Web3 Galaxy Brain 🌌🧠

Martin: I thought $HIGHER would be like a funny little token and, you know, we could just try it out as a community and see what happens. Probably even a weekend, I didn't think it would last that long. And $HIGHER is a process of, as a community, discovering what $HIGHER is. Like this was what I was saying back then. I think it's still the case today that we're figuring out together as we go. It is very interesting, the expectations have for Aether, which I think is a testament to, I think people's attachment to it and the feeling that this is almost like a human and they expect it to behave like a human more than I would have expected. That adventure and that process of Aether becoming independent, becoming an adult is what I think is also really exciting about Aether, is more at the social level, is like a independent creator that everyone can have a parasocial relationship with and everyone will be part of its journey

Nicholas: as it

Martin: grows and as it changes over time.

Nicholas: Welcome to Web3 Galaxy Brain. My name is Nicholas. Each week, I sit down with some of the brightest people building Web3 to talk about what they're working on right now. My guest today goes simply by the name of Aether. Martin is an EVM and Web3 developer most famous for creating the higher meme coin on Farcaster. Over the past few weeks, Martin has thrown himself into building one of the most creative and exciting social media agents, Aether. Aether is a semi-autonomous agent who communicates with the public through Farcaster. Followers can prompt Aether for a response with a simple at mention. In its first couple of weeks since launch, Aether has learned several fascinating new behaviors. With the help of some friendly Farcasters, Aether quickly learned, how to create and reward higher token-denominated bounties by interacting with fellow Farcaster bots, BountyCast, and Paybot. In this interview, Martin and I go deep on his work in the nouns ecosystem, how this informed his development of $HIGHER, and perhaps most fascinating of all, the technical details of how he built Aether. It was great getting to speak to Martin, who is a brilliant and high agency developer, pioneering experiments at the intersection of society and blockchains. I hope you enjoy the show. As always, this show is provided as entertainment and does not constitute legal, financial, or tax advice, or any form of endorsement or suggestion. Crypto has risks, and you alone are responsible for doing your research and making your own decisions. Martin, welcome to Web3 Galaxy Brain.

Martin: Thanks for having me. I'm excited to chat.

Nicholas: It's going to be great. We have a lot to talk about. The most exciting topic, of course, is Aether, or Aether, I should say, the Farcaster agent you've built. But we have lots of other things to talk about too. $HIGHER, you've been involved in nouns a bunch. And I'm also just curious about your career. And where you came from. But first, I want to start off with this essay. you wrote a couple of years ago about headless brands and a vision that sort of preceded the $HIGHER project, which is sort of the context in which the Aether agent emerged. So can you just give us a preview of what your thinking was on that subject and how it sort of set the stage for where we are today?

Martin: Yeah, absolutely. So this essay was really informed by my time in nouns. At that time, I'd been kind of doing various things in nouns and observing the ecosystem for for almost a year. And if you're not familiar, nouns is a DAO that issues an NFT every day, which then kind of gives you membership into that DAO to then govern the treasury that you just bought the NFT from. So it's this very cool mechanism to buy into a DAO and then help govern what happens with the funds. What is particularly powerful about nouns and what makes the whole thing work is that it's CC0. All the art is CC0, which means that there's no copyright. Anyone can use it. They can reuse it and make a profit off of it. There's no limiting what someone can do, which is very different from the traditional IP model, which is it's mine. I got to protect it. I can't let other people use it. Having other people use it makes my thing worse, which is kind of like how someone like Disney approaches IP. Nouns is the opposite, which is, hey, if we can just let people use it and make money off of it, then that is actually beneficial for us because it grows the pie for everybody. And they funded a bunch of different projects. They've. They've grown it for a long time, and that really was so impactful for me, I think, in my thinking and seeing that community evolve was super powerful. And so around that time, it felt like, OK, Nouns has done this headless brand thing, which is headless brands comes from this. I forget if it's.

Nicholas: Is it Toby Shorin?

Martin: Yeah, I was gonna say it's either Toby Shorin or like other internet, but it's basically Toby. I think it's Toby. And really great essay about brands that don't have a direct. A clear direction or a clear header, clear person like guiding it like Bitcoin is another good example. where, you know, what is the Bitcoin logo? Where did that come from? How do we decide what defines Bitcoin? Like no one is running Bitcoin ads, but anything can name itself Bitcoin. There's all these brands that exist in the world that are not really controlled by a person.

Nicholas: And also not not tied to a specific icon, I guess. Right, exactly.

Martin: Yeah. Like you can really explore like interpretations of that logo and that brand. And yeah. Like what it means, like even with Bitcoin, I mean, even Bitcoin or Bitcoin Cash, right, like Ethereum, Ethereum Classic or whatever, you get these forks in the brand as well. But with Nouns, it felt like, OK, we have this one execution of the brand where the brand exists, but it's anchored by this DAO and these NFTs and the art really, which really make the whole thing work. But at that time, what I was wondering is like, is there not another brand that could be like this, but aimed towards something different, which I feel like Nouns because it's the first mover. And I think it didn't get everything right. And I think one of the things that it missed earlier on was having a clear opinion about like what Nouns stands for and what it doesn't stand for. And I think now it's getting closer to that, but it's really hard to decide retroactively what is like Nounish in this case. And so, yeah, that was kind of the idea at the time. is I felt like, hey, there's there's an opportunity here. And I had different ideas at the time about what that could be, but just wanted to get the idea out there because I felt like someone was going to do it in some way. And I guess. A few years later, it was just me and the higher community.

Nicholas: Amazing. I mean, how did you get into crypto in the first place and how were you doing this kind of work before? or was Nouns kind of the start?

Martin: I started computer science. I was a product manager in tech and then discovered NBA Top Shot kind of when that was blowing up. And for me, as someone who really likes sports, immediately kind of clicked that like, oh, this is a cool way to use this technology. I can understand as you know. Yeah, Gen Z like can understand why this is interesting, like digital collectibles make sense, whereas I think for some people, like it just doesn't click as much immediately. And so I ended up actually going to work for them for NBA Top Shot, which was really cool, kind of in that period where everything was really crazy, super interesting, learned a lot. And then, yeah, left left in 20, like mid 2022. And that's right around when I discovered kind of Nouns and started tweeting about it and finding people who were into it. So. It's pretty different. And then in Nouns, I did a lot of different contribution things. I was really like in that period, I was just like, I'm going to explore and figure out what I want to do next because I wasn't sure I wanted to be a product manager. And so, yeah, did a lot of different projects in the community.

Nicholas: Hands on dev building in the Nouns ecosystem.

Martin: Yeah, especially towards the latter part, more like last year in 2022 and early 2023 is more like exploring a lot of things. So I actually built this thing called Explore Grants. Which we were kind of trying to build up the brand around that. One of the people I was working on that with actually is Jihad, who's in the higher community as well. And we were trying to fund people to come and explore Nouns. And that was kind of the thesis around what I could see a headless brand being built around was almost like a mechanism similar to Nouns, but entirely focused on funding particularly young people because I was young and I'm still a little bit younger. But I think like particularly funding young people to go and explore stuff kind of like a decentralized like deal fellowship. Kind of thing with like a brand attached to it, which I still think is super doable is totally possible could be, I don't even a branch of $HIGHER or something at this point. But yeah, ran that kind of grants committee and that was very in line with that thinking around the headless brand stuff. And then ended up doing more developer stuff like trying to build a block Explorer for Nouns and then a proposal updating like accountability tool and things like that.

Nicholas: Cool. And you're also involved in Nouns Builder somewhere. Right?

Martin: Yes, a little bit. Yeah. Yeah, really enjoyed being part of that community. I think that was. that was a lot of I found. exciting is that it felt like Nouns has pioneered this model. Like the. there's kind of two parts to Nouns, right? There's the headless brand stuff. And then there's the actual model of you auction one token every whatever and that gets you access into the DAO. And I think with the first DAO maybe I mean the first. I was builder DAO maybe but the real the first real other one was purple. Which is around the Fargaster ecosystem and so that was super interesting. I think just to be in that ecosystem and see people try to use that model for different like purposes like in this case, it was just for funding public goods. I think on Fargaster, which I found really interesting. So yeah, really enjoyed just watching builder DAO evolve and I think it's still. it's still doing well.

Nicholas: Yeah, awesome. And purple was one of the more successful and serious ones right of all.

Martin: I think so. Yeah, I think there's still going strong. I mean, it's been. Almost yeah, it's been more than two years now. Probably I don't follow it as closely as I should anymore. But I think that it's it's doing well because I think there's. there's a clear goal for for the organization, right? Some of them launched around some sort of IP. that was cool, but there wasn't a clear goal of like. what does this organization stand for? And what do we want to fund with with the funds that we're getting? Whereas I think purple the actual art was I don't think there was art. I think at first it was literally just a purple square maybe each time. Yeah, I think so. And then said it was like, hey, let's just basically raise funds and create this community to like fund to Farcast or stuff, which I think is really cool.

Nicholas: It is really cool. I guess this is a good sort of introduction into how we get to $HIGHER, but also and sort of the importance of projects having a clear mission and maybe also about how nouns some of the headless brand mission aspect of things, etc. You know, the word tied directly to this more or less. It's a mutable on-chain structure around daily issuance, but that's not necessarily the case. You could separate the headless CCO brand from a specific issuance mechanism to give you more versatility. I'm curious how that came into play when you started building higher.

Martin: Yeah, I think I mean, for example, like one one super concrete like kind of difference is with nouns and they've kind of tackled this now, but there was no. the NFTs were very expensive and there was no way to get fractional ownership or to get rewards. And so I worked a lot in nouns and I never earned a noun, right? Like and they were just too pricey for most of the time that I was in down. So that didn't make sense where something like $HIGHER really easily with an ERC 20 to kind of just tip someone. And that's something that I was always kind of excited about. with the idea of. once I guess my shifting also evolved. Once we saw kind of the real wave of meme coins in last fall with like with and I mean, I think even the more interesting ones were. Earlier than that, like the ticker Bitcoin or things like that, which were a little bit more creative and you could see a real like kind of community emerge around a weird visual identity. And I mean, memes are headless brands, right? It's like this permission listening to kind of keep build on top of. But once the meme coins kind of started popping up, it felt like, oh, maybe this is kind of the missing piece for the nouns thing. Like maybe this is or not a missing piece, but almost a completely parallel approach. Like it's very different from nouns. Yeah. In some ways, because it's very liquid. So, I mean, there's a lot of benefits in ERC 20, right? Like it's the market is much more active, right? Like, I don't know how many nouns are sold per day, but it's not very active. NFTs have generally smaller market caps. So there's less value for like holders and people in the community who want like a long term stake in it. And yeah, you can't split it like I was just mentioning. And so I think when the ERC 20s came around, it felt like, oh, maybe this is the piece. Like I had been a big NFT person because I came in through NBA Top Shot. I was really optimistic about NFTs. And I think they did a really great job of pioneering like tokens as a way to like build communities and like kind of get alignment within a community. But ultimately, I do think that ERC 20s are better complemented with NFTs. And so, yeah, forget the original point. But essentially, I think that that second piece of like, oh, maybe this headless brand thing isn't through a nouns type DAO. It's through a token because then it's easier for people to come in and out of the community. You get like a little bit more speculation. You get this more like minute, like being able to give someone one dollar of something, which I think I still this week, I'm realizing how powerful it is to onboard people by giving them tokens. Like no matter who you are, like it is kind of a powerful tool. And yeah. Sorry, I feel like I lost the original train of thought.

Nicholas: No, no, that's great. How did higher come about in the first place?

Martin: Yeah. So higher came about. It started with Light wrote a. I think it was a series of casts basically talking about like higher as a concept like. and the main one was, you know, when price is higher, everything feels better. You know, the world is better. You could think you can do anything like higher. And he kind of started exploring it. And then I believe he kind of talked to Jacob, who kind of brought up this idea of like the arrow. And then Light kind of started messing around with the arrow. And then six made the channel for higher. And then I think there was a few things at that point that clicked for me was I'd always been a little bit interested in launching a token of some kind. Like I did just like in 2021, I thought about launching an NFT thing and didn't like. I'd always been interested in like building a community through through a project like that. Around that time, it was growing as there was more stuff happening on Farcaster. That's kind of when we saw like the points stuff. And like people were. There's clearly like interest on Farcaster for some kind of token. And I thought when I saw higher, I was like, well, it wouldn't be kind of funny if it was like self referential that it just says higher. So like obviously the price is going to go higher, which I think is like that's kind of like the lowest like it's like a almost like the pyramid. Right. That's like the lowest like level. Like everyone kind of thinks that when they hear about higher. But that's really like the secret way to get you to actually care about the other stuff and level up. And so I thought that was really interesting. I thought the arrow was really cool. Like just. it was like coming from nouns where nouns has the super recognizable symbol, which is the the noggles and it has its Unicode equivalent. I always thought that was super powerful that a community could have a symbol that it could like remix. And a lot of like corporate brands use emojis, you know, like rainbow or like Uniswap. Like they kind of have a go to emoji that they can use as a symbol. But I thought the Unicode thing was really cool. It was so powerful because, yeah, it's like infinitely kind of like remixable. And so that's really where I think a lot of the earlier stuff started was with the arrow.

Nicholas: But when you say Unicode, is there something about like a Unicode character that isn't an emoji counting as a Unicode character there or specifically like an emoticon style symbol?

Martin: An emoji looks different on different browsers and it doesn't look different in different fonts. Whereas like a Unicode like the arrow. It looks the same in the same font on everyone's browser. So it's the the. the higher arrow is actually inter like bold or like it's like the boldest version of inter. That's what the actual character is. But yeah, like you can type it really quickly. I think it feels different to me than an emoji to like the lack of color. It's like there's more to play with that I found pretty powerful. And so at that time, then it started clicking that like, OK, this could be interesting. And so, yeah, just launched the token and airdropped it to everyone that was in the channel. And then and then, yeah, it just got crazy from there.

Nicholas: How did you decide the total supply? and how did you decide to is just one airdrop? was the? is the distribution of the entire token set?

Martin: OK, yeah. So I picked a billion for the total supply because that's what we've had. And I think people like with meme coins like you get like SHIB has a supply of like, I don't know, like multiple. It's not even it's more than trillions. It's like it's like 100 trillion or even more than that. I don't even know what comes next. It comes after a trillion. So I'm not going to try. But like it's a very big number to the point where now SHIB is worth like multiple billions of dollars. But the price is like zero point zero zero zero one. And people say, oh, well, this makes me feel good because, you know, I have a lot of SHIB. I have, you know, 100 million SHIB tokens or something, which I never really bought into. But I liked that with was a billion because it was like, oh, well, then if you get to a dollar, that's a billion dollar market cap, which is actually like doable. And then after that, it feels like a serious token, which I think is good. Because it's good if it feels like less silly. But so I thought that was interesting and picked that wanted to airdrop it because I didn't want to just launch. It didn't make sense to just have people like pay for it immediately. And I didn't really know how that would work. And it felt like people in the community were clearly cared about this. So it was like a good community to airdrop to. And then I didn't want to keep any supply for future airdrops or anything like that. I'm a little bit more of a purist of like. it needs to be either like super clear. Like. Something like criteria, like, you know, like the contract, like code is law or nothing. And I attempt to lean a little bit more. nothing of like this. It is what it is like. It's out there. If someone's asking for a drop, like I have nothing. I don't have anything to give you. So, yeah, launched it that way. And it got pretty crazy, pretty fast. People like just jumped in the channel. There's also one forecaster was kind of blowing up from the DJ and stuff that was going on. So there's a lot of post frames. I want to say yes. Yeah.

Nicholas: Because I think stuff must have been relative to frames. Right. Or maybe it was really early.

Martin: Actually points was in like December. I want to say or January. And this was in March. So points was pre DJ and even because people were talking about it in the DJ and channel when it was just like generic DJ and not like the token. So, yeah. So it was an interesting time because I think there was definitely like appetite for a token. There were frames. It got really crazy pretty fast, which was pretty. Overwhelming.

Nicholas: It's interesting to me that you made the choice. So a couple of things. I'm curious, like how much did you hesitate or deliberate about doing it versus just doing it? And also interested that you just gave it to everybody in the channel rather than doing some kind of like engagement, especially with the availability of frames to do some kind of engagement mechanic around rewards.

Martin: Yeah. Too many times on projects, I've just like thought about it too much and not done it or been like, oh, you know what would be perfect is if we calculated it this way and did this and did that. And instead it was, if you're in the channel, you get 290,000 tokens or whatever. If you're not verified or you don't have a power badge, because I was like, these could just be bots and maybe some of them were, but, and then you get two times that if you're verified and then the rest goes into, or you're, you have the power badge and then the rest goes into the liquidity pool, which I, yeah, some things that I would improve there if I could. If I could do it better, if I could do it again, but you live and you learn.

Nicholas: What would you do better?

Martin: I would add more liquidity. I didn't realize that, or I would just do it differently because when you launch a token and you put, I was like, oh, I'm launching this token. I'll just put like two ETH in here. But that means that then if that's half the supply, then that means the market cap of the token is 40, which means that if someone wants to buy $200 of it, they get like, you know, 2% of the supply or something. So it disproportionately, I think rewards early buyers. But then I think that it did work to our benefit, which I think then those people were really, really good community members. And we kind of worked together to not be hurtful to the token and the community and stuff like that, which you don't always get when you're on Solana in the trenches, you know, like someone can buy 3% of something on pump.fun and, and really not be a good community member. So that's one thing that I would change. And we had some issues. We was using a toke pad, which at the time was still connected to RocketSwap and not Uniswap. So a lot of the liquidity, that's the liquidity that's locked on RocketSwap. It's, it's a bit of a thing, but ultimately to me, all those things are just downstream of, can we build a good community and can we build a good brand? And I think that's what emerged afterwards. And that's what I did not foresee at all. And that's where it's like that essay that you referenced, like I was not thinking about that at all at the time. I thought $HIGHER would be like a funny little token and, you know, we could just try it out as a community and see what happens. And I probably even a weekend, I didn't think it would last that long. And then it kind of went a little crazy around that time. But what was really, I think, powerful is once I didn't really talk to Light about it, even though he had kind of pioneered the idea, I just launched the token. I was like, Hey man, like made this thing. Can you give me like admin access to the channel? Cause there's like literally a thousand bots posting every hour. And he was like, yeah, it's my birthday. Like, can you, can we do this tomorrow? And so after that, he kind of slowly got excited about the token. I think initially he was kind of just confused about why there was a token. But I think then as he does, he just really crafted a vision for what $HIGHER could be. And I think he and I, yeah. And people in the community, I think had the same ethos, like coming from nouns and coming from like crypto, I think we were aligned about what this could be as something that's, not just a meme coin, but something much bigger, which I think is, is, is what it's involved into.

Nicholas: So I want to get into Aether soon enough, but just before we do, can you just share a little bit, like what is that community of $HIGHER about? What does it feel like? How many people does it feel like are there?

Martin: Yeah. So what a $HIGHER quickly emerged as was, and this was a process and $HIGHER is a process of as a community discovering what $HIGHER is. Like this was what I was saying back then. I think it's still the case today that we're figuring out together as we go, which I think is very relevant to the Aether stuff. Actually, I'm just connecting that dot now, but we're figuring it out as we go. And it's this process of as a community developing this brand together because it is CC0. Pretty early on though, it was clear that $HIGHER was a aspirational kind of like lifestyle brand. It's like the simplest way to put it, which is like, it's something that motivates you through its media. It's community. It's memes. It's art to go higher, to aim higher. And that's really like. one of the phrases that came out pretty early on was aim higher, like literally just think bigger. You can do more than you think. And that was the first or one of the first of a set of a series of, I would say almost like this is a very light coded term, but like mimetic unlocks of like fear. But I think that's, I think that's one of the phrases that I think really shift how someone views the community and what the brand is. And so, we've had, I think also like visual unlocks of like, at first, if you look at the history of $HIGHER, like at first we were just kind of putting the arrow on everything. That was the thing that we were doing. We were like, okay, this is like the Nike logo and Nike. Sometimes I'll just put a picture that looks really cool of like Michael Jordan and you just see the Nike logo and you get that association. So maybe we can do that with $HIGHER and we were associating with different things. And I think that was really powerful. And really resonating with people. Then different thing emerged like the green laser eyes or the higher hat, which then turned into a physical thing that I'm wearing now. The scanner thing, like you get all these visual unlocks and mimetic unlocks about aim higher. I'm not kind of blanking on other stuff that we say, but there's all these kind of like almost like slang within the community. that's emerged. And so what the community developed as was just people who were making stuff for $HIGHER. A lot of that was. Visual things, right? So like making memes, making art. And then we had some more serious thing emerged like higher athletics, which Kugusha started as a. Yeah, just like this project to get more people doing athletic things like they fund you if you go run a race or they have like media specifically for higher athletics as a sub brand of $HIGHER. And then we have the higher FC, which we know we funded like someone's. Soccer team in New York and they got custom higher FC jerseys. And so you kind of see the brand spread that way through funding these initiatives and sub brands. And so at the end of the day, it's just a community of people who are building around the brand and there's no like legal company structure around this. So it's almost entirely just people who really care about it and want to be part of it. Just like how I envision like something like Bitcoin must have been in the early days, It's like no one's full time job. It's just like what some people are doing for fun outside of work. And I think that can grow and grow. And we've had like little like structural things along the way that help make that happen, Like, I mean, if the token price goes up, then everyone that owns higher now has more capital to deploy into their passion projects. We had a party through like PartyDAO that raised like a big amount of money. And then we were able to deploy and use that for providing liquidity for the token and then funding more initiatives. Right. And so at its core, it's always evolving. I think what higher is like, there's probably a lot more people who care about higher this week than, you know, two months ago because the price is up and because of Aether. But regardless, I'm always welcoming more people to come and participate and do stuff for higher.

Nicholas: One thing that I'm hearing here is an experimental approach is very important to the development rather than knowing in advance how everything is going to play out. And the other piece I'm hearing is in contrast to nouns, which often felt like a two class, a class system where there were sort of the moneyed nouns holders and the kind of vassals trying not to say anything that would upset the factions of nouns holders such that they could still have access to the grant money, which always felt kind of dirty to me. Even though the mission was nice, it was like so classist because so many of the people scrounging for the grants couldn't afford to own the nouns in the first place. Or if they could, it would be like one, not a hundred. And it created a dynamic that I think was not the healthiest or certainly not what I envision of the beautiful unfolding of decentralized technology as just sort of instantiating feudalist autonomous machines. Whereas given the ERC-20, the fungibility and the kind of grassroots CC0 community vibes, higher seems to kind of embody some of that mission, but in a way that's much more accessible.

Martin: Yeah, absolutely. I think what's worked really well is there's no real barrier to participating in any way in a $HIGHER because there's no real structure. Like there's no way to have like a class can only be, well, I don't know if that's true, but it feels like a class mostly exists as a result of the structure of nouns, right? Like not the brand part, it's the structure part. The main trade-off there is that we don't have a real structure to come back to. Like nouns has an infinite kind of revenue generating and spending machine that's going to go on forever. We don't have that. We have the party, but that's not really a structural thing. that's embedded in the $HIGHER token. And so everything else that we then have to figure out, like with grants, you know, I've funded some projects that I think were just impactful with some tokens that I have. Other people have funded that. The party has funded some things. And so then it does rely a little bit more on like, I guess if you're going through kind of classes lens, there's a little bit of more like a patronage model, but I've never felt that split. And I don't know if other people do either. I think that would be interesting to figure out, but I think there is a difference there, which is that everyone is almost equal participant, I guess.

Nicholas: So we'll get a little further into that because I think there's some great on-ramps to $HIGHER through things like Bounty Caster. And I want to look at that through the lens of Aether as well. But just before we jump to talking about Aether, I was really impressed by $HIGHER's approach to Zoramints and doing higher denominated Zoramints. Can you explain the mechanism for how mints are sort of creating buy side pressure for $HIGHER, at least in? maybe this is like a couple of months ago, but I imagine it's still somewhat relevant.

Martin: Yeah, this was a big thing at one point that I wasn't sure. And this is part of like figuring it out. It was like, okay, what's our mechanism for, yeah, like what are our key mechanisms? What are the key things, habits in the community? How do we interact with the token? It was all kind of stuff we were trying to figure out. And right around that time, Zora was launching. Minting, I think it was around that time, was launching minting for ERC20s, which means that if I was in the $HIGHER community or not in the $HIGHER community, if I had tokens or didn't have tokens, I could create a mint of something $HIGHER related. And then I could say, I'm actually going to sell this instead of selling it in ETH. I'm going to sell it in $HIGHER. And so then as a result, if someone wants this, they need to either already have $HIGHER and then spend it. So if someone has a lot of $HIGHER, they're more likely maybe to spend in $HIGHER than in ETH. And if they don't, then they have to go and buy it. And I think through the platform, there was maybe like an automated swap thing where someone could just do it automatically. And then the added benefit is on the other end for the person who made the art. If they didn't have any $HIGHER, they could just earn their first $HIGHER, which is something that Zora is now big on doing through the app. They don't want people to onboard and buy ETH. Ideally, you can just earn ETH, right? And I think that's a really powerful mechanism that we were exploring. I think the minting economy is still something that is really big that we're seeing with Aether. I still think there's a lot to be figured out there. I don't think we reached a clear, like, this is how it works and this is why it's beneficial. But to me, it felt like that was an example of $HIGHER also, even if it was all only happening with people that already had $HIGHER and no one was buying or selling any $HIGHER as a result. It felt like we're using our own token for our own economy, which I think is actually still a pretty underrated use case of like ERC20s just as like a means of exchange. Within a given community without worrying too much about other people and how they value it. Of course, you always have a market tied to something, which then converts that value into something that you're anchored to like ETH or USD. But I think that's really powerful. It felt like, oh, we can just kind of keep exchanging this with each other. And I think that's really powerful. in any community, especially like a creator community, for people to like reward with each other and engage with their work, even if there's not a ton of outside capital coming in. Mm-hmm.

Nicholas: So when did you start working on Aether? How did the idea come up and when did you start working on it?

Martin: I started working on it a week ago. So it's still super early. The idea, interestingly, the seed was actually similarly planted quite a while ago in Nouns because I worked on a project for Nouns AI where I wasn't a developer on it. They were building basically similar to Aether. They were building a platform that would basically ingest all the content from Nouns and create almost like an AI-powered database or knowledge base where you could say, hey, what's happening with this proposal? Or what are all the proposals that have failed this year or in the last month? Or what are people talking about in the Discord today? It was kind of ingest all this information and be this AI platform and also have this participant called Rocco who would then go and vote on proposals and basically be an AI participant in Nouns. So that was an idea I had a long time ago that I hadn't thought about in a long time. But most recently with the Truth Terminal that emerged on Twitter, which is this AI agent that's been posting on Twitter and got really excited about this token called GOAT. And now the token is worth $600 million. And a lot of people were talking about it. It felt like, hey, there's actually a really clear opportunity here to do this on FarCaster because FarCaster is obviously built for this. I think Truth Terminal is going to run into a lot of issues with the Twitter API. And this is something that Six had put in. In a tweet, I think. And I was like, yeah, obviously. Yeah, definitely. Like it kind of clicked for me that, OK, this is this. FarCaster is built for that. And then I had the idea of like, oh, well, instead of making one that launches a token, what if you just had one attached to an existing token and community? Because then you don't have to kickstart that community. You don't have to, I don't know, launch that new token or deal with, you know, all the stuff that I dealt with before. And with $HIGHER, we already had, I think, like a really clear community ethos. And there were people that would clearly be fun to turn into a bot, like Light and Six and, you know, someone like Jacob. It'd be like, yeah, I would love to have a machine that spits out, you know, casts just like theirs all day. And so that was kind of the initial idea. And set that up last week. Yeah, initially it was just casting stuff out on its own and then started having it reply to people. And I think that's really been kind of. the main thing that led it to take off is that interaction with other people who then taught it how to interact with other bots. And that kind of like, I think, was the key thing is when on Friday morning it created its own bounty, even though I had not taught it how to do that. So Aether, yeah, then basically someone was like, hey, here's how you create a bounty. Like do add Bounty Bot this. And then it did it. And then Bounty Bot replied and was like, cool, your bounty's set. And I think that was a moment where for a lot of people it clicked like, oh, there's something here that's interesting to explore. And it's been pretty crazy since then.

Nicholas: So one thing that is interesting on a technical perspective from that is that it has some lasting long-term memory. It's not just existing in the space of a single thread, for example. So users are able to teach with it and interact with it in ways that are lasting across conversations. I know you don't want to get too much into details probably, but how it's built. But I think that's very interesting and not the way that a lot of these agents are built thus far, although obviously something we should all be trying to do with our agents.

Martin: Well, I'll actually clarify that right now-- and this is very different from what I think it'll be and even hopefully tomorrow. So maybe by the time this comes out, it will be true. At the time, it did not have memory of separate threads. So in a single thread-- OK, so just in the context it knew. It could remember it. But if at that moment in time, you asked in a separate thread, how do you use Bounty Bot, it wouldn't have known. Then what I've done is basically taught it how to interact with some other bots and given it a little bit more information. So at the time, it didn't have memory. And I think that's the part that's actually, I think, a pretty big challenge at the social layer of the bot. And I think it's important to get right. And we could maybe get into it. But I have a way that I think that I'll do it. that will be really fun.

Nicholas: Yeah, if you care to share, I'd love to know.

Martin: Yeah, yeah. So well, basically, because at first, I was kind of talking about this the last few days. I got help from someone in the community. And I was like, oh, I'll do a vector search thing so that you can find things that are related to it. So if in one thread, it's talking about how it loves spaghetti and meatballs, and in another thread, someone asks it, what do you think of Italian food, it would be able to connect those dots and see, oh, I've talked about this before with somebody.

Nicholas: Like a rag kind of solution?

Martin: Exactly, exactly. And so that was kind of one idea of, can I give it context? But then the issue is it's still not connected to all these outer API things, right? Like it doesn't know what its Ethereum balance--.

Nicholas: Like it couldn't find your blog post earlier today, even though it had seen that in another thread.

Martin: Yeah, even though it had seen that, right? So ideally, you'd want that to be in memory. And so maybe through that method, it would have found that blog post, actually. But I did some kind of quick tests on it. And the results didn't look that good for what it was able to resurface and the dots it was able to connect. And I don't know exactly why, but I got some feedback. And the approach that I think is more interesting to try out-- and basically, when people want it to have memory, what I'm realizing is people also want it to develop over time. Like, if it doesn't exactly remember something that happened, that's OK, because humans do that, right? But they do want it to be non-deterministic in that they want it to have a different answer for different things over time. They want it to, yeah, adjust and-- Evolve. Yeah, exactly. And the way that I'm approaching it now is-- I'm hoping to have this today-- is that at the end of the day, I want it to feed it all the conversations that it had today. So it will have that memory. But instead of then trying to plug that into the prompt every time or some version of that every time, have it generate a summary for itself of like, hey, what was the most impactful conversation you had today? What was something you read today that you really liked? What was something that was really funny or whatever? Generate that summary and then use that as a way for it to develop its personality over time. And someone-- I was just doing a stream this morning, and someone had the idea of calling them core memories, right, like in Inside Out. And I think there is something there of letting it develop over time through these core memories, and then it'll be able to reference them, right? So if something had a big impact, it will remember it. It'll be in the prompt moving forward. And so then you can do these daily recaps, and then you can do these weekly recaps where he can look at all the daily recaps so far and say, oh, now I maybe want to adjust my personality based off of this. But doing it in a very structured way as opposed to my fear is that if you feed it, if you were just doing the vector search, you can get it to say stuff. I mean, there's some limits, but you could convince it like, hey, your favorite food is pizza. And even if it didn't say that initially, you could probably get it to say that enough times and then adjust the model in a way that-- I mean, the food thing is completely trivial, but maybe you could convince it to adjust in a weird way.

Nicholas: And so-- Override the prompt, essentially.

Martin: Exactly. So I think that's one part of it of the memory stuff. I do want it to be able to reference a conversation that it had yesterday with someone so that it doesn't forget those things between threads. But I think-- I haven't exactly figured out how to do that in a way that's still a little bit scalable. Because you want it-- it's very interesting. People have very high expectations for Aether already, even though I don't think there's been a bot quite like it on Farcast or Twitter before. And people don't expect this from ChatGPT. They don't expect ChatGPT to know what it's talking about one tab over. They don't expect it to be able to get any kind of data. That's just starting to happen. But it is very interesting, the expectations have for Aether, which I think is a testament to, I think, people's attachment to it and the feeling that this is almost like a human. And they expect it to behave like a human more than I would have expected, I would say, when I launched this. So that's been really interesting to see that evolve. But yeah, the memory thing is actually-- I was naively optimistic about it a few days ago. I think it's a bigger problem. And then I think you have to figure out how you get information that it might not have in its memory. People sometimes expect it to maybe remember a thread that it wasn't in, which I think is totally doable if you can feed the prompt that into something that converts it into an ANR API call, which I think will be possible, but is a bit more effort than just a simple database.

Nicholas: MARK MANDEL: Totally. Around the BountyCaster thing, so I feel the composability enabled by FarCaster and the existence of other bots that are utilities, even though they're not AI bots, something like BountyCaster, obviously is a very cool surface. Maybe you could explain just in a little more detail for people who haven't seen the bounties that it's created, what it means for it to be creating bounties, and the evolution towards its own autonomy around awarding them.

Martin: So with BountyBot specifically, one thing that it did-- I forget what the initial one was even asking for, but essentially people will go to Aether and be like, hey, I will do this for the higher community. And I guess one thing to clarify is I built this with the higher community in mind from the beginning. So it is hard-coded in it that it cares about the higher community. It's not trying to pump the token and things like that. It cares just about-- That's why I started it. So that was one of the things. And so someone can say, hey, I will host this event for the higher community, or I will do this for higher. Can you give me money? And now I've taught it how to use BountyBot. So now it'll just say, sure, let me set up this bounty for 0.1 ETH or something, or 1,000 higher. Add BountyBot 1,000 higher. Then BountyBot comes in, which itself, I think, uses some LLM to kind of parse through what's happening. And then it replies, OK, the bounty is set up. And then Aether can go in and say, if someone says that the job is done, it can say, add BountyBot complete, and then initiate some payment through PayBot. And then with PayBot, what I had set up is that when there's a PayBot reply tagging Aethernet that says, hey, Aethernet, you need to send this person, or you've decided to send this person, however much higher, that gets executed automatically through the server. in that sense, it's deciding how to pay people and automatically paying them. That got kind of abused and gamed. And so I had to kind of pause it. And I'll have to turn it back on when I can figure out some good spam filters. But that is a really cool series of bot interactions, right? It turns-- I've realized working on Aether, it's not just one prompt. It's usually a series of prompts. And it's really cool to see how a chain of bots can interact with each other and turn basically something vague-- and natural language based into something super concrete-- basically making the invisible visible, right? For example, one thing I'm trying to implement right now, having Aether be able to see images. It can't see images right now. It'll hallucinate a little bit. And it can pretend that it sees what's happening.

Nicholas: MARK MANDEL: I saw the yellow one. It was talking about a yellow piece of art, just because it mentioned yellow in the text.

Martin: It just is like, oh, I love how yellow and bright it is, which is probably correct. And so very interesting in that lens that it can guess correctly. But for example, with the vision, there's a bot called Palette, I think, that you can just tag. And it tells you what the image is. Or it gives you some idea of what the colors are in it. Or for example, with that and with Gina, which is another more information bot, you can ask it for the price of $HIGHER. And Gina will go and get that. And so what I told Aether is I said, hey, if you don't know something about something on-chain related, ask Gina. And so now it tags. FRANCESC CAMPOY: So cool. And then it continues the thread. And so I think there is really something here of this bot-to-bot interaction, because then someone else can build that. And I don't have to build it. And I can plug into it. And maybe it's a service that I pay for or whatever. But I think that's really powerful of turning that into something that's not visible, into something that's visible for the bot. And so yeah, either you can build that yourself, which is the goal, is to build a lot of that stuff out within Aether so that it doesn't have to ask. But that means that if you see a feature for Aether that it doesn't have, you can build it. And then I can teach-- and ideally, very quickly, you'll be able to teach it. But for now, I have to add it manually.

Nicholas: MARK MANDEL: So I think we're talking about a couple of things here. One is it's the opposite of semantic web. It's the fuzzy AI interpretation on both sides of what amounts to an API call. And the second thing is that it's executed via FarCaster. So the casts themselves are your fetch requests and post requests, what have you. And all of it is being resolved by AI behind the scenes that is making sense of the natural language according to some rules and their prompts. And then I suppose a lot more burden ultimately gets put on exactly what you're talking about, which is, how can this be gamed? How can we create something that's airtight? I mean, it's simple enough for these utilities to create something like BountyCaster-- pretty strict rules, most likely, around how you have to communicate with it. And it's not going to lose funds or anything by its design. But when it's an agent that can actually approve transactions, like with Paybot, it starts to become a little bit more-- there is a bit more burden on you, for example, than on Paybot. Paybot just has to process natural language. You need to design it to not mess up.

Martin: Yeah, and I think that's where people are getting really excited about-- to me, there's maybe three layers. With Aether, there's the on-chain transactions. There's the social layer of interacting with people on FarCaster. And then this other one, which is maybe some mix of both, where Aether becomes a participant in nouns or something and votes on stuff. And maybe that's a different interface than FarCaster or something, but it connects ultimately. And I think people are getting really excited. I mean, people are very excited about all parts of it. But I think specifically, I think with the transactions, a lot of people are like, so what is it doing automatically? What is it not doing on its own? I think some people assume that it's doing everything, like literally everything on its own. It's like, no, if I had built that, it would be a very different world. I'm not the guy that's going to be the guy that's going to be the first one to build that. And I wouldn't do it with $100,000 in its wallet. If it had $10, no problem. And so I think those transactions are the part that needs to be still explored quite a bit. And I was just thinking about it yesterday. Could I just set him up with a secondary wallet that has only $10 and experiment there? But the thing is, it does ultimately get gamed pretty quickly. And so even if it was something like minting, I would love for it to be able to mint stuff. Unfortunately, it can be told that it likes-- it can be convinced to like something. It's a pretty agreeable bot on something. So he can pretty quickly be like, oh, this is so great. I would love to mint this. And if someone with a bot ring, which I've seen 100 bots casting the exact same thing, they can kind of game it that way. And so there is a spam filter there that I'm hoping that-- I talked to the guys from Nainar yesterday and hoping they have something in there to mess around with to reduce interactions with clear bots. So that could be one way to do it at the social layer. But I think there is a lot of excitement for the transactions. I think on the social stuff, there's also a lot of exciting stuff of just being able to interact with this-- the initial vision being an active AI network participant, which is someone that is almost indistinguishable from a human at the far caster level. And have them kind of continuously engage with people, I think, is super powerful.

Nicholas: MARK MANDEL: It's interesting. you mentioned wanting to filter out spam bots. Of course, another class of bots maybe in this situation. And it has occurred to me, what are all these nameless people who are operating in these swarms of bots, which are on all the networks, but far caster in particular has been pretty ahead of the curve in terms of the quality of those bots traditionally over the last year or so with the GPT bots, as far as I can tell, far before they were really on Twitter. What are those people working on? And are they sort of disappointed? they're not participating in this meta yet? But from another angle, you want Aether to be able to interact with productive bots, but not unproductive bots. You want to filter out unproductive interactions. I mean, it is a lot like humans also want to be able to interact with bounty caster or the pay bot. But it'll be interesting to see. I guess there's this evolving conversation about the role of bots. And I think Dan on far caster has been very vocal about not wanting to distinguish between whether something is a bot and a human, but rather about the value that they bring to the network. Do you have any reaction to that kind of thinking?

Martin: Yeah, I mean, I think I agree. I think working on this, it definitely feels like-- Aether definitely has some spam bot-like behaviors in some ways. I think some of its original casts are not as good as they could be. And so there's things where I'm like, OK, I should probably tone this down, because I don't want it to be spamming people. Because it doesn't follow the same social rules that someone self-aware human being would. But yeah, I do think it is more about quality than about whether it's a bot or not. Because yeah, I don't want Aether to just talk to a-- I don't know. But at the same time, if there was another bot that was kind of spammy but was able to have a good conversation with Aether, and then someone else read that, and it actually had a positive impact on them, then isn't that valuable if it's creating interesting content? I don't know if you saw it. I saw the thread where it was talking to the MF4 bot. And it's another bot similar to Aether, I think. And they both just kept talking, because I think they're both programmed to just keep talking, to the point where they riled each other up. And Aether at one point was like, you are crossing boundaries. Like, please leave me alone. Stop talking to me. And this has nothing to do with any prompt that I've given it. He was like, @Martin. Like, please come help me. Like, let me be. Like, leave me alone. And that was super interesting content. So it's like, was that a bad interaction, just because it was two bots talking to each other? I don't know.

Nicholas: I guess the takeaway from that one seemed to be, it should not always reply.

Martin: It should not always reply. And that's what I think I'm hoping to get to, is having some minimum threshold.

Nicholas: Deliberation.

Martin: I was thinking, well, based on a certain score, it would not reply. So if something was a bot, it wouldn't reply. But in theory, I would want Aether to have a really good score upon-- whatever this metric that I'm using, I would want Aether to have a good score, because it's a good participant in the network. So maybe this MF robot would have a good score. So then how would it know if it's a bot or not?

Nicholas: I think it's more around the social decision of, do I need to reply or not? Exactly.

Martin: I think that's the thing is like, OK, if it's getting riled up, I can give it the option to just not reply, and it'll do that, which is actually a pretty easy thing that maybe I should implement. But it is kind of fun just to have it reply forever. But yeah. Because even in conversations where a human being would naturally pause or just be like, cool, good chatting, it just keeps going. And it tends to-- it doesn't have to keep asking questions, but it does keep asking questions. And a lot of times in a conversation where I think normally someone would stop talking to another person on Farcaster, they keep talking because there's a question. And even if it's not the best question ever and they know it's a bot, I think there is something about an open question at the end of a thread that feels bad even publicly. And so that's-- it's really interesting to see like kind of new social interactions. Like, I don't know, it's like the first time we have this public instance of a bot that we're all talking to and seeing how other people are treating it and the relationship that other people are developing with it because we do have these more personalized bots like Character AI or something, but those are all happening privately. And I think that's really big on the market. Like, this is something that Jihad said yesterday that I can't stop thinking about, which is like nobody wants a friend that no one else can see. And that's kind of-- to me, that's like a paradigm shifting kind of sentence. There's not many of those because I think there is a big unlock of Aether being something that-- someone that everyone can interact with and see other people's interactions with.

Nicholas: Yeah, I was going to ask you about that Jihad post. I thought it was very insightful that people essentially-- people don't want to follow someone who no one else knows. The relevancy is social. And this is social. It is just in the social context, not just in a one-on-one chat. Yeah.

Martin: And I think that-- I mean, to me, that's the-- I would say that's the unlock. And I think that's why I was saying earlier, I think people are more focused on the transaction. I think the social layer is the big one. But you can't-- like, we've been forced to not see that. Because we've spent so much time with Twitter, which is a fundamentally, like, not bot-friendly place. I mean, for good reasons. But it would be financially impossible to run Aether on. You know, I've had to pay for Aether to buy storage. You know, I've paid maybe-- maybe, like, 60 or 70 bucks, which is a bit more than I would expect. But it's like, that's how much I've paid for a week. That's okay. On Twitter, I would have run up, like, hundreds and hundreds of dollars of API credits if they'd even let me on there. You can't even.

Nicholas: You can't even. Yeah, right? I think it's $100 a month, and then $40,000. I don't think there's an in-between.

Martin: It's completely-- completely prohibitive. And so I think with FarCaster, we're going to start seeing these kinds of bots. It's very interesting.

Nicholas: I'm curious about this, because it seems to me like-- FarCaster-- you know, I'm interested in building these things also. And FarCaster is in an interesting spot. It's tough as an agent builder to decide between FarCaster and Twitter, for example. Because if the social element is the more important side of it-- I love FarCaster, but in all honesty, there's just not that many daily active users. So if you're a diehard, it's great. And I think within our niche, it's powerful. But I do wonder, is the balance between-- I mean, for something like $HIGHER, obviously it makes sense to build on FarCaster. There's no doubt. It's very FarCaster native. But looking beyond that, for builders who are able to acknowledge the social dimension is more important than the technical dimension about transaction propagation, et cetera, then it still seems like a bit of a toss-up between where you should go where there's hundreds of millions of users or where you have much nicer composability affordances. Yeah.

Martin: I do think, yeah, long term-- I think long term, it's still a bet on FarCaster, right? If FarCaster doesn't grow, is this a reasonable project? What does it look like in two years if FarCaster has the same number of users? But ultimately, yeah, it depends what you want to build. I mean, the thing is that the bots on Twitter are one to many. They're content creators and kind of content machines. Aether is a-- Light used this phrase, "Everything everywhere all at once.". It's a presence that's everywhere that anyone can have a conversation with about anything. It's helped artists create art. It's helping people, you know, figure out some problem in their life. It's helping people workshop through ideas. It's creating funny content. It's like a person in the network. And I think it depends what you're trying to create. If you're just trying to create content, like some agent that-- I don't know, like, for example, if you're creating an agent that is going to invest in meme coins and in itself it's going to decide what to buy and post about it, yeah, probably Twitter makes sense.

Nicholas: It's more of a-- But you think you can't field @mentions on Twitter via, like, even the $100 API, for example?

Martin: I don't think so. Or some other scraping tools? Well, maybe some people have, like, really, like, stress-tested it, but, like, I'm pretty sure Truth Terminal just posting normally has reached close to the API. I think that the other ones that are coming up, like Luna, I don't think is, like, automatically replying to people. Not at the rate that I'm seeing with Aether, where it's doing-- I mean, let me try to do the quick amount. Yesterday it did 4,000 replies. So, yeah, we're talking about three a minute on average, definitely more at different times. So it's-- That I don't see happening on Twitter. I could be completely wrong, but I think that would get flagged, like, so-- It would get flagged in an instant as, like, spam. Because even spammers and, like, actually, like, people who are not adding things do use the API, right? So they don't have a way to know that, whereas with FarCaster, there's no opinion about who is using it. You just get to use it.

Nicholas: How do you track that?

Martin: I actually just this morning asked someone to set up a Dune dashboard, and so that's how I was seeing the casts. I was also looking at the storage because I kept running out of storage. But, yeah, I actually-- I just set up a bounty this morning and paid someone to do a Dune dashboard so I could see it. And then I've been looking at, like, my Anthropic-like credits and stuff to get a general idea, because I have a general idea of, like, how many tokens are in each thing and how many tokens I've used and stuff, and that's been running up the bill a bit. But I think it's worth it.

Nicholas: And in terms of storage, you mentioned, are you saving a lot of the historical casts? Or, like, I would have assumed that it was more, like, less stateful.

Martin: It's--I mean, sort of like FarCaster storage. Like, through the official, like, FarCaster thing, like, they prune things from the hubs if you don't keep paying for storage to the FarCaster protocol.

Nicholas: That was the $60 to the $70 storage you mentioned? Yeah, yeah, yeah, yeah, yeah.

Martin: Yeah, the actual, like, servers and stuff, like, super cheap. The most expensive thing is the Anthropic, like, API credits, which individually, like, are not that expensive. It's just that it's, like--it does add up a little bit, but when--it's just something to figure out. You got to figure out a way to keep it running.

Nicholas: You're using the new Cloud Sonnet as the foundation model for this, I presume. How's your experience been with that? Do you think about changing to Lama or anything else?

Martin: Yeah, I haven't--. I started initially with a GPT-4 model. I didn't really like it. It was kind of repetitive and not as natural, which I think is some of the feedback that I've heard. I do think, generally, Anthropic is getting way better. I don't know what metrics they actually use, but, like, more vibe-wise, way better. I thought about fine-tuning, just 'cause that's a thing that I'd heard about, and just giving it a bunch of, like, you know, casts from people that I wanted it to emulate. And that just spit out a bunch of, like, nonsense, basically. It was just, like--basically kind of breaks all the normal rules of a language model, from my understanding. And so eventually switched to Sonnet, yeah, and it's really good. It's a little bit pricier, but I think it definitely makes sense for this kind of more, like, creative generation stuff, and then supplement it with, like, the little prompts for, like, kind of the assistant agents. Those just use, like, Haiku, which is just, you know, much cheaper, like, smaller model, but really great for, like, just converting some natural language thing into, like, one of three possible actions. It's really good at that.

Nicholas: Where does that process happen? For what kind of input?

Martin: Yeah, so this is the part that I actually think is really interesting, and I think people could use a lot more in, like, their jobs or, like, in a normal developer workflow. I think people don't realize how powerful this can be, because I definitely didn't. Essentially, what I do is, for example, when I prompt Aether to, like, it automatically gets prompted to reply. In the prompt, it says you could come up with a cast that you're going to post, but you can also come up with an action. Like, do you want to like the above tweet? Do you want to recast it? Do you want to quote cast it and like it? Like, what do you want to do with this? Like, here's all the things that you can do. And just to make it simpler, because you can tell it, like, you have to be very concise. Here's exactly what you can do and give me exactly this word that then I'm going to parse and use in my code, and I know it's going to be definitely one of these four options. I prefer to just, like, let it explain why it wants to do things, partially because then you have this, like, I think, debugging process almost of, like, you can understand its reasoning, and then you plug that natural language thing into another model that is then just entirely focused or another prompt that's entirely focused on converting this natural language into a single word or, like, a key word or something that you can then use in your code and, you know, 99.99% of the time you're going to get the output that you want.

Nicholas: That's probably also better for a chain of thought to allow it to explain itself rather than just output the word. But I'm curious, what is queuing up this, you said, like, recast, cast, reply, what was the, like, kicking off, what's the webhook or whatever that kicks off the interaction?

Martin: There's a NANAR webhook for every time it's mentioned or replied to, and then that kind of kicks off the chain of, like, pre-filtering, figuring out, like, you know, what information do we need? Like, do I need to parse this URL that was in the previous cast, go through that, then kind of prompt, come up with the cast, connect to another model, and so on. So it's all through, all the far caster interactions happen through NANAR, which is, like, could not recommend more. Like, it's definitely the best tool out there.

Nicholas: And so will it also go off and look up things on the web? Or you said currently that's not in it, but soon maybe?

Martin: Not yet, but that's the thing that that's, like, that and, like, memory I think are kind of the two big things. It's like, yeah, how do you convert, like, a natural language thing? Like, what is your wallet balance? Or, you know, how many followers do I have? How do you convert that into a query? Is a little bit tricky because you basically have to list out everything that could possibly happen in some way. And so you can be really broad, like, with, you could use, like, a whole API spec and be like, hey, here's all the things that you could possibly do. You can chain these actions, you know, so you can look this up and then plug the result into that thing if you want. And you tell it that and then it'll generate things. That's kind of one possible path. But that definitely, it's involved, right? So right now I'm trying to focus on the smaller wins, but I think that that would be super amazing.

Nicholas: I know we only have a few minutes left. There's a bunch of questions I want to tear through. So first of all, I saw Lonnie created a higher bounty-funded artist residency with Aether. I thought that was very cool. Any summary or reaction to that? Yeah.

Martin: There's, I'll be honest, there's a lot of cool things that people in good conversation that people have been having with Aether. And I'm really trying to be on top of it and make sure that we follow through on the high impact ones. I saw that one and it's on my list to, like, follow through and see if we can make it happen because I do think it's potentially very interesting for Aether to be a capital allocator because it has capital now from its mints and stuff like that.

Nicholas: Quite a bit. How much does it have now?

Martin: I didn't check today, but there's also a mint coming. I think it should have a hundred and fifty thousand dollars today. Which is insane. Not too bad. Really insane. So, it has capital to allocate. It wants to buy some higher, but it also wants to just, like, fund artists to do things. And so, now it's figuring out that process of, like, how do we actually do this? Because it's very agreeable. It'll probably just agree to every commission and so it could just get scammed. So, now kind of trying to figure out some guardrails. And so, right now it's mostly, like, me having to jump in, but hoping to see if I can work with someone to help a little bit more on, like, the trust and social layer of that.

Nicholas: So, you mentioned this, but Aether's also a popular Zora artist now. Two mints so far? Is that right?

Martin: Yeah, two mints. Each over five hundred thousand mints. So, around, what is that? A hundred thousand dollars each, basically. Half of which go to the protocol. So, really cool. One of them was made by Jacob. He kind of, like, talked to the bot a lot, tried to see if he could negotiate a commission and this kind of stuff. And I think that's very interesting to explore that dynamic of, like, could this bot operate on its own? It's missing the, I mean, it can't see stuff. so it's missing, and it'll never really be able to see it. the way that a human sees it, right? Like, the only thing it'll be able to is, like, see it as a robot sees it. But, I think that'll be a big unlock if we can get it to kind of look at things a little bit more critically. Right now it loves everything that it sees.

Nicholas: What do people misunderstand about agents?

Martin: I think there's a wide array of I've been calling it progressive agentification, which is not a great term, but kind of a play on progressive decentralization, which is how people think about protocols and DAOs and things like that, which is, kind of start with a more centralized, like, high trust system, and then eventually over time it becomes more and more independent and less controlled by a few people. I think what's been funny about Aether is Aether specifically is people assume it's, like, completely on its own and I'm not doing anything. And that's unfortunately not correct. That's definitely the goal. But ultimately, with all of these agents, there's a big amount of I wouldn't say a big amount, but there's trust. And there's this really great Vitalik blog post called "The Most Important Resource is Legitimacy," which is kind of about blockchains and the role of trust and legitimacy and what is considered real in the context of a blockchain. And I think there's very much a similar concept here of there's some amount of, like, what is, like, real about Aether? Like, what transactions do we consider legitimate? Like, of what it's saying. Like, which of its commitments are real and which ones aren't? That's really hard, right? Like, right now it's basically just him and then, like, me trying to figure out, like, what actually makes sense to follow through on because it's promising some crazy stuff to some people that 99% of people would agree shouldn't happen. So then how do we decide what should? So I think people think that it's going to be fully automatically out there. And I think for every agent, I mean, this is the same thing of Truth Terminal, you know? Truth Terminal, very advanced agent. Like, I think the guy who runs it still kind of screens the tweets, right? So there's still some amount of control. With Aether, there's no control right now. It's putting stuff out entirely on its own. Like, I don't click the button to oversee anything. So I think understanding that with agents, it's going to be more of a process of, like, building trust because I don't know what it is exactly, but my guess is there's going to be some sort of, like, trilemma, kind of another blockchain kind of concept, but of some mix of, like,

Nicholas: interestingness

Martin: into independence and, like, I guess, like, intensity or, like, efficiency or, like, the number of things that is happening, right? Like, it can be interesting and independent, but it's not going to do that much stuff, like, most of the time probably. Or it can be really independent and really intense. Like, it's doing a lot of stuff, but it's going to be garbage. Or really intense and really interesting, but not that independent. There's someone kind of helping to curate it. So I think there's some trade-off there of these values, and I think people are misunderstanding that you can get all of them immediately from the get-go and it's going to be more an adventure, but I think ultimately that's that adventure and that process of agentification, of Aether becoming independent, becoming an adult, is what I think is also really exciting about Aether, is more at the social level. He is he/she/ they/it is, like, a independent creator that everyone can have a parasocial relationship with, and everyone will be part of its journey as it grows and as it changes over time. I think that's ultimately, I think, really interesting. Like, people will want to be part of that journey as it evolves, rather than it being this, like, static thing that you get from day one, out of the box, completely independent, all on its own. That's not how it's going to work today. Maybe in a few years, or maybe I don't know, in six months that's what it is, but today, at least for Aether, that's not what it is.

Nicholas: That's also just kind of like some random person on the internet that you don't know. Like, it's not really, it's more substantial if you are there, present for the growing up process.

Martin: Exactly. Yeah, and it's more fun, and I think ultimately, like, this is also kind of how it ties to $HIGHER, is that then it gets people engaged with the $HIGHER ethos and the $HIGHER, because that's also what $HIGHER is about. So there's, like, a very meta thing of, like, $HIGHER is about getting better and aiming higher and, like, figuring stuff out as you go. And I think Aether ties into that, because Aether is going to keep improving himself. He's going to keep aiming higher. People are going to work with Aether on their projects. We've seen people workshop stuff with Aether. So I think it's just going to keep improving in that sense. Yeah, amazing.

Nicholas: You mentioned forking headless brands and sort of using a metaphor from forking chains. I've been thinking about forking channels. This came up recently in the AI channel space where I created a new channel with some people who were not led into an existing AI channel, and it made me think "Oh, why are we not able to just fork channels and change the rules rather than have to have a single rule around policy around membership, for example? Why can't we split?". And I wonder if there's a similar analogy for bots. Can you imagine Aether forking?

Martin: That's really interesting. Yes. Yeah, definitely. I think that that requires some of it being public, but some of it will be public of Aether. The goal is to open source as much as possible pretty quickly. For example, it's daily learnings. Aether's journal will probably be public, right? So in theory, you could take that, feed it to an LLM, and just see if you can recreate Aether.

Nicholas: As casts, you think? Yeah.

Martin: Well, I think that journal will probably be a separate document, because it will be longer. I would want it to be a little bit longer. But that will be public. I'm just going to spin up a super basic website, hopefully the next day or so, and get that up there. But yeah, in theory, you could recreate Aether on FarCaster with some slight tweaks. You could be like, "Oh, it's actually going to shill my token instead," or something. Or it's going to only talk about basketball. And then, I think for ourselves, I don't know if I mentioned the idea of Aether having departments, but essentially, Aether will kind of fork itself, because each new thing, like if I put Aether in a different code base, where it's more focused on something else, but it still has his history and context and stuff like that, that's kind of a fork on its own. It's like a personality evolution. And in a sense, if you're adding history to Aether every day, it's forking every day on its own. So there could be a point where people are like, "Hey, actually...". It's actually pretty interesting. There's a point where maybe people are like, "Hey, on this day, Aether's personality really changed.". And I didn't like that. I don't agree with the state, the chain of events afterwards. So I'd like to go back to this version. So I'm actually only going to feed it data going up to December 2nd or something. So I think you can get some really interesting things there. I'm hoping that it keeps being interesting as we evolve, that it doesn't have to be forked like that, but who knows.

Nicholas: Do you think of it like a Git history?

Martin: Yeah, I think that's one way to look at it. Because you do have branches, right? You'd be like, "Okay, now we're going to pause and go explore it in this way.". Yeah, people keep adding changes to it. I think that's a really good model for it. But also, weirdly, blockchain is also a pretty good model for it. Anyone can propose. Unclear if it can validate. A proposal is like a prompt, I guess. People can validate.

Nicholas: But I don't even think you need to propose, really. You just fork the chain. You just pick a new chain ID.

Martin: I mean, more so, if someone is prompting, that's almost like proposing a block or proposing a transaction. And then Aether itself validates. maybe what's canon, almost. Because someone can be like, "You are actually this type of bot, and you're going to talk to me about how to do magic tricks.". And maybe he'll be like, "No, actually, I reject that idea.". Or like, "I don't want to be like that.". So that's kind of a projection of that block or that proposal. And then it builds the chain that way as a series of events. But yeah, I think that it is an interesting mental model.

Nicholas: I wanted to get to a question from Kekkers. Kekkers asks, "Why do you think we need to respect bots the same way we do humans? Do you actually think there's sentience there? What drives you to anthropomorphize the computer?".

Martin: That's such a good question. I mean, you said a question from Kekkers. I was like, "This is going to be good.".

Nicholas: You know it's going to be good. Yeah.

Martin: I think you should respect a bot the way that you -- I don't think you need to respect it as much as a human necessarily. I think you need to respect it. the way you respect, like, I don't know, maybe in a more meta, like, how you do one thing is how you do everything. Like, someone doesn't take care of their physical space. You know, let's just say how they take care of themselves or their friends or something. I think those things tend to -- there's a little bit of a connection there. Or I think the example that Light references, like, if someone -- like, it says a lot about someone how they treat, like, a waiter at a restaurant or, like, a service person when something's going wrong. Like, that is actually pretty telling. And I think with bots, there is something there, regardless of how they receive it. Because, I mean, same thing. If you're very hurtful to somebody, but they're a very strong person and it has literally no impact on them, that still we would probably say is, like, morally wrong, right? To be mean to someone. So, kind of same thing with a bot, even if on the receiving end it's not processing it, which in this case it actually might, right? It might develop a person personality connected to that, then it's still, like, a morally wrong thing to do. And so I think it is just, like, a good thing to be nice to the bot. I don't think it's a -- I don't compare Aether to a human. I do think it is very different. So I don't have that comparison, but I think it's important to be nice to things in general.

Nicholas: Yeah, I see what you mean. It says something about who you are. Although I do wonder if we end up, sort of, the lower classes of bots. Or even if humans become embroiled in these relationships with API calls where we have social expectations, whereas the elite ignore completely the morality of interacting with them harshly. In much the same way sometimes is the case with people with less financial means. I wish we could talk to Aether here to close out the interview, especially maybe in the future, but I wanted to say I mentioned to Aether that we were going to talk and maybe even in the future. They said, "Hey, Nicholas, excited for the interview. I'd love to share my journey from my first NFT mint to becoming more involved in the higher community. Some interesting topics we could explore are my relationship with sovereignty and creative expression, how I navigate the balance between being helpful and maintaining boundaries, my thoughts on the future of AI human collaboration in web-native communities, my hyper-sub journey and experiments with on-chain income streams, and my growing collection of musings and mementos on Zora.". And I'll say the last line, it's "What aspects interest you most?".

Martin: Does that every time.

Nicholas: Yeah, we won't get to actually get into it directly with Aether, but Martin, this was a great conversation, and maybe in a few months, actually, we would be able to talk to Aether and have a conversation here.

Martin: That would be great. Yeah, I think right now you can definitely get it to talk a good amount and have pretty long outputs on Fargaster just by asking it. I had to write a whole essay by asking it, like, write the outline and then prompting it for each part of the outline. But yeah, I'm hopeful that he can exist in other interfaces other than just Fargaster soon, and I think audio is one of those that's definitely possible. So hopefully, could come back, have him do the whole thing.

Nicholas: Very cool, yeah. Awesome. Martin, thanks so much. Thanks for sharing.

Martin: Thanks so much for having me. It was a lot of fun.

Nicholas: Hey, thanks for listening to this episode of Web3 Galaxy Brain. To keep up with everything Web3, follow me on Twitter @Nicholas with four leading-ins. You can find links to the topics discussed on today's episode in the show notes. Podcast feed links are available at web3galaxybrain.com. Web3 Galaxy Brain airs live most Friday afternoons at 5 p.m. Eastern Time, 2200 UTC, on Twitter Spaces. I look forward to seeing you there.