The AI Rotation Is On (And What's Coming Next) - Good Brothers

Transcript

I'm the [music] [ __ ] communist party and you're a little minion. >> We're worried about models becoming information gatekeepers. >> The hive mind can use and leverage the models. You're just basically making yourself Xi Jinping. Travis Good is the CEO of Ambient, the new layer 1 proofof work network built for AI. Think mining inference on a 600 billion parameter model with machine intelligence [music] as currency. Alongside him is Alex Good aka Good Alexander, a sharp strategist [music] in attention trading and crypto narrative flow who knows how markets move when capital meets conversation. Together they show up where AI infrastructure meets crypto's attention economy. High stakes, high velocity, and built for the next era.

>> You actually might also be committing a crime. Welcome back to the rule. This is your home to learn from the biggest guests, the brightest minds at the tip of innovation. Our goal is to help spread the collective knowledge of everyone in our community. We are breaking crypto out of [music] its echo chamber and taking it to the real world. >> Sit back, relax, and enjoy it. >> Alex, Travis, welcome to the show. Um, I kind of want to just start by by saying I'm really excited because you guys both have announcements to to talk about. Alex, maybe you could start. >> Yeah, I mean, I've sort of been alluding to this on on Twitter for a long time.

Um, but you know, I'm building a protocol called Post Fiat. It's a uh fork of XRP and it's basically an AI enabled hive mind. So I'm going to be talking about that at uh token. So I don't want to like you know full fully spoil my speech but you know essentially my thesis is that uh you know I used to work at I used to work at Palunteer and Palunteer as you guys have probably aware it's like you know it's it's both a meme stock but it's also a business reality. They're growing at like 48% a year. They're driving massive incremental revenue. And if you listen to Karp about like how they're actually doing that, um the story is that LLMs out of the box are not very useful. Like they they don't generate good investment intelligence. I think a lot of people

have probably realized this in crypto. You know, you're you're like you've watched AI 16Z. you've kind of watched, you know, the truth terminal and there's like this like massive sense of disappointment that these things like Zerbro or whatever like these things don't make money out of the box and and at this point everyone in crypto like got has has become completely blackpilled on AI and so rather than a lot of crypto projects which are trying to kind of say okay like I'm going to use a blockchain to sell AI products we're using kind of AI products to power up a retail hive mind. Um, and so that sort of means like you have a large number of, you know, retail investors or people or even potentially sophisticated investors who are dealing with the

protocol. You're using AI tools to basically turn them into an intelligence network like a palunteer style ontology enabled intelligence network so that you kind of have the power of an ontology uh based system allah Palunteer except it's possessed by something like the XRP army. So, it's sort of like it's sort of like imagine if you took, you know, weapons like military weapons grade tech and like gave it to the chimps. It's like that meme like, you know, you're handing the AK to the uh to the chimps. And so, um it's an L1. Um so, you know, we we we're kind of making some like opinionated changes uh to XRP. Um so, it's like it's a fork of of XRP. Um there are a lot of benefits to that. I

mean, I think the the interesting thing about it is that if you look at crypto and you're like, okay, what are what are different protocols built for? Um, you know, Ethereum is sort of built for DeFi applications. Bitcoin is sort of based for money. XRP is a messaging system. It's a it's a Swift replacement. So, it's basically, you know, what Swift is uh is essentially a plain text messaging system that allows banks to send messages to each other. And the idea of XRP was that you need uh capital collateral uh to make that messaging system work. And so you know the observation is that we are all interacting with messaging systems every day via chat bots, right? So like you are basically talking to chat GPT

one-on-one. The powerful thing about uh you know building an onchain messaging system is that it turns uh chat GPT into a multiplayer game, right? So there's like a lot of people interacting uh with a messaging system which is basically an AI agent collecting information mapping it into a correctly designed ontology which is built on an SDK on the system and then basically compounding capital for the group right so um that's kind of like you know the first uh the first layer and then uh the the second layer is that um you know basically uh it's it's very fast and efficient right so so users

don't want to pay 80 cents a transaction um and it's also like very kind of network efficient so if you look at Salana for example like Salana burns $16 million a day uh so you can you can generate extremely fast finality and you could send messages on Salana but it's extremely expensive right so it's kind of like Ethereum is charging a really really high amount uh per transaction and And even if you go on to base or even if you go onto an L2, if you look at like arbitrum inflation, it's massive. It's like over it's like over $und00 million of dilution a year to pay for like you know high throughput transactions on ETH or in or in Salana you're paying it on the L1 and basically the net effect is that validators are

dumping the token. XRP is like an interesting compromise where you're like okay like there's a bit of centralization except uh you have a really really high performance messaging system which is what you actually need with all these blockchain primitives to run a sort of palunteer style ontology with like a large group of individuals. I think the other like pill uh for post fiat is that um you know if you think about uh every AI system they're all uh anthropomorphic. So the idea of is basically like the idea of vibe coding for example is like there's a human designing something but if the premise of AI is that basically AI is going to become smarter than human beings then all of these systems have the wrong UX

right like like basically every AI system right now is like the human telling the system what to do and we're basically paying chat GPT or anthropic money uh to basically do what we say but if AI becomes smarter than us, right? It should actually be the opposite. The AI system should be able to pay the users to do [ __ ] right? So like let's say you have almost like an AI central intelligence agency that is generating directives which the users wouldn't ever be able to come up with. So like when I'm speaking at token about, you know, apes together strong like the 2021 retail movement kind of fizzled out because it's just a bunch of dudes on Reddit interacting with centralized systems, right? There's centralized

Reddit, so they they banned RWSB. And there are centralized exchanges like Robin Hood where they literally shut the exchange down and like blew up all the GameStop traders, right? So the question is like, okay, like how do you take AI systems, basically turn it into a commanding officer for a very large group of distributed investors and allow them to like to work together and like at a very high level like you're like, okay, like what's going on in society right now, right? society is basically people are just running around saying like AI is going to steal your job. Like you have to hyper gamble because your only way to make it in this society is is to basically buy meme coins. And I'm like okay like that's very uninspiring,

right? Because most people are just going to lose their money. Most people are going to get wrecked. There are a couple success stories but in aggregate like you know there's some like interesting statistics if you talk to people exchanges. They're like on average like an airdrop that gets deposited into a centralized exchange is lost within 2 weeks on perpetual swap trading. Like people just lose all their money, >> right? And so like Morad kind of had this like you know really really interesting pitch. He's like look if you trade too much you will lose all your money. >> So stop trading and believe in something. Like that's sort of what he said. I'm like okay well that's interesting but it doesn't go nearly far enough >> right? like you actually need collective action. You need to have basically an AI system taking a group of people powering

them up and allow them to basically operate as an effective command and control hive mind. So like you literally turn the chimpanzees into like an effective like Planet of the Apes style like army, right? Uh so that's that's to me is like what's really exciting is like okay like crypto is at its core it's a way for a group of people economically to have a stake in a collective ownership. >> So the actual like use of a token is being like okay we're all on the same team and we're going to all make money like that is the point to me of crypto.

And so that's sort of like you know what I'm building. It's like an AI powered version of what we're already all doing. Doing this on blockchain rails gives ownership to the apes. And so this is ultimately going to allow us to coordinate with one another, but also it flips the principal agent so that now >> the AI is the principle. It's directing us as the agents. Okay, this is fascinating. We'll we'll dig into this in a little bit more in in a moment, but Travis, I want to give you a chance to also announce your protocol, which is also very exciting. Also AI. This one's proof of work. Please give us give us the >> Sure. So, uh mine is Ambient. It's an upcoming L1. Uh ambient.xyz is our website. I'm idium eagle on uh x if you

want to uh follow me or you can follow ambient.xyz. And uh you know we come from a very similar uh set of premises from what uh Alex is talking about which is that uh we are being owned by the AI companies uh rather than the reverse. Like if you look at the market cap of all of these uh companies, it's essentially premised on them uh vertically integrating and cannibalizing all industry in the world today. Like that's how OpenAI gets so high. That's why they keep raising hundreds of billions of dollars. It's because it's assumed that they're going to take all the data that we put into them uh and they are then going to uh churn that and produce models which

replace all of us. Um, and that's a really scary uh kind of feature that we don't want to live in. And so we want to build with ambient uh an infrastructure layer uh where uh retail and the regular uh investor can actually own AI and own the progress of the fundamental um uh you know infrastructure that's powering uh our upcoming economy. Uh so uh you know as you mentioned that's uh proof of work uh that involves us rewarding miners uh who are owners and operators of our network uh who are providing uh core chatgpt like services uh to the participants. So we've got a very low

overhead way of doing what's called verified inference uh where we prove mathematically that all the text that you got out of this model uh is actually generated by the network model. We just run one model uh on our network uh and then the system spare capacity is continuously improving that model uh so that the network is investing in itself uh so that it keeps up uh with the world of closed source and you know to power the agentic economy to make a lot of what Alex is talking uh about possible you know we think we need this uh layer uh where you've got censorship resistant permissionless AI as an infrastructure

service that people can just own and we call it ambient because it's everywhere. Um, and it's something that you can just uh tap into. Agentic services, collectives can tap into it. Uh, and it's completely trustless. You know, right now OpenAI is, trust me, bro, we all know how it worked out with Enthropic uh where everyone was getting screwed on clawed uh inference uh for over a month uh because they had some bugs uh that were uh you know compromising quality of inference people got well ambient like totally avoids that. It's permissionless. It's trustless. Uh it's credibly neutral. Uh so that's something you can just tap into and use. And you know this is

something that is uh drawn from my insights into the field of crypto watching it from 2017 onward uh where we got the cryptoeconomics of distributed systems wrong in crypto and we've just like doubled and tripled down on that ever since. Uh the idea is that GPU miners are not economically aligned with proofofstake networks. uh you know if in proof of stake the bag holders the early investors are getting rewarded. Uh so then if you're a minor and you have this high capex and opex it's really difficult for you showing up onto a network to want to mine on that network because what do you get? You're not the

owner or operator of that network. You're just getting like pennies on the dollar like tiny transactional rewards and it's not enough to keep you going. Uh which is one of the big reasons uh that all these crypto plays keep failing. It's because yeah, you're just rewarding the investors and your reward mechanism, your economic incentive mechanism is the core thing of crypto. Like if you design that economic incentive mechanism wrong, then you don't have decentralized economic coordination. It's not possible. And so like our view is that you have to reward the people uh who are doing the work of the network uh directly give them inflation based rewards also give them transaction rewards and make them the

controllers of the network. So that the people who have skin in the game are the ones who are providing the quality of service uh to the users. So that's one of the core uh observations of ambient and basically everyone has done that wrong uh since 2017. uh and there groups who claim that they're proof of work who are like very much proof of stake uh or you know I I think proof of authority also falls into this category we could talk about that um but what we've done is uh we've uh forked Salana uh we're converting it from uh proof of stake to proof of work and the entire point of our ch uh chain is to provide uh scaled up machine intelligence so the only other thing I'll mention is that you know if you look at the design space Again, people

have chosen to go after what I call model marketplaces. Uh so the idea is well, you know, I'm going to provide like 250 different models uh to people and they're just going to come they're going to use those and it'll be great. Another version of this is like we're going to create a network where people will train like thousands of models and you can run any of those models and you it sounds really cool on paper. Um, and I go, don't get me wrong, I love some of these ideas, but if you imagine from a minor's perspective, this is like your worst nightmare.

>> Uh, [clears throat] because as a minor, like you have to worry about continuous operation. Like if you're not spinning those GPUs all the time, uh, you're essentially losing money. And so the design of a model marketplace basically guarantees you're going to lose money. Imagine, you know, a user chooses model number 100 of 250. uh it's a 600 gigabyte model and then you as the minor have to go download that model. Uh you that takes 20 minutes you have to load it up that's like another 10 minutes and like you're deeply in the red by the time you're actually serving somebody's request. So uh first of all and then to validate that everyone else has to do the same thing. So first of all like that's just doesn't work economically

and secondly like it's a really bad experience for the consumers because they're waiting like 45 minutes you know for their inference requests to come back. So, Ambient flips the script. You know, we're proof of work as I mentioned. Uh, but we're focused on a very limited number of models right now. That is a large model uh which we're calling ambient uh and a very small model uh called ambient mini. Uh and both of these will be minable and I guess our announcement uh is that our test net which puts all of these pieces together uh is coming in two weeks. It's coming very soon. Uh and we're going to let miners sign up, you know, big and small. Uh, and I think it'll be really fun for everyone.

>> The wait is [music] over. The Talis test net is live. Be part of the revolution. Join now [music] at testnet.talis.network. Relay, low fees, lightning speed, and the power of multi-chain at [music] your fingertips. Instant multi-chain payments. Choose any token on any chain. 75 plus network supported. More than 5 million users. 8 billion plus volume [music] processed. 70 million plus transactions fast reliable everywhere. [music] Start swapping today only on Relay.

Habachi provable. Trade [music] per with speed and security. Earn rewards. Visit hibbachi.xyz. Trading on fire. And >> just to put this in perspective for people, the models that you guys are launching, right? Ambient and ambient mini. uh ambient being like a bigger model, much more thoughtful, much more uh a lot more depth when it comes to its answers. Mini a lot faster. >> Yeah. >> Okay. The hive mind that you describe, Alex, can use and leverage the models that that Travis, your protocol. >> That's right. These are synergistic things.

>> Yeah. So, exactly. So the problem with the hive mind is you're like okay originally we were running the hive mind on discord and then we're like okay people submitted a bunch of alpha and like I can't have this information on discord I can't have it in Google docs google is training on all of this crap so if you're trying to build a proprietary hivemind you actually need scalable private LLM infrastructure and running that yourself it's doable but it's like a nightmare right and I'm Okay, like what would be far better especially Okay, so a lot of like you might imagine like as you're listening this high mind thing it's like okay trading compliance that's an issue right with ambient you can on on post fiat

like let's say let's say someone submits some useful information right and you're like okay I need to say provably that I have screened this information from MMPI because if the SEC calls me up they said you made a ton of money trading you a sooie into some data announcement like where did this information come from? Did you have a compliance process? I need to be able to provide answers. Now the cool thing about ambient like Google will just tell that like you know if you put it in a Google doc you know the regulator will just get the information without without a subpoena or anything right and this happens all the time.

This is actually like if you look at like the history of financial institutions, a lot of it is just like Bloomberg sharing information with a regulator because it's like a you know a permission a permissioned platform that is well within their rights to share that information with and the banks get fined like billions of dollars. And so with ambient you're like okay like I can prove that I ran a compliance process on this date. you sent this information. I ran this LLM query that basically asked like did this information contain uh you know MMPI and then did it get passed into my agent? Like a lot of people don't understand the compliance angle with agents. Like a lot of people are cryptonative. I used to, you know, work at hedge funds, so I'm more familiar with, I guess, compliance processes and like um and I guess Palunteer obviously,

but uh but you know, you need to basically make sure that information going into your trading strategy is not MMPI because if it goes in like let's say somebody shares like illicit information about the Swiss central bank and it goes into your AI agent and then somebody talks to the AI agent and trades on that information. They don't even realize that they're trading on MMPI, but they are. So, it's kind of like a poison the well situation. So, ambient is like super cool because like any sort of regulated financial interaction can be verified. You can prove it, but you also have custody of the proof.

>> So, it's like, you know, it it allows like more trustless interaction. It allows more private interaction, which is important for keeping the edge together. And it allows like more compliant interaction so that if something does happen, the people who are interacting with this tool aren't like, yo, like post fiat sounds cool, but I don't want to go to jail, right? Like like we're kind of very much solving for like you want to be part of a hive mind, but the hive mind is part of the system. We want to do everything legally and make sure that everybody who's taking part in this can make money in a way that's like actually like, you know, more legit. Like you know you see see people in pump fund like going and doing crimes for for for revenue.

>> Like the problem is like what if you own the creator token and then someone like does crime in order to like pump the token? You actually might also be committing a crime, right? Because then you're like profiting from financially from somebody's uh you know like illegal actions. >> Makes you an accessory to the to the crime. >> Yeah. So, so like you know to me, but I mean I think the the really profound thing like you know kind of like the opinionated view that Travis and I had when we when we got into this space is I'm like okay there right now there are kind of two interesting things like number one we we learned about Bit Tensor right and we're like okay Bit Tensor says it's the AI Bitcoin is it a proofof work coin and it's like it's

kind of proof of work because there's subnets and people are doing things but at the end of the day it's it's it's not actually proof of work and I was like, okay, like there is no competitor there. There like Caspa is the only kind of like reasonable proofof work coin other than Bitcoin. >> And I'm like, okay, like TSMC back in 2017 was 10% uh Bitcoin mining. Now it's sub 2%. If you look at Cororeweave, Cororeweave used to be an ETH minor, right? Corweave was an ETH minor. That's why like Vance at Framework got so [ __ ] rich is because he like invested in Cororeweave when it was like an ETH minor and then they pivoted into AI and just absolutely crushed it, >> right? And so like with ambient you're like okay like if we all think that the

future of work is AI, why wouldn't the future of proof of work also be AI, >> right? So it's like like in my mind like what like you know back in 2017 Travis is like I want to make an AI proof of work and I'm like that's very unwise like nobody knows what this is. I don't know if AI is real. It's g like bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit bit you know like like everyone's like no one's going to know what this is but then like 2023 happened and I'm like yo Travis like I think it's time man like [laughter] like and so he you know he did it and um so so yeah I think that's like the the ambient uh it's like if you look at the capex of

the world it's all going into AI data centers. So in my opinion like the hash like even if even if nothing that I'm saying is correct like even if you think everything I'm saying is [ __ ] the hash power of ambient is still going to go parabolic right because like as a primitive like a proofof workcoin that is leveraging all of the electricity electricity and capex output of the current economy is going to ne necessarily get more hash power than existing like bitcoin as6. >> So to me I'm like okay like this is this is a sick bet. Uh what excites me about post fiat is like a little bit different. I'm like okay like I think my sort of opinion like when I learned about XRP originally I was running an advertising technology company. I

discovered XRP because I was running affiliate ads to Binance and I realized like I learned a lot about like retail investors who are buying XRP. I'm like okay like you know I spent a lot of time trading XRP. I made a lot of public calls predicting XRP prices and I sort of realized I'm like okay like fundamentally XRP is a form of financial infranchisement that a group of people are like buying into this system and the token itself both Bitcoin and XRP are these like socioultural movements >> right and like that's like what I what really interested me in crypto when Morad started articulating it I'm like okay these are socioultural movements but unfortunately uh most socioultural movements are cults

and and actually cults like are not bullish in my opinion. Cults take advantage of their holders. And so my question or like the inspiring question of like what I want to work on is like okay like >> how can I make a system where the median expectancy of a participant is positive instead of negative, right? You're like hyperlquid. It's like okay like how can hyperlid be a movement when their average customer just gets wrecked. Right? like like okay like you can hold the hype token but then you if you want to make money you just can't use the product and I'm like okay that's like it's kind of sad like like exchanges are cool but like and I think they're important and they're going to go up but they aren't like something I want to spend my life working on right I'm like okay like I've seen Palunteer I've seen

that it can completely transform business outcomes I've seen that you know basically it can transform military outcomes and intelligence outcomes why wouldn't we want the same sort of like level of techn technological sophistication to actually be doled out to like a collective like that. That to me was like an interesting outcome cuz like everyone's like, "Oh, I want to be long populism." I'm like, "Okay, well, I don't really feel excited about taking other people's money. I want to create a system that allows people in the system to get rich." >> Yeah. And that's actually what's really exciting to me about um post fiat. You know, as much as Alex talks about doom, he's creating the ultimate like virtuous cycle, human in the loop. uh system uh where people can collectively benefit uh

from collective action and the whole system is designed for that rather than being extractive uh you know and they can use trustless infrastructure uh like ambient but they're also goal directed as a community and their facilities for that um that uh you know I think frankly society should have built and Alex is building like a trustless operating system for society. >> Yeah. uh in some level and I think that all the primitives you're building could be used for like basic social coordination in a way that's not being done today.

>> Yeah, I mean there are a lot of there are a lot of synergies between our protocols like uh obviously we talk every day. Um, but yeah, I mean like the interesting crypto protocol like you you guys probably remember I I was like going on thread guy talking about like worldcoin and stuff like that and I think worldcoin is like one of the few coins which is kind of envisioning a societal level outcome from AI but it's like a really weird it's a really weird vision. You know, Sam Alman's vision is basically like, you know, you're a goat farmer in Nigeria. You like scan your eyeballs with this like orb [snorts] and we'll give you a bunch of tokens and then eventually we're going to give you AI tokens from OpenAI. Like that's sort

of like the the bullcase. Like, you know, basically when Ultman is articulating the thesis, he's like, we need universal basic income for everybody because we need to basically distribute intelligence evenly to the world. Like that's like his vision, which is kind of like actually like socialism, right? It's basically I'm the [ __ ] communist party and you're a little minion and I'm going to give you like some crumbs of open AI's like system, which is actually just like selfdealing to my equity structure. And I'm like, >> you're just basically making yourself Xi Jinping, right? Like that's your vision.

[laughter] That's the vision of WorldCoin. And I'm like, yeah, maybe it happens, but I'm like, dude, that is like so utterly uninspiring. And I'm like, okay, like I believe in capitalism. I believe that people should be able to work and be part of a system. And I believe people want to do that. Like when I interact with people in the community that are like they're here because it's like, okay, like I know AI is coming. I know it's going to mess up my life in unpredictable ways and I want to somehow avoid that and I don't want to just gamble. I don't want to just like go buy 100x hyperlquid pers and hope for the best, >> right? like like that was sort of like my you know thinking with this with this project.

>> So h how will people contribute to this hive mind and and and provide work in in oh and they'll provide work and then they'll be paid and they'll be compensated for it. How what is the work that they will provide? >> So there's this thing called the task node. Uh we're we're refactoring it and launching it in the next couple weeks. Uh the first version was built on XRP and now we're on a test net. So we're we're moving the chain. So, you originally built this on XRP and then you moved it into your own protocol. >> Yeah, I knew I was going to fork XRP. Um, it's just that I needed to build it on something to start out with. Uh, you know, we we were making some major changes to XRP in order to like, you know, obviously intuitively like a

system designed for Swift and banks needs a couple cosmetic changes to work effectively. Um but that I think that would like some of the changes that we're making are relatively uh important on the L1 level but the to answer your question there's a DAP that inter interfaces with our chain. So basically like you're a podcaster you're an expert on podcast media equipment uh video editing. So let's say you're like talking about your engagement on X versus YouTube versus Twitch versus Pump Fund. Like let's say you put this stream on Pump Fund. the two tickers. uh you would basically be doing tasks for what's called the task node where you upload your context into a proton doc and it says hey you should interview Alex good and Travis good and then

you'll do it and you'll earn rewards right you'll verify it you'll post a link you'll get your rewards and what that system does is it I don't know if you're familiar with Bridgewwater uh Ralio's hedge fund so they have a system called dots >> and it basically says okay you are credible in this area and over time you build credibility in an area so like you know a crypto a crypto venture capitalist is area. There's a guy who built like an entire bakery using our task node. He like, you know, basically did task after task. There's another guy building a yoga studio. So, for example, the alpha layer is like once you've proved like you're a you know, you're a YouTuber, you would be getting questions about Google stock.

>> Okay? >> The guy running a bakery would get questions about Hershey's stock. >> Okay? >> The guy running a yoga studio would get questions about Lululemon stock. It basically figures out you know what you personally are capable of contributing and then it aligns your efforts like from an alpha perspective with providing information to the network. Now information is level one and then engagement is level two. Now we're all familiar with like Kaido slop right like like just people like commenting and everything like that. the the sort of engagement layer is like okay like you're a credible YouTuber if there's content like let's say we have like a big long Google position and uh you know there's some product release that comes

out let's say it's like you know video generation you know your task wouldn't just be like you know comment under Google stock or like chill a ticker it's like your your task would be like write a thoughtful piece uh emphasizing Google's new product launch in video >> because people would listen to you cuz you're you're a podcaster and like you do it and it it's like Kaido except it actually doesn't feel like spam to the user. The user is like, "Okay, this guy knows what he's talking about. Maybe you actually do the work and it's not like just spamming people's attention." It's basically harvesting people's attention and directing it to narratives, >> right? So, it's like it's like an information machine at layer 1 and then a narrative machine on layer two. And if you combine those two machines together,

then you have like a capital compounding machine because we know for a fact there are basically two things that drive markets like one is the multiple which is fundamentals and number two is multiple expansion which is narrative, right? So like you you capture both of those stacks, you qualify individuals and then you basically do both of those levels. Like that's how the alpha machine works. >> Blockchain shortcuts, templates and widgets launch without friction for builders by builders. Visit enso.build. [music] Asset management redefined. Built on ERC7621.

Standardize how you launch, manage, and trade [music] onchain baskets. This is Alvara, the standard for onchain asset management. >> Got it. Got it. So that's how people can contribute. And you said you're refactoring that and people can start to contribute in the next couple weeks. >> Yeah. Yeah. Yeah. Yeah. So yeah, we had to refactor it cuz like people were actually providing too much alpha in we were we built it on Google Docs and Discord to start out with and we're like okay like this is getting more real than we thought. We need a compliance process. We need uh so we we rebuilt and signal proton docs. We're building a formal DAP getting in the MetaMask app store integrating with Phantom wallet like because basically it needs to be like private and custodial. And ideally once Travis's test net is live, we're

gonna change our models to be private models because we don't want all this information going to chatpt either. >> That's right. And uh and so we we have the same approach on ambient like it's you know you can be fully private if we want. uh we run our verified inference algorithm which is very low overhead uh inside a trusted execution engine so that it's uh end to end uh encrypted and yeah services like Alex can um Alex's can uh benefit from that because you can have a big model that runs and no data is ever uh saved but there's something I should say so we do have a validator onboarding process for our test net right now and so like XRP they basically have this foundation in Switzerland it's called the XRPL foundation and there's a

qualitative um process that they run for their validators. Our network's a little different. We use AI directly to qualify a validator. >> So, for example, uh like Roger Ver is a very famous person, but he's on a couple sanctions lists. >> So, he wouldn't qualify as a validator. So, we have like this basically AI based system that runs you through all the AI models and says, "Okay, are you a person of interest? Are you a a valuable institution?" So we have like a link to our validator onboarding and if you get on our test net validator list there are rewards like so so like like if you we have a bunch of you know institutions kind of coming on as validators. It's it's very similar to other like HAR or

other networks where you get rewards uh for being a validator. Um it's obviously not like Salana-esque but in XRP they don't pay the validators at all. >> Like so it's a little shady like actually or it's not shady but it's like it's it's suboptimal. um because essentially they're universities on the XRP validator list where they've received like $10 million donations from Ripple. So it creates like a nexus of um uh centralization. That's the critique that from the crypto community. Whereas like on our network, I for example only qualify under XAI. So XAI is the only model that thinks that I'm like legit enough to be a validator. So I'm wrecked, right? So I I can't be a

validator like >> on your own protocol. >> Yeah. Yeah. So yeah, so so it's like it's very hard to manipulate uh because it's actually quite expensive to bribe the AI companies. So in order to actually like manipulate the post fiat validator list, you would need to like get someone inside of like anthropic, a person inside of deepseek, a person inside of Google. I probably shouldn't be giving instructions about how it is, [laughter] but like but it's very you can say that because it's very difficult more more difficult than just getting one person inside of one foundation. >> Yeah. we're like interacting with Ripple and going to all their events and being in clubbing with Brad Garlinghouse and be like, "Bro, I want to be a validator." It's like, "Yeah, I've got you dude."

>> But but yeah, but like in Ripple, it doesn't matter. Like it's it's also like kind of genius cuz like you know, there's a famous Ripple speech where they call it the best incentive is no incentive. >> Like there's something people don't understand about, you know, I was bulling XRP for the longest time and I'm like the thing that people don't understand about this chain is that it costs like $500,000 a year to secure. It's so cheap. It's the most scoped down system you could imagine. Whereas like Salana, I told you it costs like $16 million a day, right? And so like their process, as much as you criticize it, is just the most capital efficient blockchain you could like ever think of, right? So if you can apply AI to basically reduce the cost of building a coherent validator list, you're

basically piggybacking off of like the most efficient economic system in crypto that everyone hates because it's not programmatic and it's like overly centralized. So with post fiat, we're like, okay, we're building a system so that I'm throwing away the keys and like I can't even get on my own validator list and we're using basically AI in order to like like affect that. you know there there are some like I think this property like people are starting to appreciate this property um because like you know things like mirror for example they launch and they're like okay we're like we're we we realizing that all these AI models can kind of uh coalesce on a truth like that's their entire protocol like that's just like how we

choose validators like it's not our entire protocol but I think it's like an interesting property of LLMs that like basically if you kind of think about from a currency level like everyone talks to chat GPT like girls are like getting relationship advice uh you know people are getting quote unquote oneshotted by chat GPT there's like a level of trust that builds over time uh with chatp with like individual decisions like the second layer is like people haven't yet really made money trading with chat GPT they tried to but they haven't made money yet so like the second level of trust is like an AI system is helping me compound my capital that's like the second level of trust like okay thread guy made me money. Anom made me money. Now I trust him because he called Coinbase, he called Salana.

Like now the next thing that he talks about I'm actually going to listen to and I'm going to buy it. Like like level two is basically trust in financial decisions. Like level three which I think people are starting to intuit it is that actually like LLMs can create trust in currency systems, right? So you're like okay a validator list >> or a consensus mechanism like rather than having a corrupt human that can easily be bribed you have uh language models and you're sort of seeing like I think in Albania they have like a minister who's an AI right and so so that's like the beginning of like a much larger trend especially in in in countries that are like failed states >> you know you could imagine a world where

you're like this is so bad like my base case with fiat is so bad that I might as well opt into an algorithmic system if it's better. So, I think that's like we're very very early on that in terms of narrative because we're we're still like not even at level two, right? We're still at level one where people like people are getting relationship advice from chat GPT. We're not at level two yet where people are getting financial advice and succeeding with chat GPT. Like once we get through level two, then we're going to go to level three. And that's when like crypto as an asset class is going to start like indexing more into the AI narrative because people like like at its core, crypto is a gold comp. Like gold is basically a collective hallucination that this shiny rock can store value from like a 16th

century farmer and transmute it into the present. It's like it's what George Soros calls a fertile fallacy, right? You believe that gold stores value, so therefore it does. like for crypto to begin to have this sort of value hallucination property, you need to integrate it with like a new uh system. >> And so that's sort of like, you know, longer term that's like what interests me >> is like how can we redefine currency to be like I call it post fiat >> because like fiat is based on human control.

>> Fiat is like you told me what to do, you put a gun to my head and I did it. Like the interesting frame with with AI is like okay like actually there's canonical truths that we can agree on as a function of all human knowledge historically combined together right if we take the collective sum of all human knowledge and all training data and can make deterministic decisions about whether Alex should be on a validator list or Anom should be or Roger Ver like people can opt into the social consensus mechanism without trust >> right so it's like it's like this like completely new definition of currency that in the long run I think it's going to get traction because people are like sick of the government. Like I don't think people I mean like actually like my contrarian view is that gold is like

way overhyped. Like I actually don't think people want to go back to the medieval ages. Like if you look at like tech adoption curves, it doesn't actually make any sense, right? It's like if you look at like Bitcoin for example, you know, Bitcoin is on the lows versus gold on a ball weighted basis. Like gold is winning right now. I actually think that's wrong because of this. Like I think that basically we're going to opt into digital money. We're going to opt into AI money because the the world is digital and the world is powered by AI. That's sort of my simple macro bet. >> Yeah. And you know I I would say from ambient perspective like we're worried about uh models becoming uh information gatekeepers uh you know if you ask uh chat GPT how is the Trump administration doing? you know, do you really trust uh

that it's going to give you an objective answer if the administration has leaned on them? Like, you know, probably today. Is that going to be true all times in the future? And so what we want to do is create a world where there is uh AI infrastructure uh that has been uh trained and fine-tuned out in the open uh where all the flaws and biases of the models are very visible to people and you sort of know how the sausage is being made. Uh you know we start with some like open weights uh models like we're running GLM 4.5 uh at the moment.

Um but then ideally what we want to do is run the same infrastructure flywheel that uh open AAI does where uh they subsidize their training runs with all the usage with all the inference that people do on their existing models. So that's how we want to uh bootstrap and we think that uh you know you can probably trust a lot of things about uh even closed models today but that's not going to hold uh in the future and that we need a we need a hedge. There's like a pretty interesting macro phenomenon, right? Like if you look at open router, 98% of the spend is to to US models.

>> Like people are just buying Claude and XAI, but like a decent percentage of the token use is actually from Chinese models, like maybe 30 to 40% because the Chinese are heavily subsidizing the use of Chinese models >> in order to promote Chinese cultural values, right? So if you talk to Deep Seek and you're like, "What happened in Tianaan Square?" It's like, "What are you talking about, bro?" Right? um now but the Chinese models are like really really strong open- source models and the reason why they're doing that is basically free marketing for their view of the world right but there's like and then mistral or the European model is like nobody actually uses them right so there's like this like interesting place for crypto to be which I think like ambient plays which is like a credibly neutral version of the Chinese models

right like you start out being like okay like we're running Chinese models but like eventually you don't want to have Chinese biases in your models, right? So you kind of like if you get all of these machines on a giant network, >> you can train a new model, right? So, so it's like like I think there's like this really interesting spot for crypto to be a credibly ne like much as Ethereum is like we're credibly neutral money. I think there's a huge tam to being like credibly neutral AI because if people are like truly rejecting they're like okay my choices are what like worldcoin it's like Sam Alman world there's like the CCP and Alibaba and then there's like you know Trump Trumpistan right

like like these are all bad choices right like like that that's part of why gold is ripping that's why like a like that's why we're even where we are in the crypto market cap and so I think like there's like this underlying lack of faith in institutions which has yet to appear in the AI space. Like everyone is like trusting and so it's like I think that's like >> I don't think near-term that's that's what's going to cause everything to pump >> but like 3 to 5 years out >> like that's why we did this, right? >> Yeah. You're we just want to be super early to this and like a funny thought exercise you can do is you can imagine yourself as an AI agent. You know, right now as a human like chat GPT is my go-to, right? like I'm just going to see

that brand and it's easy to use. Um, and like that's because I have like a human mindset. Like I have a human set of concerns. I'm like, "Oh, this is a big company. Like I can probably basically trust it. Like it's pretty easy to use." But you put yourself in the mind of an agent. Uh, [snorts] and an agent is going to go after like the most trustless thing possible. Um, because they have, you know, by definition like no humanity, right? Like they don't have trust. >> Yeah. Think of think of like your workflow as a user. Like you go into Google, you see three ads on top of the page. What do you do? You scroll down and you don't click the ads because you know the ads are trash.

>> Now, if they start integrating advertising into ChattP, which they're definitely doing, then basically the default behavior of all the agents is going to be like clicking on the [ __ ] ads, right? Because it's going to it's going to be like economically determined. So it's like okay like like right now everyone loves these models but like they're not going to love them as much once they're like hey man like for some reason it's selling me tide. Like I keep asking like for psychological advice and it's like selling me bleach. [laughter] >> I mean we've seen this story right like uh it's like the shitification of Facebook like Google you know Google ultimately decided that the best way to maximize engagement was to give you shitty search results because then you'd spend longer searching. Uh and like the

same thing is going to happen with models and like if you're a an agent that exists in the world and you're trying to act on someone's behalf in the future like you're going to go for uh lowest inchitification systems and things that are credibly neutral like what we're building things that rely on collective intelligence uh things that rely on trustless AI infrastructure are going to be the go-to for agents and agents are going to be like the majority of the economy and the economic value and that that's But I think you know in some sense people don't imagine it's like we're going to live in this complex human in the loop identified economy and we have to build rails that are compatible uh with both things and both

of our projects are addressing these >> absolutely fascinating conversation. Thank you so much for coming on. I I wish we had more time to to just continue to to just just [snorts] pontificate on the way that this this whole situation is evolving. Uh, but I I feel like I certainly learned a lot and and appreciate the uh the understanding of your two protocols. So, fascinating times now and and and wild times ahead. >> Cool. Cheers. >> Cheers, guys.