TFTC: A Bitcoin Podcast - #322: Building the reputation layer for social with Maciek

Episode Date: April 6, 2022

Join Marty as he sits down with Maciek of Hive one Follow Maciek on Twitter Follow Hive.one on Twitter Read more about Hive.one Shoutout to our sponsors: Unchained Capital Braiins HodlHodl Bitcoin 2022 - use the code TFTC for 10% off

what is up freaks this is Jigsaw from oh what the fuck was that it's all you want to play a game that's all I got it's actually Marty Ben

Miami maybe we'll switch up the ad reads here Bitcoin 2022 is in Miami April 6th to the 9th like you just heard, Matt and I will be doing a live show

on the 7th in the open source tent or on the open source stage. Join us. We're going to be physically and verbally assaulting Bake. It's going to be fun. We're not even going to talk Bitcoin. We're just going to mess with Bake. That's on the 7th.

The 6th is Industry Day. We're going to get a whale pass and bump elbows with everybody in the industry. You can do that there. Seventh and eighth are regular conference days. A bunch of people going to be talking, giving presentations, doing panels, making announcements, big announcement week. We're going to be live and in person for big announcements. Miami's going to be the place to be.

And day four is a music festival and a comedy fest. want to get cake thrown at you stevie aoki will be there you just stay close to the the front of the stage and i'll throw a cake in your face i don't know why you do that but some people get off on it use the code tftc to get 10 off after visiting b.tc slash conference get them while they're getting hot 35 000 people are going to be there a lot of people b.tc slash conference use the code tftc 10 off this group is also brought to you by good friends at unchained capital very well represented at the empower conference in houston parker lewis hardest working man in bitcoin was wheeling dealing telling people how to custody bitcoin unchained helps you do that

if you're a business an individual a family office you want to custody your bitcoin Unchained Capital has a two or three multi-sig setup they call it their vault you hold two keys, they hold one it's a collaborative custody you always have full control over your Bitcoin

if you have your two keys if you ever need Unchained to be the second of the two or three multi-sig quorum they are there for you they are here to help you get off zero and get holding your own keys if you're a rich person out there It's like, dang, I really like Bitcoin.

I really want to realize the full benefits that the protocol provides me, but I'm scared. I'm going to hold all my Bitcoin on an exchange. Shame on you. That's a single point of failure. Hit up Unchained. Tell them TFTC sent you and you'll get $50 off their white glove concierge service.

That's going to take you from zero to having a multi-sig vault set up with a thousand cuck bucks worth of sats in it. They're going to send you hardware wallets. they're going to have video conference calls with you they're going to get you comfortable it's not as scary as it seems it's not as hard as it seems

if I can hold my own keys you can hold your own keys Unchained wants you holding your own keys go to Unchained.com figure out how to do so it's important freaks this week this rip

this week's rip was also brought to you by good friends at Brains a really good time hanging out with the Brainsteam in Houston this week, Edward Evenson dude likes to party he's a bad influence he's a bad influence he's a bad man

you want bad man working on the products that you're I'm not saying bad like evil and nefarious I'm saying bad like damn that dude's badass you want these bad men working on your Bitcoin mining firmware. Organizing your mining pools.

Running your mining pools. Edward Evenson's a bad man. Works at Brains. Brains is a team behind Slush Pool. They're also a team behind Brains OS Plus Firmware. If you have an ASIC that is compatible with Brains OS Plus Firmware and you're not running it, you're leaving sats on the table. Don't leave sats on the table.

It's for losers. Are you a loser? I didn't think so. Go to Brains.com and see if your ASIC is compatible with it. And if it is, download it. You don't want to be a loser. Do you?

No, I didn't think so. Go to Brains.com. Download the firmware. Check out everything else they have. Last but not least, this is brought to you by a good friend at HodlHodl. HodlHodl is here

to bring you a peer-to-peer exchange, a peer-to-peer lending platform, no KYC, no AML. They're staying true to the cypherpunk ethos of Bitcoin. They have their lending platform, lend.hodlhodl.com, allows you to put Bitcoin

up in a two or three multi-sig quorum. You hold one key, your counterparty holds one key, Hodlhodl holds the third key. You have no control over your Bitcoin, but you do have visibility into that escrow account, so you know that your stats

aren't being re-hypothecated. You have certainty that if you pay that loan back, you are going to get your stats back at the end of the day. As long as how to holler your kind of party signs the second of the necessary two or three signatures. You put your Bitcoin up as collateral,

you get stable coin liquidity. You can go spend that as long as you're paying back the loan plus the interest. You're going to get your stats back at the end of the day. Lend.HodlHodl.com No KYC. No AML.

HodlHodl is doing incredible things right now. They've been doing incredible things. They continue to do incredible things. This lending platform again, it's a good way to get some liquidity without having to give up a bunch of information.

It's a great way to get some yield on your stable coins. You're a stable coin guy or girl. You enter the other side of that marketplace. You lend out your stable coins. You get what you lent out plus interest if they don't pay it back you get the sats it's better than you know the stable coins in my mind lend that huddle huddle.com enjoy freaks you've had a dynamic where money's become freer than free if you talk about a fed just gone nuts all all the central banks going nuts

so it's all acting like safe haven i believe that in a world where central bankers are tripping over themselves to devalue their currency bitcoin wins in the world of fiat currencies bitcoin is the victor i mean that's part of the bull case for bitcoin if you're not paying attention you probably should be probably should be probably should be logic um i'm sorry for berating you the first time you reached out to me about Hive One. That's okay. Why was I wrong? And what did I say?

I think that you were maybe not too excited about the fact that we put the rank of all the Bitcoin Twitter accounts in the ranked order, if I recall correctly. Yes. To me, back then I'm putting myself in past Marty's shoes. I guess I just, yeah, I don't know. I don't, I don't like the Bitcoin that's supposed to be this meritocratic open source, anybody can come in and bring good ideas. And I thought the ranking maybe could have been somewhat of like a, a, a dick

measuring contest more than like a quality contest. Yeah. And I can understand that sentiment. I mean, you're not the only person that, um, let's say this didn't sit well with because there is some something inherently um maybe i know it just wraps people the wrong way when you suddenly have a ranking right and um i get that but the the difference between what we're trying to do and what traditionally happened with lists that we're associated with right so you have ranking of influencers and all of that bullshit where you have a bunch of people make

subjective choices and say, oh, okay, these are influential people, right? While on the surface, it looks the same, what we're doing is fundamentally different. So we're not trying to be a subjective judge of who's influential, who's the leader. We're trying to algorithmically find existing dynamics in a given group and just provide you with a metric to understand what's happening in that group. So we're providing tooling. We're not we're not being a judge of it and I think that that's something that we have a lot of work to do in terms of conveying that better because that sentiment that that you had I've actually encountered quite a bit so I don't take it personally don't worry I understand that this

is this is going to happen so I'm just trying to take time to and I'm glad that we're starting with that because this is an opportunity to address this and explain why why we we use that format why there is a rank why there is a list because there is a good reason behind it which i'm happy to go into um so i don't know if that's that's the direction you would like to take it to yeah before we dive into the why and how i think of the particular mechanics of hive one but just like why hive one in general what drove you to to build this product originally sure um well it's uh it can be a long story it can be a short story so the the reason i started working on this is that i was just curious and before starting this company i had um i had built this well a complete

different company but um what was similar is that i was interested in how groups of people work together in a specific environment and how the type of this cooperation can affect like how effectively they work together so i 10 years ago i moved to berlin and i wanted to be to join the startup ecosystem to be part of that that community over there and i hit the wall because it i i was an outsider i was moving into a new city and i thought it would be much easier turnout was more difficult and at the time there were plenty of other people moving to berlin trying to join the startup ecosystem and it turns out it was actually a very common problem so initially i just created small event to uh try to make it easier for people to get into startup ecosystem and that kind of

like evolved into international franchise of events where i was designing a format of a decentralized event where we were replacing um any event like a venue with software and my goal was to understand how a startup ecosystem works as what kind of dynamics what kind of different actors are in this group that enables this amazing innovation that's somehow all of these companies and products and so on are coming out from these handful of cities that have these tarpico systems. What's different about the way

that people interact with each other? And that was really fascinating to me. I spent almost five years working on this company, even though I hated organizing events. But what was fascinating to me was trying to learn how these dynamics work

and how could I design a format that would enable people to interact even more effectively with each other. Or it would be even easier for outsiders to join these communities to become part of this, you know, like mechanism

that creates all of this innovation. And when I finally sold that company, I was frustrated with the aspect of organizing events. I was taking some time to think of what I want to do next. And at the time, I started becoming interested in Bitcoin and cryptocurrencies in general. And what was even more fascinating to me than the cryptocurrencies themselves was that I noticed the same thing I had seen in Berlin, as a community of people somehow coming together and collaborating in a different form than usually we see that, to create some, some products and companies. I started seeing this on Twitter. I started seeing the same dynamics happening there. And this was really mind blowing to me.

So initially I just wanted to like 10 years prior with Berlin, I just wanted to become insider myself in that community. And again, I kind of hit the wall. It was kind of difficult to, you know, to break in. And I started literally on a piece of paper. I started like thinking of like what kind of model I could design to find a sort of like key nodes in this network that if I connected to those people, that would be my shortcut to that community. And that one, one thing led to another. I started like designing simple models and we turned them into a script where

a friend of mine, it, you know, we, it spits, started spitting some lists of people, I showed these lists of people to a handful of friends, they said, wow, this is very useful. so I kept iterating kept iterating and at some point it started getting so accurate that I thought like okay there might be something to this idea and in order to test whether this is real you know whether okay maybe I'm seeing things or did we create something really useful we tried to run a test and back then we we ranked what we thought was crypto twitter and we published this as just a simple list on the crypto influencers io so we just bought a website we did this

completely anonymously. I tweeted about this and I went to sleep and I woke up the next day to like literally hundreds of people, you know, like tweeting about this or retweeting and saying, holy shit, this makes sense. How did they do this? And when this happens, I realized that, okay, we have something real, right? So that's how the company came to be. Like after this experience, I realized, okay, it's time to start another company. I raised some money. I built a team. We built a technology and, you know, like long story short, here we are. But what I learned in the process was that, um, we found a very effective way of understanding how communities form on Twitter and then understanding who are the leaders in those communities. So I did lots

of fun, like first principle thinking about like how influence works, for example. And I wanted to understand, I want to understand what is influence, how it works. And I realized that in order to understand it, really, I have to be able to quantify it. And the only definition that I come up with that would be quantifiable is that influence is a shared collective attention of a group. And the way I like to explain this somehow is that,

sometimes, is that every group has some form of pecking order. I know that Bitcoiners don't like the word hierarchy necessarily, but there is a difference between top-down hierarchy, like in a corporation, in a government, right? When it's enforced from the top and there is a hierarchy that emerges from the bottom, right? And every group has a hierarchy and that can be a hierarchy enforced from the top,

but it can be a hierarchy that just will emerge in a bar. You have a group of friends hanging out in the bar and there's always going to be someone in the group where if he or she says, hey guys, let's go to another bar. Everyone will automatically stand up and, you know, start moving. And there are some other people in that group.

if they say that hey let's go to another bar you will say yeah maybe later you know like they will generally get ignored right and in theory if you could put a chip in the in the head of each one of these people and quantify how much attention each person receives from the rest of the group over the course of the night i like i mean obviously i cannot prove it but my theory is that you would be able to very accurately predict what would be that reaction if each one of them stands up and says hey let's go to another bar if the rest of the group will follow or not so what our model does and what you're seeing on the website is an equivalent of quantifying that attention flow just on twitter because while we cannot put chips in humans in people's heads

while when you're interacting with others on twitter you're doing this through a pipe and there's a signal that flows through the pipe and we can observe that signal and that's what we're doing we're trying to quantify we're trying to estimate um how much attention you're receiving from a given community. And based on this, we can predict where do you stand in that pecking order. And so you said, when you came to the quote unquote crypto Twitter,

it reminded you a lot of Berlin and what you saw in that scene. What are the particular traits that you notice and that you're now trying to quantify via Hive1? Well, the fundamental similarity is that a startup ecosystem has a more efficient capital allocation in some ways than like traditional forms of capital allocation in traditional economy, just because there is a very deep interconnectivity between certain groups of people within that ecosystem, right?

So you have the community aspect is really a reputation system that is used by this ecosystem. Okay. That's a little bit of a word salad, but let's say you have Silicon Valley, right? The reason that Silicon Valley was so effective at allocating capital was that people knew each other. And because they knew each other, there was a reputation system, informal reputation system that decreased the risk of that capital allocation. So it's sort of, it's a, it's a equivalent mechanism to how a bank will request a bunch of documents from you when they want to give you a loan, right? They want to decrease that risk. So they're using different mechanisms for that.

They will also check your credit score and so on. That's a reputation system, right? That's a formalized reputation system, but there are also reputation systems that have not been formalized. And the most effective reputation system that is not formalized, that I'm aware of,

is a community and by far. The problem with communities is that they only work on, like they have limit in terms of like how large they can grow. Like Dunbar scale and stuff like that. Exactly, right? So communities are extremely effective

as long as they stay relatively small. And that's the reason these groups are also, so for example, Silicon Valley or the Berlin, you know, like startup community when I arrived, they have, well, or at least they had the reputation of being very open and welcoming, but really when you wanted to become a real insider and raise capital or, you know,

get access to some of the resources this community had, you actually had to work very hard and very long to build up your reputation, right? because the function of that is to decrease risk of capital allocation, right? And the same thing has been happening on Twitter. So in Bitcoin community or Bitcoin Twitter, there hasn't been so much of venture capital activity,

but you can see this with all this Ethereum, crypto, you know, DeFi and so on. They're doing quite a bit of that, right? There's lots of scams and so on. And they're trying to use these community mechanisms these days to try to decrease that. I feel like Bitcoiners are using this community reputation system for other resource allocation primarily, but this is happening as well, right? The point is that this community is really a means to decrease risk when it comes to allocating resources, right?

And the fact that this is happening these days primarily on social media is, well, in some ways exciting. because as I said, you cannot put a chip in someone's head, right? You cannot track that attention when it flows in a bar or in a conference room, but you can very precisely track these interactions when it happens through social media. And while communities are super effective

at being these informal reputation systems and in some ways very egalitarian because they're bottoms up, they're also limited by scale because they're limited by how much information can our brain process and store. But if we can move part of that computation, part of the information storage

onto a chip, well, then suddenly this dynamic changes quite dramatically, right? Because suddenly we, perhaps we can use this super effective community mechanism in places where so far we had to use state or corporations to do that reputation management for us, right?

Because we just didn't have any system that would be able to work on a scale of millions or billions of people. you can go above dumb bars number exactly yeah um don't worry about the books we got books falling in here but that's it's fascinating right and the egalitarian aspect of it is again like i mentioned in the beginning it's like a meritocracy people rise i mean that's i mean you mentioned before we started recording lop and like i just like remember back in the day uh when i was first getting on bitcoin twitter like it was like lop beauty on peter todd luke dasher john

carvalho bitcoin era log was on matt o'dell nick carter uh pierre and bitstein uh like back in like 2014 2015 those were like when i was just getting into bitcoin twitter i was like all right these guys are they know their and i've followed them ever since still hold them in very high regard and credit them with uh helping me come to an understanding of what bitcoin is yeah like i didn't find them in meat space i was in chicago at the time and in new york city and they were spread all throughout the world and i started creating a list on twitter my own list my bitcoin twitter list i i still have it i still use it to this day but it was just like me self-filtering and so that's just my my brain

creating filters and pattern recognition being like all right i'm gonna throw these people up and i've seen their body of work over many years and have the ability to trust their opinion on many things and not only trust it but also evaluate it critically understanding the perspective of their perspective and how that fits into my worldview. And so you're trying to mechanicalize this. Yeah, so what we're trying to do is to observe exactly what you did, but scale it up to every single person on that list.

And we're trying to observe dynamics, which tell us who do you pay attention to, right? So the fact that you pay attention to Jameson and to Pierre and other people, right? That's a signal that we're using and the more attention you yourself receive, then stronger the signal, right? And then we, it just goes back forever. So the first signal that we've been using is who follows you, obviously.

So we will look at who follows you and who follows the people that follow you and who follows the people that follow them and so on forever, right? So what we arrive at in the end is effectively what you just described, right? You create, you curate a list for yourself because you want for yourself, you want have a list of trusted voices and everyone else does that or most people do that too right and what we're trying to do is all that we're trying to do is to collect that signal and combine this into a single list and this turns to this turns out to be very useful in the end right but it's also it's it's a bottoms-up hierarchy because it's that whether you are high on that list or

lower than that list you have no ability to influence this the only way to influence this is by having other people make that choice for themselves that they want to listen to you, right? You have no way to coerce them to do that. And that's very different from how we traditionally think of hierarchies, right? Because in a corporation or in a government, you're in the top of the hierarchy

and you control that hierarchy if you're at the top of the hierarchy. In a bottoms-up hierarchy like here, the only way you can stay at the top is if people at the bottom see value in what you're putting out. Yeah. So you mentioned follow, calculating who's following who and all that. What other metrics go into this? Because I'm very interested to hear how you guys are calculating everything on the back end. because, again, this drives me back to,

I mean, obviously this is a podcast. I have a background in podcast advertising from my time at Barstool Sports. And there's one thing that's notorious about podcasting specifically, especially when you're trying to monetize it, is that podcast metrics are notoriously opaque and inaccurate in a lot of regards.

And one thing we really stressed at Barstool, and what I stress here now at TFTC, see was what you really have to do like to to understand the impact of an ad campaign on a podcast you can't look at the number of downloads what you have to look at is the follow up on social who's engaging with each episode maybe even the advertiser and that's what you should hone in on don't care about the number of downloads per episode um or or month-on-month growth rate like what is the social engagement 100 who's listening not not how many downloads there are right like that's the metric you would like as a podcast host yes we actually built a

prototype of that metric uh quite a long time ago um and it worked really well so the way we did this was that we took we calculated um uh also a score for how influential each tweet was so based on because we already have information about like each person on bitcoin twitter right so then we would look at every single tweet and who interacted with that tweet right like who replied to that tweet who retweeted a tweet and then we built um we scraped all the podcasts from you know all the major podcasting platforms and we exported urls and now we matched all these tweets um with specific podcast episodes right so regardless whether someone shared your podcast episode with url to spotify or to i don't like apple podcasts or soundcloud whatever else we would we would

merge all of this into a given episode and we calculated how influential each podcast episode was in Bitcoin Twitter or Ethereum Twitter or whatever other Twitter. And eventually we stopped working on this because we needed to focus on some other things and we didn't want to get distracted. But the experiment worked remarkably well. Remarked remarkably well.

So we're definitely going to go back to this. And I see this actually being, eventually this might be the way that you would price your ads, right? Because if as an advertiser today, the online, that's kind of like a tangent, but online advertising works in a very weird way today. So the whole gigantic market is done through two forms of targeting. So one is targeting for intent. You go to Google, you type in shoes, you get Nike, but like all social media

platforms and podcasts will use targeting for audiences, right? So, okay, I want to show this that to mail between 20, 25, link in New York, making over $70,000, right? Or whatever other demographics that you pick. And this just doesn't make sense in most cases, right? Like for your podcast, for example, like most of your advertisers, I would imagine, would want to say, okay, Marty, like the more, let's say we want to speak to Bitcoiners, right? The more Bitcoiners listen to your podcast, the more we're willing to pay for it. that seems to me a way, way more reliable way to price, um, you know, how much they should actually

pay for time on your podcast, right. Rather than, you know, what age or, you know, like income group or whatever other demographic demographic data today is being used for those things. Yeah. No, yeah. You can get very granular. So that gets to the point, like, from what I understand, the Hive 1 meritocratic ranking product that you've built, it means to a much larger end. Yeah. I mean, Hive 1 has always been, in our minds, more like a demonstration of what's possible with the technology we're working on. These days, the way I would explain what Hive 1 is, is it's a really good way to discover reputable accounts to follow. And maybe soon we're going to be adding way more communities because now we're we're scaling our index so

you'll also be able to discover interesting communities on twitter but really what we're working on is a new um a new way to index uh social graph so because we invented this technology that enables us to index social graph to index these relationships and status of individuals in a group. Well, we can, with this, we can index social graph the way that Google indexed the web, right? It's actually a very similar model in principle, but instead of looking at the links between websites, we're looking at attention transfers between individuals in the group, right? So what we're trying to do is to index the social graph, like Google indexed the web and social graphs are everywhere, not only on Twitter, but also obviously on all the other

social media platforms, but also in podcasts, right? Who appeared on each podcast, guests and a host as a social graph in academic papers, in venture capital deals, in lists of conference speakers, and a whole bunch of other things. Like there are hundreds, if not thousands of different social graphs. And in theory, you can bring all of them together into a single index so that you could build search engines and social feeds and, well, essentialist identity systems. Like it's really, it's an endless list of possibilities once we have that index that can, you know, anyone can query.

Yeah. And so how do you designing hive one design for like the edge cases and maybe when I would consider like low effort engagement farming, right? Like the breaking tweet, like how do you, cause people like ripping Bloomberg headlines, putting breaking in front of it. and knowing that that is not necessarily meritocratic, like putting ideas out there, but realizing how to hack people's dopamine receptors. Well, that's actually pretty straightforward because you can, so what we're interested in, and again, this is all just probabilistic estimation, right? So our algorithm is always improving.

It's never perfect, but we're trying to estimate what share of collective attention you hold in a given group and over time, right? So if someone goes viral with some tweet and then everyone forgets about this the next day, well, our algorithm is not going to pay too much attention to it, right? But if there is someone like Adam Beck who has been receiving attention from Bitcoiners, from other influential Bitcoiners for a very long period of time, well, then this means

that most likely he's very influential in that group, right? so actually these viral tweets that's fairly easy to account for and what even happens if someone does this repeatedly then if you start looking at we often

generate these data visualizations attention will come from certain groups of accounts so very often when someone publishes these sensational tweets those very influential accounts don't really pay too much attention to that and you see that at the edges of the cluster right so those accounts that are fairly new

they're the ones that get excited and that doesn't result in much of a difference in score if at all yeah it's fascinating it's uh again going back it's weird being ranked yeah i checked it before you came in here it's so weird like i think i'm number seven on the bitcoin list congratulations it's like do i deserve to be number seven that's like the the question. I don't, I don't know. Well, there's a thing that this is the most objective way, how you can get an answer to this as, as possible, right? Because we're not deciding this.

I'm not deciding this. I'm just telling you after crunching all of this data from tens of thousands of accounts on Twitter, that's who they pay attention to. Right. Um, so good for you. Seems that what you're putting out there, like, uh, carries value to lots of people. So weird to me. It's so weird.

Yeah. I can just hear like the, yeah. But it's, so how, again, like ranking all this, getting reputation, identity, how does it, we've talked about how it can change the scalability of groups above Dunbar's number, how it can help advertisers more granularly target and podcasters price their product. if they continued on the advertising role? What else is this? Like, does this help? So there was a lot of talk recently about Twitter and, you know, misinformation and all of that, right? And I'm squarely in the,

well, I'm on the side of the debate where I believe that Twitter should not decide what should be allowed or shouldn't be allowed, right? Like I believe that this is not the role of the platform. But at the same time, I do agree with the fact that there is a lot of like nonsense and bullshit that should be somehow filtered out from my feed, right? What's possible with our technology is that you could just decide, okay, well, I have a bunch of communities that I trust and a bunch of communities I don't trust. And I can have now a very simple set of filters that on given topics, I want to hear from this group and on other topics, I want to hear from that group, right?

So let's say COVID was one of these like hot topics, right? Well, maybe I want to choose that I want to only hear from a group of virologists, right? Or epidemiologists or whoever else, right? And as long as that group delivers quality to me, I'm going to continue paying attention to that group. And as soon as I become suspicious that maybe this is not the best quality information, maybe I will switch to a different cluster within that discipline, right? Because the same thing happens like you have on crypto Twitter, where you have all of these

different cryptocurrencies and tribes and so on, right? I know that Bitcoiners don't consider themselves part of it, right? But let's say this section of Twitter is very interconnected. If you look at academic Twitter, you will also see those clusters. And very often there is going to be a disagreement. There is going to be more than one cluster within a given discipline, right? virologist, you will have two groups and they have very different viewpoints on COVID, for example. So instead of having the platform decide, okay, what opinion should be allowed or not allowed,

right? You have a bunch of groups of experts. And now as an individual, you can choose, okay, I don't want to hear from random people on this topic. I want to hear from experts, but now I see, oh, okay, there are two groups of experts and they disagree on these topics, right? So now I I can understand it in a much more depth. I can still make up my own mind. I can make up my own decisions. I can filter out all the people who are talking out of their ass

on the topic they don't understand, but I can still make these decisions myself. Yeah, so you can have competing views. You can have like the Malones of the world, the Peter McCulloughs of the world, and then the Fauci's of the world and his acolytes. And you can see that they have different viewpoints, but they're somewhat subject experts

and you can decide for yourself. Yeah, and you can also see who pays attention to them, right? So which groups pay attention to them. And that's, I think this is particularly helpful because we are not able to be experts on everything, right? Like if someone is talking about particle physics or I don't know, like some biology or chemistry or, you know, COVID for that matter, right?

Like I'm not able to really understand every argument, right? Because I don't have this depth of knowledge. But so what we do is that we relegate some part of that trust to some indicators of like how credible that person is. And for the last several hundreds of years, we've been using primarily degrees for that, right? So this person got a debt degree from that university. And on that basis, we're going to trust them more than that other person.

And that's far from ideal. but we still need some some proxy right for those topics that we simply don't have the expertise so just jumping from the topic of of covet right maybe you're not able to evaluate to to tell between two experts arguing between themselves but you see that okay that group um that that group of experts uh is getting attention from lots of physicists and mathematicians and so on and that group of experts might be getting lots of attention from journalists and from um i don't like you know like different fields that are not really very stem related so your probabilistic estimation could be in that situation oh okay i probably want to i want to probably pay more

attention to that group that physicists and mathematicians and you know those people are um you know considering them let's say more credible and that's of course not perfect there is no perfect solution here but in my view that's at least an alternative we should seriously consider to having a centralized arbiter of truth deciding okay this is allowed this is not allowed yeah i guess one problem you have to think about too is like the galileo copernicus problem where in retrospect they were obviously very right but the the crowd the the popular opinion at the time was very much against them and how do you try to like well that happens every day in science actually you have um you know you have a bunch of old people that um are dominant

dominate given fields and it takes well they might not change their minds and um it's kind of like a old um old old truth or i think coon wrote about this that uh uh you know that the science changes because the old people die off and then you know we can have new ideas yes percolate right but it's not that no one listens to these new ideas it's just that the people who are really dominating those fields they they they they refuse to listen and they have majority of the largest cluster but there is usually a cluster a small cluster emerging that's slowly building up that um entertains those new ideas right and that's the reason it's so important that we have um space for those ideas that are not say considered um the right ones by the dominant

cluster we have space for these small clusters to develop and to entertain those ideas because they over time some of them will be nonsensical some of them will die off but sometimes this new new cluster that's split off it will turn out to be right and this idea will grow larger um but again with the technology we're building you actually would have a mechanism to to notice when there is a group splitting off and forming a new set of ideas right and you can actually start paying attention to that group and okay sometimes you will say well this sounds like nonsense to me but sometimes you will want to pay attention to this new set of ideas and right now we don't have a mechanism to notice that we on like there is a huge delay um and it's not only a problem in

public discussion but also it's a it's a serious problem in science like um in terms of like how quickly we can progress and how quickly ideas percolate in science you know my mind's going straight to like trading fund fund management there was a lot i mean when i worked in the the fund world like one of the big themes in like 2013 2014 was trying to have these um i forget what they call them but these thematic funds that oh quant funds quant funds but yeah Yeah, they were quant funds, but like they would try to gauge, gauge, what's the fucking word I'm looking for? They try to gauge sentiment.

Sentiment. There we go. So the problem with, and I spoke to some, several of quant funds in the past when we started working on this. And I was surprised that they're actually not doing too much with sentiment. And the problem with doing the sentiment analysis is that, let's say you were trying to gauge the sentiment for Bitcoin, right? And you did this on Twitter.

Now you're going to have a bunch of accounts that are expressing positive sentiment towards Bitcoin, right? And then you see, well, these are Bitcoiners, right? Of course, they're expressing positive sentiment towards Bitcoin. Then you're going to have a bunch of accounts that are expressing negative sentiment towards Bitcoin,

like Ripple, you know, Ethereum. Peter Schiff. Gold Bucks, right? A bunch of those clusters. And now, well, they're unlikely to ever express positive sentiment. The same way as Bitcoiners are unlikely to ever express negative sentiment, right?

This doesn't matter. But if you took all the tweets, this is what you would get as a baseline of that sentiment, right? Like you would get a lot of noise effectively. But if you could separate, you could understand, okay, this is the Bitcoin community,

this is the Goldback community, this is, you know, XRP, whatever else, right? Now suddenly you can organize these different communities and you can understand, okay, well, actually this fintech community or this community, these really matter in terms of a sentiment.

towards that asset um so what we're building i believe is actually going to play a large role in those models down the line where we've been already having some conversations with some hedge funds about potentially using that um because you get um you get understanding like this algorithmic understanding of very large groups of people something that again this is no one no one has figured out how to do that yet and with the technology we're building it becomes actually relatively easy yeah i mean if you're able to visualize a cluster breaking off into a smaller cluster and then i just imagine like the way quants think and they just see like growth over a certain period of time like obviously a tendency like all right let's allocate some money towards

an idea that or an investment that this idea would flow capital toward yeah it's all probabilistic estimations yeah um so what does this do for society ever yeah i mean so it's hive one are we creating a hive mind kind of so the reason that um we called it hive one is that i wanted I would hope that in the future you're going to have thousands of hives. And what I mean by this is that what we're really building is an algorithm that can index all of these different social graphs.

And then we want to make this index available in a permissionless manner. So the same way as, imagine that Google, instead of building their index, they index all the web, but they build a walled garden around it and they build their own search engine and they've built fantastic business, right?

But imagine that instead of building their search engine and building a walled garden, they just made this index available to anyone so that you can query this index you have to pay them a little bit but you can query this index and you can build your search engines or you can build your maps

or whatever else right we would have a completely different web full of innovation full of different products and now imagine what if not only they made it permissionless and open but they also decentralized control over that index right so you would like

that would become one of the infrastructure layers of the internet itself where suddenly everyone can build on top of that. They cannot be cut off from it and so on. And that's really what we're trying to do down the line. So Hive 1, as you see today, would be just one of the applications built on top of that index.

And others can build another Hive, another Hive, another Hive with slightly different functionality, but tapping into the same index that we do. Does it make sense? Yeah, no, it makes a lot of sense when I'm trying to think of at scale, what does your server cost look like?

Like how much data? Tens of billions of dollars per year. That's what we're estimating would cost to maintain that. So yeah, how big, if you don't mind me asking, like is the server burden right now for you guys?

Oh, right now it's not a lot. We're spending under 10,000 per month on the AWS, but it's expanding rapidly. We just started growing our index. So we spent the first three years on fundamental research really because we had to solve some problems

that have been out there a very long time and we managed to solve all of them. So we kept it small intentionally and now we got to the point where our model is fully generalizable so we can we can use it to map any community on Twitter and literally several well less than a month ago we figured out how to make it make it also run in unsupervised fashion which means that now we can deploy a model in a given section of the Twitter graph and it will find all the communities that are out there and and also we'll be able to name them automatically. So the way you see on, if you go to Hive 1 today,

you will see Bitcoin, Tesla, community Ethereum, right? We gave these names to those communities manually, but now our model can not only find those communities, but it can also figure out what should be the name of that community based on what words do people in that community use in their bios, right? And this was the last piece that we needed to be able to just run crawlers on Twitter,

just like Google does on the web. And now we expect that the costs will explode quite rapidly in terms of how much the infrastructure costs. Well, it's crazy to think, what were some of those unsolvable problems that you guys fixed? Oh, well, some of them are just like in graph analysis, so pure mathematics.

And the others were just like, okay, how do we put these things together, right? So a really difficult piece is still, for example, where does a community end, right? So this is an ongoing process, how we're solving that, because as I explained before, the way we're going to find this group of people and then we're going to rank them, it's like, who do you pay attention to and who everyone else pays attention to? And there is going to be slight differences in terms of who you would consider a Bitcoiner and who another Bitcoiner would consider a Bitcoiner and so on.

So it's never one-to-one, it's the technical term for this is that the community is non-discrete, right? So that it's not a zero to one, the boundaries here, but it's rather zero between one, right? So it's a scale. It's amorphous in a way. Yeah. So what we're now trying to do is to find better ways to finding these, to painting these boundaries, and this is still not perfect. So at the moment there is something called insider score on the website which is which is still in beta and it's it's far from perfect we're working on better ways of doing that but um being able to find communities and name them give them boundaries this is something no one has has ever done before in such a way that was accurate

right so that's i would say that's the primary problem probably yeah damn it's uh it's crazy to see how far you guys have come because what did you reach out to me what two years ago yeah i mean we spent more than three years just you know heads down um trying to prove that this is this is even possible because it seemed possible to me in theory but we didn't want to make too much noise before we actually were confident that okay we can we can really pull this off because there were many companies that raised lots of money before i mean you might have heard of cloud they raised i think 40 million dollars or something like that and they started indexing you know like the social media but they never got their model to be accurate you know i didn't want to build another

company like this i wasn't interested in in building a tool for marketers i wanted to build something so accurate that you as a bitcoiner you would find it useful right or someone in vegan community would find it useful on daily basis to check their community right because this is the ultimate test whether the model is accurate enough that it satisfies insiders in those groups right so how do you prove that it's accurate well the ultimate test is whether it's useful to you right um and the reason we're confident is that we just have at this point um around 10 of all the bitcoiners on twitter um have used hive1 right um so that that to me is quite quite a good indicator so how are we how are we using it oh just signing up on our website

right so that that's what we track we're trying to instead of looking at i mean we of course look at different metrics but instead of looking at the total number of users we're interested in what percentage of members of a given community what person what what percentage uses our um you know like uses that that that list because we see our number one user is a member of the community right because this is the edge case that you mentioned before right like if we can satisfy um we often talk about uh um we are like internally we we always think about wiz uh you probably know him right he's this hardcore bitcoiner who spends like he told me that oh yeah i spent five to six hours every day on bitcoin twitter right and every time i explain to my team who we're building for i'm

saying we're building for wiz right we're building for this guy who's who's been you know part of the community for 10 years he's still five to six hours every day using that like it has to be so accurate that it's useful even for him right um so that that's how we're that's how we're thinking about inaccuracy and the only way to do this is just we're there's a reason i'm here right like there's a reason i'm in austin i'm hanging out with bitcoiners there's a reason you know i'm here on this show because we're building for this community this is bitcoin bitcoiners were the first community that embraced what we're building so um and also i happen to be a bitcoiner so it makes it easier to build for myself but ultimately we're we're going to build processes to as we're

expanding as we're indexing more and more communities we want to be building relationships and also more automated ways to gauge how satisfied each community is with our ranking because ultimately we're not building for advertisers we're not building for anyone else we're building for those communities and only as long as we serve those communities with something that's useful that they find that it's making their community stronger we're going to succeed yeah that makes sense it's crazy no as uh somebody was inherently skeptic when it first came out of the box it is strikingly accurate um again even though i don't think i should be number seven everybody is above and around me i'm like yeah it makes sense

that you guys are up here um and uh yeah it's uh so when you guys open this puppy up like Are you obsoleting Hive 1, the company? No. So, um, I mean, uh, how the company is going to be structured, we don't know yet. I mean, so, so this is, this is secondary, right? There's a team that's working on Hive 1 and there is another, um, let's say website-like, um, product that we're thinking of that we, um, we haven't, um, like talked about it yet.

And, um, we're for now keeping on the ROPs that's also going to come out. And apart from that, we're going to open an API. And we're like, long-term, the API is going to be the main product. And down the line, we're going to figure out how to give up control, how to decentralize control over that API, right? Over that index and over who has access to it and what terms. But the company is going to still exist. It's just that once there is this decentralized network that's operating,

you know, like the index, there's going to be multiple different participants in that network and we'll be one of the participants. Nice. and so when you open up the API do you have any plans to integrate something like lightning

like yeah yeah are you allowed to talk about it I mean I'm allowed to talk about this it's just that I don't want to

make a fool out of myself like speaking like before we have really designed this properly but yes I mean so just high level

how would you think about it at a high level like the network has to make money right so if if there is an index it's

as we talked about it's going to be very expensive to maintain like tens of billions of dollars per year potentially in terms of like indexing computing and so on so um when you're accessing that that index as a developer you will have to pay for that access right and it seems natural to me to leverage lightning for that um or something that will come maybe even later right the beauty of the approach

that we took is that we're first solving all those hard problems and we're waiting for this infrastructure to develop right and in five years when or three four or five years whenever we're ready to finally start putting this together as a you know on decentralized rails this infrastructure is going to develop even more right so that we'll be able to take advantage of all this innovation that has happened in the meantime yeah it's crazy because that's something that's always inherently made sense to me is like api calls just pay like exactly five sets ten sets whatever it may be if you're making thousands like it adds up and you can have like two sides of that right like so you don't really have to build like a very um complicated decentralization you know network

whatever you just need to have like authentication that happens you know some kind of reputation system somehow some kind of pricing system and then you can have like thousands different providers of like api like apis that um serve you know hundreds of thousands of customers and you have this just small piece in the in between that um uh that in this the intermediary between them right is this what web 3.0 looks like i have no idea you know like um uh yeah i mean the the term for this would probably be smart contract but i don't even i don't even know if this is the right uh the right way to describe it well it seems like the right way to build this type of infrastructure.

Can I give you a set? You can give me some data and I can plug it into my product. Yeah, exactly, right? So there's a reason I'm saying that I'm trying not to talk too much about this, not because it's a secret or anything like this, just that we haven't really had time

to really sit down and think through how exactly this is going to work. And I don't want to say something silly that later on I'll have to walk back. Yeah, no, the authentication part of it. That's something we think about here