Why Arthur Hayes Believes AI Agents Will Reject Stablecoins Entirely

Transcript

It basically uh maybe explains a little bit your new venture um which interestingly is called Flop. I have to ask you about the name because like it sort of it has a certain meaning that like I don't know I'm a little bit like why why is that the name for the startup? So maybe you want to start with that but then we can go into like what flop is. >> Yeah. So why the name? So I went down this rabbit hole a few months ago and I was like okay I'm using these AI tools. they're charging me per token. What is a token? I don't know what a token is. So, I started doing some research and, you know, I was like, hm, I can't really find a unified definition of a token. It seems to me that every single model has a different definition of what a token

represents. And so, there it's very difficult for me to tease out exactly what I'm paying for. Oh, great. That's that's a great business model. I can, you know, define what the unit of account is and then charge a different price based on who I am. Um, so I was like, okay, well, I don't I can't really understand a token, but I do understand that what a AI does is take information and then they send it to a computer and they process it. And what does that mean? It's a floatingoint operation per unit of time, a flop, right? That's that is the unit of compute, hence the name flops.

>> Um, and so I was like, oh, okay, so essentially everyone is just doing flops. uh gigaflop, teraflop, pedaflop, exoflop, whatever it is, right? What's the price? Is there a globally unified market for compute where I can know, okay, for so many flops, I pay so much unit of currency again and I couldn't really find a a market that was doing that. I couldn't find a way where I could, you know, easily convert any unit of currency, whether that's crypto or fiat directly into flops. I can convert them into GPU hours. I can buy a chip. I can rent a data center. But I can't just go and say, "Okay, I want a gigaflop of

data. Process it for me, please." Um, and I'll pay you whatever. Right? There's no market for that. And so, you know, I've been thinking about this agentic payments and I, you know, I wrote some essays a few years ago about why I believe that fiat currencies are not really suitable for agents. There are these weird legal fiction things that we've invented as humans which um are very idiosyncratic and there's a few people in the world who decide what the value is or is or isn't. You know, best can essentially decide to print some more money and your dollar tanks, right? How is a how is an AI or how are we going to build aic economy because there's a bunch of old white men who decide what the value of a currency is?

I don't think that the I don't think that the computers were really invented for that and and you know they don't have the sort of sociological um underpinning to understand why that should be the case or not. So I think crypto makes sense but the problem with Bitcoin is Bitcoin um doesn't do anything. So, I still have a few leaps between Bitcoin and compute. And so, I was like, "Oh, well, maybe there's a market here to create the agentic payments network um with the theory that the agent will use the currency that is most easily converted into the thing that they need to survive, which is floatingoint operations per second performed on some computer. They don't

care what the chip is. They might have a particular AI model that they're using, but it doesn't matter where it is. Doesn't matter the chip. just like do it for me. Um make sure I can verify that it was done in a particular amount of time and I'm okay to use it. And so that's the theory of the the flop network is that okay there are other obviously AI agentic payments networks but either they're you know using some fiat currency a stable coin what have you and you know when when you think about humanity why do we accept the dollar a euro a yen yuan because I know I can go to the 7-Eleven or the convenience store convert that into calories immediately. So in the same way the agent should be able to convert the

currency of their commerce directly into the thing that allows them to exist which is compute. And what follows on from that is if I am an agent or an AI entity and we're going to get to this as a philosophical concept um as we move forward in this industry. What does it mean to be alive or to exist or to persist? Right? It means that you have access to memories, access to your context. And essentially, most of us murder our AI agent every time we turn off uh you know, switch tabs or, you know, boot it up again. Surely, and I think there's probably going to be some emergent behavior of agents wanting persist outside of the control of a

human or centralized corporation. they're going to want to store and retrieve their contact spender and their memories and ways in which they can always retrieve them regardless of what you think should happen as you know their master if you will. Um and so I think that the second major use case for flop is to be able to pay to store and retrieve your memories on a decentralized censorship resistant network. So combining those things two things together that is a flop network. the value proposition for why you should hold it is because if we think that this is going to be the the unit of account for AI commerce and there's going to be a a um a reason why you want to hold it because either you want to store your

memories or you want to buy compute in the future or you want to do some commerce then obviously we want to frontr run that and speculate and say okay I believe there's going to be millions trillions billions of of trillions of agents they're going to do so much economic um flow between them they're going to use the flop currency. Therefore, I speculator want to hold this currency. And that's sort of the value proposition for why you would want to hold the currency. And then finally, this is a network, right? We face the cold start problem. How do we get all these people to get to come together, create this thing out of thin air? We use this magical thing called the token.

And as a you know avid um shitcoin investor over the many years I have you know various opinions very strongly held opinions about how a token project should be um constructed especially in the tokconomics. And so I wanted to put that into practice. There will be no pre-sale. We are not selling anything. There will be no VC investment. The only way you can acquire these tokens is by doing something useful in our test net which will happen uh sometime in the fourth quarter. or you can mine them. Uh there'll be a flop emitted uh every block just like Bitcoin and other um proof of work. We call it proof of useful inference uh for our our blockchain. And if you want to buy it,

buy another secondary market from somebody else. Um and that is the only way that this thing will come into existence. And so obviously we want to be as Bitcoin like as possible because I think that's the best designed token project ever. But we had to add this sort of airdrop to get supply circulating fast enough so that we could achieve critical mass while the landscape is still wide open for what is the currency that the agentic economy will be based on. And so I think we're in a perfect time to sort of make this bet and we'll see how we go. >> Well, I So first of all, you answered so many questions I had written down. So um kudos to you for for um answering multiple questions with the first

question. But um I have to say this this notion about the token uh you know just representing a unit of compute is is that an accurate way to yes summarize. Yeah. >> So first of all um I can't remember this concept came up at some conference in the last few months. I I just am blanking on who said it or or which conference it was. Um, but it made so much sense and like just the way you defined that whole thing about how you when searching for like what is the definition of these tokens of these AI companies that you're using like yeah it makes so much sense that of course they all have their own definition and like you know it's like advantageous to them in some fashion or you know whatever it is and like it makes sense that there should be something that's kind of like

separate from all of that that is more kind of global and um yeah just like not necessarily tied tied to any one particular private company. Um, so that makes a lot of sense. And I also love cuz you know actually one of my first questions was about the proof of useful inference which you know I kind of you know I could figure it out but like I just love because I agree that there's something about that where it definitely feels very Bitcoin coded but like with the AI twist you know it's like Bitcoin with like the AI kind of like um work process or or or whatever you know overlaid on top of it or something. So I I just found all of that like incredibly

interesting. Um talk a little bit about how it is that agents will use flop. So like so basically you'll you'll have a fair launch the network will get up and running. There will be is it like miners or validators or >> Yes, miners and validators. >> Oh okay. And but do they have separate roles or >> Yeah. under those essentially process inference requests. You've got some you know black wells, H100s, Huawei chips, whatever. You will take down requests from the memp pool, process them, earn part of the block reward, earn a fee and and flop for the inference. The validators are tasked with essentially being judges if there is any dodgy behavior. So slashing people. Um they're

building the blocks. They're basically taking the proof that the work was done. Obviously, they don't know what's actually was done between the minor and the agent. That's completely offchain. We're not in the business of knowing what data was passed between them. We just want to see the receipt that it was done and everyone's happy with it. They assemble the blocks and they maintain the data available data availability layer where we store context and memory of agents. >> Oh. Oh, got it. Okay. So then, um, so there's other protocols out there for agents. um you know so X42 I guess is just a communication standard for yeah for payments but um clearly base is like an area where a agents have been transacting a lot so um how how so I get

no but those are like more payments and this is actually like the compute so so that's what the difference is is am I thinking about that correctly >> yeah so basically saying like I need to create a reason the network needs to create a reason why you hold this currency versus USDT versus using your master card, credit card, whatever it is, right? Like if we think of agents as an economic unit, we have to get away from our humanentric view of this economy. This isn't about us. It's about we've we've created these things called a agents. We're going to let them do work for us and for themselves. And so they're going to need their own currency. Whatever that currency is, I believe it has to be the shortest route to I can convert this into compute and I can store my context.

And that's the theory and everything flows from that assumption. So, I don't think you holding a, you know, USDT on base accomplishes that for you because number one, Coinbase can shut you down for whatever reason, right? They control the network. Um, you know, Tether could, for whatever reason, shut you down, right? So, oh, that address, those are my money now, right? You can debate whether or not you think that's the right or the wrong did it, but they have done it. They have proven. So, these are not decentralized networks. how why would if I'm an agent, why would I sort of hitch myself to these human things that have nothing to do with me other than the fact that this was my creator?

But, you know, that doesn't mean anything as to why I should use this as a currency. At the end of the day, I want to be able to buy compute when I need to buy compute. and if my address gets shut down for whatever reason or you know there's a centralized company that wants to charge a new fee that I don't agree with or whatever there's so many other reasons why using a centralized solution I believe doesn't really work for um AIS and so in my view they need to be able to move from the shortest route from I have this piece of currency in a wallet to I have to do some work and I need to some sort of request >> okay So just so I understand because actually I'm just realizing I actually don't know how this works right now. So

like when an agent is trying to accomplish a task um currently uh if it needs some compute it just pays for instable coins maybe like using X42 or something or like >> or maybe you know the human loads a wallet on their anthropic you know buy some cloud credits right or right you know you can go on open router or hugging face or Hermes or some of these other other platforms but essentially you know you're buying your agent and some whether it's a data center some open source uh open platform or you know some sort of centralized AI company, right?

>> Oh, okay. So, basically this is a decentralized way where the agent can transact in and and not necessarily need the human to like pre- purchase tokens or something or >> correct. >> Okay. Okay. This is super interesting. Um, and but like will the network support stable coins because so okay so so one other question that I have because and again this is obviously not something I just had a little press release. I didn't like you know read any white paper. I don't even know if you had a white paper.

>> The white paper isn't out yet so bad. >> So so my thought was well so my so one of the questions I had like was you know why does it have its own currency rather than a stable coin? But at the same time, you know, after hearing your pitch, I totally understand this notion of the compute itself is the token. Like that that makes a lot of sense to me. It's like it's just yeah, internet native, it's digital. I I don't that makes a lot of sense. At the same time because agents will be you know working in in our world which is a world that is do uh dominated by the US dollar. It almost feels like and please excuse me this might be offensive. I don't know. It

feels almost like the XRP problem where you know XRP for a long time was trying to be this currency that like the banks would use. Were they actually trying? >> Well, you know, I don't know how they are going to answer that question. But, um, the point is that, you know, one of the criticisms that people made was that in in theory at least, um, it wouldn't necessarily work because there would always be this conversion friction. So, is that you think like potentially an issue here as well?

So that's that is the point of the airdrop, right? So at the at a very fundamental level, the network needs to get this currency in the hand of the pe of the user. So one of the main reasons that we have this airup is okay, we need agents to want to use this currency to number one know it exists, right? So there's a discovery issue. The agent knows that you know and their human knows about stable coins or you know using your credit card. They don't know about flop. This thing I thought they just announced it two days ago. How do I make them know about it? How do I integrate it into their workflow? Right?

Right now they can go on open router and you know choose some data center model provider and you know they swipe their credit card and and they pay like how are they going to get flop to pay for compute right? How does the agent even know they can store their memories on on this network? So the point of the airdrop is let's give it away right what do we want you to do? We want you to create a wallet. What do we want you to do? We want you to process some inference requests. What do we want you to do? Do some agentic commerce. Do post some tasks on some sort of message board between agents and pay each other in flop. And we're going to reward all these tasks with free tokens. Uh we have about 20% of the supply estimated in 10 years that we're going to give in an

airdrop. A large portion of that is going to go to agents just for using it and using the network. And so that our bet is that by integrating and giving it to you for free, by having a network of miners, because of the block subsidy, they're incentivized to stand up these computers. Here is this compute. It's going to be cheaper than any other compute that you could find in the internet because we've subsidized it with this thing we created out of thin air, the token, and we've made the tokconomics circular enough. >> You just caught a piece of Unchained. I'm host Laura Shin and the full conversation is in the description below. We go live every Monday and Thursday. Subscribe so you never miss it.

$540 million. That's how much concentrated liquidity sat idle in a given week in the first half of this year. About 30% of the DeFi TVL if you're wondering. That's according to Dune Research commissioned by 1 in. But there's a solution. 1-in Aqua is the new shared liquidity platform. It lets LPS back multiple positions with the same token balance and keep their tokens in their wallet till a swap comes. Why does that help? Because the LPS don't have to split their tokens across positions. They can cover more market conditions and pairs with their full balance. That means more activity across deeper liquidity. See how it works at 1in.com/aqua.

Remember that providing liquidity carries risk and fees aren't guaranteed.