is, you can't rewrite as a linear problem like I = A * I
where I is the Influence vector.
Before jumping an trying to compute it, you should prove that there exist a solution, and that it is unique.
Otherwise, if it's not unique, you run the algorithm and you get one solution, you run it again and you might get another valid solution, which is completely different from the first.
That's non trivial to do I think. I'll give it a shot but I am not particulary optimistic

2024-06-27 11:19:27

- reply

That's right.
I think this is it. I suggest you add it to the article, I think other people might find it useful.
https://m.primal.net/IzkY.png

2024-06-27 08:28:51

Yesterday was my last day at Geyser. From now on I'll work on social graph shit.

2024-06-27 08:26:46

- reply

DAMN IT. @npub12vk…pugg @npub1w0r…cu4x and @npub1nz6…0w5y I wasn't able to upload a .svg image using your clients.
Amethyst showed "Error 500 failed to upload the image"
https://m.primal.net/IzdD.png

2024-06-27 08:24:43

- reply

It didn't upload the fucking image 😕
Let me retry @npub1u5n…ldq3

2024-06-27 08:17:47

- reply

Hey, I've read the post and I've worked out the final formula that you can see in the image. Can you confirm it's what you had in mind?
- I(v) is the influence of node v in the graph
- N-(v) is the set of nodes that rated node v
- r(w) is the rating node w made about node v

2024-06-26 20:03:36

- reply

how do you compute proof of liability?

2024-06-24 17:31:16

- reply

@npub1fjq…leku
@npub1nxa…a6q8
I have been busy designing an algo to approximate the local Pagerank based on the global Pagerank.
This means massive speedup because you compute the global Pagerank only once a day, and then use the approximation for each users.
I think this will become a new cool building tool for WoT stuff. Publishing likely next week.
(example, you are searching someone, the results can be ordered by this metric)

2024-06-24 16:41:49

- reply

Inverse square of the distance, yeah, that's a reasonable model, because there is some parallel with how other things work.
btw (1/x)^2 and 1/(x^2) are the same thing.
However, the biggest discussion for me is whether there should be these scores at all, and I am in the "No" camp.
If you define one particular problem (like impersonations), you can solve it without scores that can alter social dynamics (think how horrible it would be to walk around in meatspace with a number on your forehead).
So I would remove followers count, zap count, mutes count, WoT scores. Show people not numbers

2024-06-24 16:34:37

- reply

nope, inverse square (1/x)^2 behaves quite differently from a log

2024-06-19 14:28:58

- reply

Interesting 🤔
So, the WoT filter allows for people up to 2 hops away, unless they are connected to you only by someone that has too many follows, right?

2024-06-16 09:12:27

- reply

Hey @npub1zte…wcyy, great article.
First, I am all for using spontaneous user interactions as input for discovery algorithms, as opposed to NIP77 where the user is asked for a trust judgment that will likely require bootstrapping a new type of user interaction.
Second, I like zaps because they carry a value signal (assuming the sender is not colluding with the recipient).
I think your proposed algorithm is part of a more general class of algorithms that I've described in my
Navigating the Social Graph https://pippellia.com/pippellia/Social+Graph/Navigating+the+social+graph
Here is how I see it:
- The input is some social graph whose edges represent some form of social relationship, such as who follows who, or, as in your case, who zapped who and how much.
- some propagation rule is applied (in your case, it's direct propagation)
- calculate a certain set of candidates
- apply a sorting rule to sort the candidates (this can be pagerank, local pagerank, trustrank, zap weight...)

2024-06-12 09:02:40

- reply

https://m.primal.net/IhnN.png quite strange descriptions

2024-06-12 09:01:03

- reply

you are showing too many details at the start of the chat.

2024-06-10 22:27:00

- reply

well, the candidates grow literally exponentially.
I follow directly follows hundreds of people, but if we expand at people that are distant 2 hops we get to 10k-ish. 70k is 3 hops, and then you start to saturate because Nostr is still small.

2024-06-10 19:21:58

- reply

basically nothing
btw I already sent Pablo a video recording showing the issue

2024-06-10 19:13:23

- reply

Thank you!
Oh I like threads, as they:
- require to be concise
- show images where you want them
I tried to post it from Highlighter but it didn't worn for some reason, so I had to do it the manual way lol

2024-06-10 19:11:46

- reply

will take a look!

2024-06-10 18:39:50

- reply

If you liked this 🧵, follow me for more of these and consider retweeting it to share it with your audience.
If you are a builder that wants to integrate the social graph into your project, reach out to me.
I would be more than happy to help.
note1jwu5qpyh02uk7neyj4tn2aq6fq5amtsv67t59xkffyqekm65lfmqq6y62e

2024-06-10 18:39:22

- reply

I want to thank @npub1wf4…dgh9 @npub1zaf…26k5 @npub1pgf…gh8x @npub1t3g…upxq @npub1jkn…tcvj for reviewing before release, and @npub1jlr…ynqn for having me on the @npub1mlc…8jpa to discuss these topics.
https://www.fountain.fm/episode/SJJKS7tr2WI9R4gaVXxO

2024-06-10 18:37:19

- reply

We have a chance to create the freedom web, but to get there we need to understand the tools.
This free website is a way for you to save time and learn about them in a distilled way.
https://pippellia.com

2024-06-10 18:37:00

- reply

It's possible to use the social graph to create ratings that are
- personalized
- relevant
- dynamic
- hard to manipulate
How, you ask?
Well, that's too much for one tweet.
Fortunately for you (wink wink), here is an article in which I summarize hundreds of hours of research on the social graph.
It goes into the weeds, the implementations, and the philosophy behind it.
https://pippellia.com/pippellia/Social+Graph/Navigating+the+social+graph

2024-06-10 18:36:33

- reply

4. Ratings
Ratings today are mostly fake, irrelevant, and censorable.
But they are broken on an even deeper level: you value the opinions of your friends more than those of strangers.
(I meant to tag @npub1dtg…up6m)
https://m.primal.net/IfrR.png

2024-06-10 18:35:22

- reply

As @SvetskiWrites points out in his latest 'Reviewing Reviews', there is another crucial piece of commerce that will be radically changed:
Ratings.

2024-06-10 18:35:07

- reply

When you have an open, public social graph and the ability to filter spam, the incentives change to the point where the advertising industry gets fixed.
Yes, advertising will still be a thing, but you won't be forced to watch ads because you can't be captured.
Companies that optimize for attention will be disrupted by those that optimize for value.

2024-06-10 18:34:44

- reply

You see, the problem is that big tech companies now have a monopoly on this information.
But with open social networks like Nostr, it's all public.
Sounds scary? I'll tell you why it's not.