Liminal and Project Alexandria
Liminal Joins me today to talk about Project Alexandria that is being built on nostr. This is a chunky episode about using nostr in a very different way so strap in, hang on, and get hooked on the ideas presented in Episode 937 of Bitcoin And!
The Project's Wikifreedia: https://wikifreedia.xyz/gitcitadel-project/npub1m4ny6hjqzepn4rxknuq94c2gpqzr29ufkkw7ttcxyak7v43n6vvsajc2jl
The Team:
Liminal:
https://primal.net/p/npub1m3xdppkd0njmrqe2ma8a6ys39zvgp5k8u22mev8xsnqp4nh80srqhqa5sf
Laeserin: https://primal.net/p/npub1m4ny6hjqzepn4rxknuq94c2gpqzr29ufkkw7ttcxyak7v43n6vvsajc2jl
The_Beave: https://primal.net/p/npub1q6ya7kz84rfnw6yjmg5kyttuplwpauv43a9ug3cajztx4g0v48eqhtt3sh
Chip Turner:
https://primal.net/p/npub1qdjn8j4gwgmkj3k5un775nq6q3q7mguv5tvajstmkdsqdja2havq03fqm7
Michael J:
https://primal.net/p/npub1wqfzz2p880wq0tumuae9lfwyhs8uz35xd0kr34zrvrwyh3kvrzuskcqsyn
Find me on nostr
npub1vwymuey3u7mf860ndrkw3r7dz30s0srg6tqmhtjzg7umtm6rn5eq2qzugd (npub)
6389be6491e7b693e9f368ece88fcd145f07c068d2c1bbae4247b9b5ef439d32 (Hex)
Twitter:
https://twitter.com/DavidB84567
StackerNews:
stacker.news/NunyaBidness
Podcasting 2.0:
fountain.fm/show/eK5XaSb3UaLRavU3lYrI
Apple Podcasts:
tinyurl.com/unm35bjh
Mastodon:
https://noauthority.social/@NunyaBidness
Support Bitcoin And . . . on Patreon:
patreon.com/BitcoinAndPodcast
Find Lightning Network Channel partners here:
https://t.me/+bj-7w_ePsANlOGEx (Nodestrich)
https://t.me/plebnet (Plebnet)
Music by:
Flutey Funk Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 3.0 License
creativecommons.org/licenses/by/3.0/
Good morning. This is David Bennett, and this is Bitcoin and, a podcast where I try to find the edge effect between the worlds of Bitcoin, gaming, permaculture, podcasting, and education to gain a better understanding of all. Edge effect is a concept from ecology describing a greater diversity of life where the edges of 2 systems overlap. While species from either system can be found at the edge, it is important to note there are species in the overlap that exist in neither system, and that is what I seek to uncover. So join me in discovering the variety of things being created as Bitcoin rubs up against other systems. It is 6:0:9 pm.
Yeah, baby. Yeah. We're in the evening time. It's the 16th August, 2024 and this is Episode 937, I believe, of Bitcoin. And I am going to be interviewing a gentleman by the name as Liminal. He is part of this thing called Project Alexandria, which is seeking to build kind of a library on the Nostra protocol. It's a little bit deeper than that. Sadly, the first 20 minutes of this interview was completely and thoroughly trashed. That's right. I was using HiveTalk and something really weird happened all of a sudden without me pressing any keys or blinking twice or I don't know calling down the, you know, Beelzebub upon the planet to wreak havoc, it just muted my microphone.
And then I never could get it back. I kept unmuting my own microphone. I thought I could fix it in the mix, but it would just remute my microphone. And I don't know what the hell happened, but when I detached it was an I was unable to recover that 20 minutes of audio. So there's where introductions were done and a little bit of liminals history. But when we came back, HiveTalk was working just fine. And the reason I'm choosing HiveTalk is because I would rather use a pleb built platform and somehow or another pipe them satoshis than to pay another single red cent to legacy anything like Zoom or Google Meet or whatever.
I honestly don't care. I'd rather deal with the heartache and pain and know that I'm trying to support plebs and plebs platforms because that's what Liminal and his crew is trying they're trying to build, right? They're trying to build not just another Nostra client. They're doing something very, very special. It's going to take them a long time to build because they all have other jobs and whatnot like that. But I wanted to get into a deeper mindset of what exactly is going on with this Project Alexandria. And if you don't know, Alexandria is the name of the city in, like, northern Africa, Egypt, I'm pretty sure. Anyway, it had its massive this massive library.
And 1000 of years ago it burnt to the ground and everything was lost. There's conjecture that humanity was set back maybe 500 years by the knowledge that was actually lost in this library. So that kind of gives you a little bit of a little bit of background as to what Project Alexandria is. But I guarantee you that it ends up being much more than a library. So what I'm going to do here is I'm just going to drop you guys right into where Liminal and I picked up the conversation after, I don't know, Beelzebub decided to wreak havoc upon the face of the planet. And you can make your own determination as to what you think project Alexandria is. Alright.
Let's see. I don't know how much of that was lost. So let's do let's just do this. Let's get to the, project Alexandria
[00:04:27] Liminal:
and Yeah.
[00:04:29] David Bennett:
Start talking about what that is and where and, who's working with you on it.
[00:04:37] Liminal:
Yeah. So it's basically a we we call it the the Noster library, and we got, Stella, ChipTuner. We got there's there's quite a few people on here. Michael j. Yeah. Yeah. So I I mean, there there's quite a few, and it's difficult for me to remember everyone. But we have about, like, about a 10 10 people 10 person team on Get Citadel itself. Yeah. Some are working on their own our own Nasser library. And with Alexandria, there's the aspect of displaying these specific notes from the spec that we have. But, ultimately, we call it Nasser library, but I also call it a knowledge base where, you know, using using, the features that Noster have Noster has, I think about you have a private relay and you're you know, if you're interested in biology and another group, they're interested in chemistry, they can have their own private relays, where they can put whatever they want, and they don't have to worry about being intruded by anyone else that's not part of their group. But their content is on for this is up for display and to be broadcasted to other to other relays.
And a third a third party, maybe they are interested in computational biology, so they have to pull in stuff from the biology relay, the chemistry relay, and the mathematics and physics relay. And they just, you know, they're just pulling it in and kind of taking and collaborating with the existing notes, whether they are permissioned by, you know, the original group or I mean, if if they are, like, they don't need to actually know the original people. But just by pulling in the notes, they can actually edit it or comment on it. And the but if, you know, someone if those original relays see those notes, they can also welcome them in and,
[00:06:55] David Bennett:
have more collaboration. It's just kind of having it up for display, and I think there's a whole real cool aspect with the monetary layer and how how people can both share knowledge and, get paid for it. Well, now it for for the listeners, the pro project Alexandria I I mean, I presume that you guys named project Alexandria after the library of Alexandria that burnt down centuries ago with untold amounts of knowledge just gone that we so the it seems like the idea if we if if I kinda, like, take that idea, it just rub with a little bit that you guys are trying to build some sort of library that exists on relays, but maybe the first iteration is just so groups can like, computational biology groups can work together, but they can be sure that all of the work that they're doing together is going to kinda remain in some way, shape, form, or fashion if we have our if we have our way about it forever.
Is that kind of where that direction is going?
[00:08:05] Liminal:
What do you mean by, like, our way about it forever?
[00:08:09] David Bennett:
Well, like, the the whole thing with the library of Alexandria burning down and and laying waste of this vast storehouse of knowledge was a blow was, I think, a blow to humanity. We don't know what was lost. And when I look at what you guys are doing, you and Stella and and the rest, I what I think is going on is that you're kind of reconstructing a way to have a library in a completely different way that takes on more of the decentralized notion of relays, that takes on more of a a a notion of we that there's copies on all these different relays. And maybe not a complete copy on every single relay, but would you take the whole thing in aggregate across the network that you would almost be able to rebuild the entire library out of, you know, 10 relays if, like, a 1,000 relays that had parts and parcels of this information got burned down?
Is that that that's sort of what that's sort of the way that I think about it. And like I was telling you earlier before we had audio problems, I may be getting this wrong. And I just I mean, if I am getting it wrong, I don't mind being wrong. I just wanna sure that I I am thinking about this correctly.
[00:09:36] Liminal:
Well, yeah, I think that the relays really offer a level of redundancy to the, to the content that's on there. So, you know, if you wanna share your knowledge out there and other people find it interesting, they can broadcast it to their own relays, and they can also continue the conversation outside of the relay that was there. So, you know, I think about also the NIPS on on Nasr. So right now, it's the the NIPS and and the conversation for Nasser development is on GitHub. But I really don't think that there's a problem with having a private relay that hosts the repository. And the main developers that contribute to that repository, they can develop the specs. They can have conversations on it, and it's completely visible to everyone and also accessible through Noster clients because it's actually really it's nice to have that kind of integration.
And anyone can take what's on there, clone it, edited it, because the the whole point is, yes, you you have interoperability, but you can also kinda do what do your own thing. And the the problem there is with signal. So if if you have a 1,000 different forks of these different specs, well, how do you find the the signal, and how do you find this cohesion? Well, you can always look to the this kind of quote, unquote official relay that has the official specs. And if you want to be interoperable, you can stay there. But if you wanna, you know, have the specs completely reversed and, you know, do some sort of, you know, artistic, you know, Noster client that changes everything around, you can, and you're not really bothering anyone because the the source signal is on the private relay.
Right. It would
[00:11:27] David Bennett:
I guess my question is is, like, you know, I've got a start 9 little server, like a server 1. It's actually I don't I mean, it's small, but it's actually pretty powerful. And I can run a Nostra Relay on it. And I have one spun up, but I've I've got the service stopped because I was only using it sort of as a test. Would I be able to just, like, say, here, here's my Relay, and then have you guys over at, you know, Project Alexandria or Get Citadel say, yeah. Since you're offering, we'll add it to a list, and then all of a sudden, whatever it is that you guys are doing gets copied over on my Relay. Is that is that something that helps, or is it something that's not very helpful? Or because what I what I'm getting at is I don't understand relays the way that I don't understand that I think I don't understand the extent to which you guys are thinking about project Alexandria.
I think I know what a relay is, but I'm probably getting it wrong. So I, you know, I kinda wonder that would be my question. When we're talking about the Nostra ecosystem, exactly what is a relay?
[00:12:42] Liminal:
Is that easily explained? Server. Yeah. I mean, it's it's the I I use the, let's say, Twitter or x. You you have the data layer, which is where everyone's tweets and etcetera go and and is stored. And then you have the client, which takes the data and displays it according to the the specifications that they want to display the the data. And with a relay, it's similar in it's similar in in that sense, but now anyone can spin up a relay or a server. Like, you I I would. My understanding was just, like, keep them, relay and server, more or less the same as a from the same perspective. But so it's just like it the relay holds the data, and it can also broadcast the data. So you can you can connect to any relay relay and look at the data in there, and then the clients will will connect to specific relays. And using the specifications of Noster, they will they will work with the data how how they see fit. So, there's, NIP 1, which displays the standard notes, and now you have all these clients that, you know, it's there's they decide the format, how it's displayed on the screen.
But there's a separation between the client and the relay where the the client developers don't actually need to worry about the relay. They just need to pull in data from the relays, and then you have relay developers that are just working on the on the data on the data layer.
[00:14:21] David Bennett:
Okay. So a relay, it it sounds to me like what just so that I get it right, that a relay really is just a data server, but it's addressable over the web as a public entity. Yeah. So I can address it and say and then start looking at the data, and then other people can address it and pull data from it. Yeah. Okay. So now that we've got a network of these relays running around, that's where I that's where I keep thinking about going back to project Alexandria and then the crazy ass ideas that I've that it's kind of ignited in my head about it.
That's sort of where I feel safer with something like knowledge being the a project like yours that is starting up to provide a long term container for addressable knowledge for the public. And I I wanna make sure that the the that the listeners because, like, there are several times when I've listened to podcasts that I wish that they would just explain it to me like I was 5 because there's so much that I miss on even if it's a simple should be a simple concept. Somehow or another, when I'm listening to it, it becomes something convoluted, and I start, you know, missing messing it up. So that's why I'm I'm kind of like a dog with a bone on this. But one of the things that you said earlier was well, actually, I I don't wanna do that yet. I wanna talk about, like, a specific example because you were talking about, like, a group of people, like, doing computational biology, working together through this project, and then the sum total of what it is that they're working on remains contained in these public addressable server containers.
What about just putting on a book that that already exists? I mean Yeah. How does because that seems to be a a different way of addressing the, project Alexandria. Is that something that you guys are thinking about as well, or is this just for more like group work and contain containment of what was worked on?
[00:16:41] Liminal:
So there are 2 main specs that we that, have been created for kind of knowledge bases. There is what I'm calling modular articles, and it's a way of transforming the famous note taking method, zettelkasten, and applying it to articles. And, for anyone not familiar with that, a zettelkasten is called or the slip box method. Basically, you have an index card with a title, a unique identifier, and then the content, which is a text. And it's supposed to refer to previous content, like, whether that's a paper or another note. You're just you're just referring to it, and it and it's like it's a small enclosed idea about, about something, basically. Right? It's it's enough that the meaning gives rise to itself because you don't need to look at anything else. But, you know, the the ideas are derived from older or other content.
And a modular article basically takes these index cards and it's assembles them in a linear fashion, which that's how articles and books are made. Right? Like, we have you have the introduction section or you have pages. It's it's all it's linear. And but if you can partition it in a in a in a, I guess, meaningful way, whether that is by chapter or by paragraph or maybe even by sentence, you can still compose them together in a sync, like, from beginning to end, and it'll look exactly like an article. But now the difference is that it's very, searchable and indexable because you you don't have to search through a whole, you don't have to search through a whole textbook. And it's it's one of my gripes when I'm reading papers. They'll they'll cite a, you know, a really long paper or a textbook and, like, yeah, this this idea comes from there. And but then, you know, if you wanna verify that, you have to really dive into the source material.
And in this case, if you are actually working with these modular articles, you can specify the location where this the knowledge that you're writing, where it, like, comes from. And it was like a very, almost like a locational citation, but it's it can generalize to other things. It can general like, Stella wrote an article in Highlighter about publishing the Bible or publishing, like the Gutenberg. Right. What whatever whatever's on Gutenberg because you can still work by chapter. You can work by paragraph, and it nothing is really different. It's just that intrinsically, even if it's hidden by the user, but intrinsically, they're all separated and searchable.
[00:19:30] David Bennett:
Yeah. And that's where I'm I'm really happy that that we're having that discussion about where where you're saying because that was one of the first things that I started thinking about when I saw this project was that, you know, I have I have my own Obsidian vault. Actually, I've I've got 2, and I I really shouldn't. It should just be one vault, but I do have 2. And when I'm looking at either one of the vaults, I see all these all this connectedness. You know? Like and you kinda you kinda can't help it as long as you're if you like, because the way that I do it is I'll read a book on my Kindle, and I'll highlight, you know, a passage. And then I've got, Obsidian has a plug in that allows you to download all of the notes that you've take or all the highlights that you've done in Kindle, and it gives you the location. And then that way I can go back, and I can go, okay.
Now take this note and or take this highlight that you did this. You know, what were you thinking? Why did it catch your attention? Is it connected to anything else? And, of course, if I've got, like, you know, 40 highlights from the same book, they're gonna be connected. That's fine inside that one container that is that book, But now I read, like, all kinds of stuff about like, I've got many soil books that I've read, many mycology books that I've read. Yeah. I've I've read a whole a whole book about just carbon. The the symphony and c, if you haven't read that book, it's actually really good. But there's things in the in symphony and c that relate directly to one of these other authors or one of these other books.
And I'll make those connections, but my problem is is that it's all just me. It's all just I mean, which is great because that way, that keeps it separate. But what I really when I saw, project Alexandria, I was like, and the fact that it was being built in Noster, I started seeing the potential for my Obsidian vault to start making links outside to other people's vaults. And then that's sort of where I was getting getting at, what we were discussing earlier. We'd had a whole discussion that got destroyed, but, you know, the mycological networks.
And I'm looking at my Obsidian vault as its own network. And then I know other people have their Obsidian vaults and and, you know, it looks like a network. But if using Nostra, I can make at least one connection to somebody else's vault. All of a sudden, there becomes a huge potential for real knowledge creation to start exploding. And how would that is that I I've talked to one of my good friends, Huddl bod, may has made Korkle, right, Korkle dot social. And we've had a discussion about this. I I'm not sure. It's just interesting to get a different perspective. So knowing what you know about how Noster works and knowing what you know about Zettelkasten, how would it be possible for a note out of my Obsidian to connect with a note in your Obsidian or Rome and not have this end up being noise? Does that make sense? Yep.
Okay.
[00:23:05] Liminal:
So first, I wanna talk about, like, the network aspect. Naspers actually has you can think of very, different levels of networks. There's the the clients that can aggregate different relays together, and, so that that's one way. Right? You can actually by doing that, you're kind of creating a single node where relays all kind of mesh together. And there's also the network of, you know, follow, follow we, where I follow you, but you don't follow me or we follow each other. Like, that's also another type of network, like, almost like a social network. And then there's also the almost individually at the relay level, if you have kind of, as you said, like vaults, right, where only you you or maybe a few people can access. You can have a relay where you have your 12 people talking and communicating, trying to, you know, build something together, and you can do it in isolation.
And then another group can have their own relay, same thing, private. And maybe a few people from both groups are able to to communicate to both relays, or there's a third per a third group that isn't allowed to either of them, but, they're watching. They can read from those relays and pull in that stuff or or, you know, any kind of combination like that. So you have different types of, connections and networks that can actually really work and and synergize together. And with the, yeah, how how do you handle, noise, right, where where we have large language models that people are gonna be using for any number of reasons, whether that's actually making things useful or or flooding people's periphery with noise. So that way, they're just like, oh, well, I I can't even make any sense out of what's going on, so I'm gonna you know, I'm not gonna even care about it because everything is just noise.
And this is where the 2nd spec, NK BIP, 02 comes in where you're working with embeddings. And what an embedding is, it's just a you you have words and you are represent representing it with numbers, which is just a coordinate. You can think about it, like a, you know, x y z. That's a list of 3 three numbers. Well, language models will have a list of 700 numbers to as a as a single coordinate. And that is, the way I think of it is like it's it's an approximate way for how we actually think of meaning or how we represent things. So, you know, and it doesn't it, you know, the specific coordinate changes depending on how how this model is trained and, the specific text it's trained on. But, ultimately, after enough training and after, you know, whatever learning is going on in these models, you can have a word that says, you know, the mitochondria is the powerhouse of the cell. Right? And that's a specific coordinate. Right? Right. But that will also have some sort of proximity to, you know, powerhouse of a city is the electric plant. Right? There there's a proximity to that that is different from saying the there was a fire in in a forest. Right?
The real cool thing about these language models is that they are a computational approximation for meaning. That that's how I kind of think of it. And so in a in a world where we we can be flooded with AI generated noise, essentially, we can actually navigate the space with embeddings and find, related ideas. And, also, if we're just working with Nostra, you don't need to embed everything. You can embed whatever is most meaningful to you. Mhmm. And because it it takes it it takes it takes resources to train your model and to also get embeddings for that or to even work with, like, a lot a large number of notes. Uh-huh. And so if you know, when you pair these kind of partitioned articles, right, like a paragraph, and you have an embedding for that, you can and you you have, say, a 1000 or a couple thousand other notes, you can you can really see what what other notes are actually semantically close to it. And and this works.
It's it's like, it's the there are 2 types of ways to navigate. Right? You can you can navigate with direct links, whether it's a citation or just pointing to pointing to a source, which is what, like, Obsidian works with. We have we have hyperlinks or tags or anything like that, and that's nice. But there's also limitations. So you can there's been problems recently about kind of this computational spam where you have a bot that just, like, putting all hashtags and labels and, you know, that they can imagine to increase their reach. Right. And so that that muddies everything everything up for everyone else.
And it's the reason for that is really you want to increase the region and the intention on your ideas, which is why they were created. Right? Like, the the whole web 2.0 use hashtags to to organize and search information, but then you you just have this bad incentive where you don't know who is you don't know how to effectively select hashtags. So not only do you have to put the specific words relevant to your content, but you also want to, you know, kind of add more or put everything under the sun just so that way you get more increase, but then that becomes spam. Right. Well, if you have embeddings, it's a it's a different, it's a different, situation because now you're actually working with meaning. So anything that's not actually related, you know, through some distance, which you which you can actually calculate. If there if it's farther than that, then it really doesn't mean anything to you or at least in that situation.
[00:29:29] David Bennett:
Yeah. It that's what I keep wondering. Here's one of the things that I get on my own just within inside my own vault, and I'm not using a like a like an AI bot to crawl it or anything, is that when I go to add a new note, one of the first things that I start thinking about is how can I connect it with something something else, like what's called a you know, for the listeners, what's called a backlink? So I can like, I'm writing a note and, like, if it organically occurs to me, oh, that's like this one note that I wrote in symphony in c from from the book symphony in c, and then I'll do, like, some light searching, and I'll if I can find the note, great. And then in my new note, I will put a direct link to that old note so that forever and beyond, every time I come across this note, I'll see this backlink, and when I hit it, it takes me to this other note. Well, that's great if you know to link these notes.
What I'm starting to to have not I'm not having problems, but it's starting to creep up on me as something that I'm going, I'm writing notes. And if I don't write notes, like, every day, and that's supposed to be a habit, like this the Zettlekasten guy, I can't remember his name, he wrote 8 a day, and he did it for, like, 45 years. Well, I don't do that. Nicholas Newman. Yeah. And I because I don't do that, and I get kind of I kinda forget what it is that I've written. And then now I find myself reaching like I should or I ought. I have this, like, this moralistic or ethical need to put in a backlink. And I'm like, you know, what you're also probably doing is causing noise for yourself.
You're you're fighting to do this thing because you think you should do this thing, and you probably shouldn't do the thing if you don't immediately have this idea in your head. So that's just me as a single individual organically. And now I start thinking about the the notion of embedding and what can happen how does like, what's the analog of what I just described when we get into embeddings? And and that's where where I get back to keeping noise out of the system. Like like you were saying, just, like, just spam hashtags so that everybody and everything can, like, see this one note, and all it does is make me not wanna look at hashtags anymore.
So I guess and I I don't know how to answer that. I mean, is is there a way that that we can limit that kind of thing where we're just fighting to link everything to everything else? Is there a parameter that we can attach to an embedding saying, I don't know. Don't fight too hard or something like that? Well,
[00:32:27] Liminal:
I think of the labels and hashtags or backlinks. These are manual and really good for human navigation. And then the, the the embeddings are more context free. So if you have multiple knowledge bases that you wanna pull in from, but then you're writing something that you think is related, but you don't exactly know what the original source content is, you can take that embedding, and basically calculate the distances between that one note and all the other notes. And then, you know, you can say everything under this distance or the calculate the closest neighbors. Right? Like, you know, there's basically a a decreasing list of, you know, coordinates that, you know, the farther out it goes, the less relevant it is. So you don't actually need to worry about linking everything. You can actually do that later once you find, you know, the relationships between different notes.
[00:33:34] David Bennett:
Okay. So this this notion of of embedding and this context free is just kinda looking at what's most similar, and then you make the decision later. It's like, I'll show you what's closest, but you're gonna actually have to open these things up and make the determination as to whether or not this is relevant
[00:33:56] Liminal:
to you. For human for for the human navigation, I would say. But Yeah. Yeah. You you can you can use both. You you can, yeah, you can definitely use both. If you have a series of a 100 notes and you're writing something and you wanna figure out which one is the closest semantically, and you apply these this distance out. It's, you know, it's called the k nearest neighbors, and there's extensions to that of, like, figuring out which points are closest to the the point that you care about. Mhmm. And you can just say, it's just a a distance that, you know, that increases as as you go further away and just say, oh, just give me the the top seven.
Give me the top seven closest neighbors to this. And then you can see what's related. Or if you're doing co coding and you have, GitHub Copilot up there, what I think about is actually, if we are doing this collaborative knowledge base and we have NOS devs on that are kind of submitting their code and it's partitioned by however they they code, like, by function or by class, by file, whatever, you can essentially, write something that or put a comment that says, oh, I want I want to, pull in, notes from a specific relay. And I just don't know because I have no idea how to code. Well, the language model could take that and assign it to or or basically figure out semantically what other pieces of code are closest to that.
Like, because other devs, if they submitted their code with maybe even comments as well, it's there's there's a relationship. Like, you can actually calculate. Even though even though between programming programming languages or comments, even though they're all using different syntax, the meaning is similar. And that's what the these large language models are actually working with.
[00:36:02] David Bennett:
Yeah. That's what, you know, people use the term AI, and I I I just don't like using that. So I'm gonna stay with large large language models, and and there's, like, a whole camp of people that just think it's nonsense. And then there's, you know, other people that think it's going to, like, become so overarching in society that everybody loses their jobs. I know the truth is in the middle somewhere. You know? And and so I exist I I exist pretty much, I think, on the left hand side of the exact middle between complete nonsense and welcome our robot overlords. Right?
Because when I look at the like, what you're just saying, like, I'll go back to, like, you know, my time when I was doing a lot of 3 d modeling. When you're making a model in 3 d, your best friend is figuring out a way to make selections so that you can modify the, you know, what it is. You need to grab something that you need to modify. And if you're working with, like, you know, you know, thousands of polygons on a model, you need to be able to select which ones that you wanna make, which makes selection tools probably the most important part.
Like, the I I actually think selection tools are more important than the model generation tools, like build me a polygon. It takes Yeah. Second order to select that polygon layer. And I think that that's what LLMs in this context is able to do is become a selection tool. It doesn't mean that it knows what the hell it's doing, and it doesn't mean it's gonna be a robot overlord, but it also means that it's that's not nonsense. That is a highly that is a very, highly utilitarian tool. Right? But as I think about this, you know, I'm thinking about the notion of of embeddings, What I kinda wanna do is what and you may not be you may not think about it this way, but I'm just gonna go with it and and pray for I'm just gonna throw it out there and pray for the best.
What does a note look like? Like, let's say that I've written a note. I've got one ready to go in Obsidian. I actually have the ability to publish right out of Obsidian to Nostra, which is a a a plug in that I cannot believe is there. I thought that was awesome. Whoever whoever made that, thank you. So I've got, like, is that James Magoo? Is that is that it? I think so. Yeah. Okay. Actually, I could probably find out real quick who who did that one. I'm just gonna crank that up for just a sec. But, so I've got, like, all of my notes are done in markdown.
They're like, you know, like, Rome does markdown. Obsidian does markdown. Markdown language is, like, one of the easiest and most simple things that in the in the world. So I can write a note that has a heading and then maybe a couple of a place for a couple of hashtags, you know, for whatever reason. And then I'll get into the body of the note, and then I'll have the location or, like, the reference, which is which is a backlink to, like, I don't know, Merlin Sheldrake's mycology book that I read. Mhmm. And then there will be the possibility that I've backlinked to another note that is not hasn't nothing to do with Merlin Sheldrake whatsoever.
And then at the very bottom of of the way that I did my note structure is the Kindle location to where I can go directly to the file and find that highlight that the that the whole note came from. If I hit publish to Nostril, it's just nothing but words, which is which is fine. But your the notion that you say of add and embedding, What I just told you, is there a way that you can explain how an embedding might function with the note that I want to publish and make it functional out in the Nostraverse?
[00:40:22] Liminal:
Yeah. So let's just start off with what a note looks like. You have a unique identifier, you have the time published, the specific content, and then there's a section for tags where you can put whatever hashtags, labels you want. And, you can also have within within that tag section, you can also have these backlinks, as you can see with other notes on Noster, we can refer to. We can comment on, other notes, and we can also quote quote notes. And, that's also within, you may see that within the content because that's how it's displayed. But within this, this JSON object is a reference to an original note. That's how that's how we see this organization.
And, with embeddings, I'm it's more like, it's its own no, it's its own JSON structure, but it's referring to the original source material. So that way you can actually look at just the just the numbers. If you wanna look at if you wanna preview, if you wanna look at where it comes from, you can also take that unique identifier and pull it in. But it's it's kind of a thing that happens after, I because you can have multiple different models will have a different coordinate system essentially for, you know, this this meaning that it's computed.
And so you can have you know, I used OpenAI's ADA embedding, or you can use LAMA embedding. They're all they're all different they're all different, like, coordinates, but the end result is that, like, notes that are similar or text that is similar is closer, even though the space might be different. Okay. But if you if you wanna work with the same space or you wanna actually compare things, you have to use the same model.
[00:42:22] David Bennett:
Okay. So because I I think what I for for whatever reason, I keep thinking of the embedding as, like, something that you would include inside of a note that is a flag that says, please look at me and do something with me. But that's not what this is, is it?
[00:42:42] Liminal:
I would I mean, that's that's definitely a capability of, like, hey, please. Like, I want this to be searchable through embeddings. I think that's that that's definitely something that can be done. Oh, okay. Nothing is stopping anyone from connecting to any relay and finding notes that they care about and embedding them without without the original person's permission. Like, that's what, nasa.band does. They they add labels to to notes. I think nasa.social also does that for, you know, specific terminology. They'll add a little label there saying, hey. This has profanity or whatever and, you know, that because it's permissionless. Right. Yeah. I I think and also if you're if you're working with a knowledge based client like what we wanna develop, Yeah. Once you submit a note, we can also automatically use an embedding to, you know, produce another note that is linked to that one.
[00:43:35] David Bennett:
Right. So I guess at the end of this, there would be a client that you could choose to use. And if you do, maybe it's got a really slick graphic user interface that shows the the, like, the way the Nostraverse is for, like, I don't know, the last 4 hours, right, from from your followers or maybe you could even say from like, if you're using a web of trust model. Say, only the top 10 guys that that you know, in my web of trust model, I only wanna see their notes for the past 4 hours. And then all of a sudden, you see there could you see if they're connected at all. Right? And then you can start going through and visually picking those and say, why is that one so close to these other 3? What what's going on there?
And then all of a sudden, you might find that it's fired up an idea, and now you've gotta write an article about it. So it it it forces content creation. And I think that that's, you know, one of the ways that I look at Obsidian is is a forcing function on content creation, not just a repository, but something that actually says, when you've done enough, you'll be able to look at this thing and go, I can knock out 500 words in a day, and I would have never thought about this before because I read these three notes concurrently because they were so close together. And then to have that, like I said, now we've got this this thing where we can reach out from our Obsidian vaults and start connecting to other people's ideas.
And now we can have a client that looks at that that interaction. I I can't see what's inside your vault. I can only see what you've sent out. And since that is now out in the public domain, a client can actually look at those notes and say, here's the connections. And I can't I honestly can't think of anything potentially more powerful, but also at the same time, potentially, you know, without these, not safeguards, potentially as noisy without the issue of saying, I really need a client to show me what's close. And that that closeness the the farther away it is, the more noise in between is sort of like a way to for us to visually identify meaning.
And I how close are you guys to releasing something? I mean Well,
[00:46:11] Liminal:
let's see. Well, there there's ongoing work with the the actual interface for, project Alexandria. I'm not I I tried having a little demo client up, and that's what's being worked on, but I'm not natively a web developer. I just work mostly work data and try to visualize that. So and, you know, we're all doing this in our spare time with our, you know, our full time jobs. So Right. But, we're all really passionate about this stuff, and we're looking for, like, if we can find any any type of funding to, like, help get this off the ground, we were we're all we're all for that to so we can put more time into that.
[00:46:54] David Bennett:
You guys had submitted a grant to Open Sats, but it was rejected. Right?
[00:46:59] Liminal:
I Or something? Well, well, I think Stalin and I may have individually done that. I I initially submitted it, last last summer, But I like, I you know, that was also when it when these ideas were still forming. Right. And but also acknowledging, like, hey. I'm not a web developer, so I don't know exactly. I'd I'd like funding, not just for to for continued work, but, to also pay developers to make the client, the actual web part of this. And, I think Stella has also done the same thing. But yeah. But I think if we really if we find people that really believe in what we're trying to do, you know, we are taking, you know, any any type of support we can, whether that's donations or developer support.
[00:47:56] David Bennett:
We're actually really just we really wanna help,
[00:47:59] Liminal:
increase access to, like, education and learning.
[00:48:02] David Bennett:
Do you have a Linktree or anything with all the We have a geyser dot fund. Is that is that your main For for Git City Delia. Okay. I'll make sure to include that in the show notes because I I I know that address. I just for whatever reason, I thought I I thought I was following it, but I had to refollow it. I don't I don't know what's going on there, but I've got that address. I'll include it because I think that this project is really important. I don't think this is something that we can just say because either y'all do it or somebody else is gonna do it. I know that there's already like, they're doing it over I know they're doing it over at Twitter. They're do they've they've had it at Facebook.
They've had it at Instagram, but they're never going to release it. It's it's it is proprietary. It is the way it's one of the ways that they make money off of the people that use their network. And what you're doing is twisting that entire thing inside out and making it a a useful tool for knowledge development, content you know, potential content creation for everybody that is on the Nostra network, and that's what helps destroy Twitter and Instagram and TikTok because, honestly, I can't even stomach getting on Twitter anymore. I mean, I've been banned three times. And honest and now that I've been on Nostra for well over 2 years, the only time that I go back, I I finally I got a 4th account just so that I can post up, announcements for this, you know, like, for the Bitcoin and podcast.
That I when I go there to do that, I ins like, it takes 30 seconds of of scrolling to find something that pisses me off, and I'm like, I'm done. I'm just done. I can't do this anymore because they're using what you're describing in a terrible way. I know they are. And if we don't get it's for yeah. Well, if we don't get that for us, if if that's why it's important for for, you know, as much people as many people that are listening to this, if you can support this project, please do. Oh, by the way, do not either either DM me or email me your do you gotta get Alby address or a lightning wallet address?
Yeah. Okay. I need that so that I can split you in this in this thing 5050. Okay. Because I've been I've been meaning to do that, but most of the people that I've talked to don't they're not most of the people that I've talked to on this show so far haven't been so far down the rabbit hole that they have that they know what I'm talking about when it comes to zap split. So I just don't even I just don't even ask. But make sure that you give me that. Otherwise, I can't put it in. So that's that's the way that I can support you directly from this show. And the listeners that listen to the show will know that they're supporting you directly, but the guys are fun. I'll put that in in the show notes too.
How how can anybody get ahold of you outside of those 2? I mean, this is like what's, your, liminal on Noster. Noster. Yeah. And that's about it for you too? Yeah. I I yeah. Pretty much. Nostra only. I know. I can't like I said, I can't do I can't I just can't put this thing down. But, is there anybody else that that, people should be following out of this out of this project, or what's their Nostra?
[00:51:42] Liminal:
We got Stella and, Lasserin. That that's her that's her, NIPO 5, I think, or her Noster address. Okay. We got Michael j. I can actually pull up
[00:51:57] David Bennett:
pull up the group. Will you do me a favor? Will you send me the names of the people that because that way, I can go pull their Nosters, and I can include those in the show notes as well. Cool. Yeah. I've lost them that. Yeah. I I think people should be following, like I mean, I love following Stella. You know? It's like, I follow you, and I follow at least 2 others from the group, and it's always good content. We're we've been talking about project Alexandria, but let's for the last part of this, let's go up about 5,000, 50000 feet. With macroeconomics, Bitcoin, Lightning Network, ECash, Fedimint, Nostr, where where do you see humanity going? What direction are we going?
[00:52:54] Liminal:
Oh, I don't know about the direction humanity is going, but I just kind of see Noster having a real with and all these interoperable ecosystems being the kind of mycelial network that connects individuals and servers, developers. It's just you know, you have these pockets of organization, whether that is, you know, developer groups, communities, and that's where people are aggregating. And right now, as it as most of Noster is, is, we just kinda throw all the relays into a single client, and, you know, you get problems with, kind of intrusion into your conversation by, you know, by just people outside of of the core group. Maybe they don't they're lacking the context. Maybe they're actually malicious and and trying to just, you know, stir stir up something.
And I think that, especially with and with this kind of the financial layers that that work with it, we can actually separate, these communities so that way they can start flourishing. So actually putting kind of walls around communities so that way they can let people in, but also keep whoever they don't want out. And and I think the like, you you bring up the edge effect a lot. Right? I really think that, having this kind of separation between communities and individuals will help the one, help those communities flourish into their own thing, building in in silence without, you know, without external forces letting letting whatever they're building collapse through through noise.
Mhmm. But then also you have individuals that are kind of traversing between these communities. And whether that is just individually, they they look at these walled off areas or they're broadcasting, a specific note to other related communities, it's gonna really bring up a lot of collaboration and and mixing, which is, you know, it's really I feel like like just how nature has happened. Right? You you have everything mixed up together, and then at some point, there's differentiation, between organisms or between tissues. And because they're protected, they're able to have their own boundaries and specialize in their own areas, but interesting happen. And if by having this, like, traversing between communities, you also have very interesting things happen because it's this very specific and selected,
[00:55:42] David Bennett:
mixing that is allowed by the people that that connect those communities. So one of the things that I I wonder about quite a bit is the refer not reformation. The renaissance was basically brought on by the invention or the cobbling together, not really invention, the cobbling together of the printing press. First book, Gutenberg Bible. Now you didn't have to go to church to read the word of God. This it was not just that, but it was, like, the existence of the printing press and the existence of the stripping away of the power of the church by putting the word of God into the hands of anybody that could actually get one of these books, which while a peasant wasn't able to afford a Gutenberg Bible, a lot of people a lot of people were able to get these, you know, were able to get these.
And that kicked off the renaissance. And a lot of people said that the worldwide web, the, you know, www, the Internet is gonna do the same thing, and it didn't. So what what do you think what's the difference between like, because I think I got my first, like, when I was, like, first undergrad at Texas Tech was 91 or, yeah, 91, 92. And right around that time, they, gave you an email address, and that was, like, emails is, like, sort of just infiltrating into the colleges. And then about, 6 months after that, they said, if you want to be able to navigate the worldwide web, come down to IT and get your PPP account.
Alright. So I go do that, and I get Netscape. Right? And I was thinking I was hearing about, you know, how it was gonna be the second renaissance, and it just wasn't. But and I did and I never felt it. I I was like, I didn't call bullshit, but I just I never had a feeling like, yeah, this is gonna change everything. For me, it was like, why is this picture downs downloading so slow? Right? That that was what I was concerned around about around that time. But now we've got money interacting with ideas in an open framework that just seems to be ingesting if it even remotely looks like a network and it's similar, it's almost as if they're automatically seeking each other out. What changed?
Do you have you got any idea? I mean, do you you know what I'm talking about?
[00:58:18] Liminal:
Yeah. I I think part of it is, you know, kind of going off of the framework of companies even before the Internet, existed. You just have these individuals kind of building their own monolith and, you know, seeking funding to help contribute and build to build that monolith and whatever, we now we have web 2.0 where you have platforms that are trying to keep you in, and then whatever developments in that platform are built to actually have, keep you in there and keep you engaged because that brings in more money through advertisement. Right? But, you know, these same tools can be used for, more educational. Like, you know, rather than recommending the most, insaniary, content to help you, you know, glue your eyes to the to the screen and get a rise out of you. You can recommend other things that are maybe not if that's if that's the world you wanna be in and and, you know, be, you know, throwing fire at people, yeah, you can be there. Right?
But you can also get recommendations from just actual similar content that is just not necessarily the easiest to consume, but because it's related, you're gonna actually put the work into actual to to learn it. And this is what what Noster is developing. And I guess a lot of these decentralized technologies is really having this permissioned or unpermissioned interaction between communities and between the silos that that have been created. And, you know, silos are great. That's you know, having having protection and boundaries is how things have gotten to, develop into, like, whether that's a floor flourishing ecosystem or, you know, a a medieval cast like the castle with the walls, and then you have a whole community that's helping helping build and and protect the the community that that is being, you know, kind of cultivated and under the control of a single ruler. That that's fine, but I think now we we're really able to move beyond that and have more, more collaboration and more agency to individuals as opposed to just more the, like, a top down perspective where the individual at the top is deciding what what the individual at the bottom do.
And, that was pretty effective, but now we I think that this is giving us more agency.
[01:01:00] David Bennett:
Yeah. I I that's a great answer, because something has clearly changed. But one of the things that I'm looking forward to the most is is to see y'all's continued work on this project. And if there's any if there's any way if there's any announcements, if you guys wanna get the word out about something, please just hit me up. I'll I'll put y'all back on, in a heartbeat to get this out because I think watching what I'm seeing is one of the most amazing things I've ever seen happen, and I'm so happy to be alive during this particular time. I could have been alive, you know, anytime. But this particular time, while it's hard on a lot of us and it's still hard on me as well, just I can't imagine living in any other time to watch these kinds of thing happen. So I'm just praying for your continued success and your continued work.
If, guys that are listening, please help them out. Go over their Geyser Fund. It'll be in the show notes. And whatever it is that you're, dropping on me on this episode, half of it is gonna go to my friend, Liminal Liminal. I'm gonna stop the recording. You hang on for a second and just say just say bye to the crew.
[01:02:14] Liminal:
Alright. Well, thanks for listening. It was a great, great time to be on, and I really appreciate you having me on. Oh, absolutely, man.
[01:02:23] David Bennett:
Okay. So just so that you're aware, I am dropping Liminal has given me a functional lightning network address or wallet address, and I am dropping that into the split to help support the Project Alexandria. So, please, when you're boosting, understand that not only are you boosting me or streaming me satoshis, but half of everything that this episode makes in perpetuity will always go to Liminal's wallet. So spread this episode far and wide, boost generously because they don't have any other financial support, to get this project off the ground. And I think that this project very well may be one of the more important projects in not only the nostrils space but in in the space for humanity where we try to have new ideas, and come up with new concepts, and effect change on this planet. And this is one way that we might be able this might be a discovery tool for discovering other people and other people's ideas that we can directly interact with in a way that we've never really been able to interact with somebody else's thoughts or ideas or writings ever before in the history of of humanity.
And while that may be just a bit of hyperbole, I don't think it is. I think that we are have been trained, as good little girls and boys in public school, to never have any vital thoughts and to, God forbid, ever actually verbalize what you think might be a vital thought. Because people will think you're stupid or people will think you're gay or people will think you're I don't know. What am I trying to say here? A nerd. And that you're just embarrassing yourself when you when you say things that that really mean something. And this this is not just this project that really means something to me. It's watching a potential network of humanity build something together that they might not even know that they're building it together. And I'm not talking about building Project Alexandria. I'm talking about being part of it and build using it as a platform to build ideas with each other that we may or may not know we're actually building together.
It's like I said, this is very much different than anything that we've ever seen. And I want to see these people get supported very heavily so that they can come up with a very nice client that makes sense, has a user experience that makes sense when it comes to interacting with each other's ideas. So with all that said, again, boost generously. Stream generously. And if you can't afford to do that, then spread the episode far and wide so other people can look at these ideas and say, I want to be a part of that, or I have an idea for that, and I'll see you on the other side.
This has been Bitcoin, and and I'm your host, David Bennett. I hope you enjoyed today's episode and hope to see you again real soon. Have a great day.
Introduction and Overview
Interview with Liminal: Project Alexandria
The Vision of Project Alexandria
Noster Library and Knowledge Base
Understanding Relays and Data Servers
Modular Articles and Zettelkasten Method
Navigating Noise with Embeddings
Human Navigation and Embedding Context
Publishing Notes to Noster
Project Alexandria Development and Funding
Future of Humanity and Decentralized Networks
Closing Remarks and Call to Action