The $650M Bet on Crypto's AI Future with Haseeb Qureshi

Transcript

Welcome back to the rollup. I'm Robbie. >> I'm Andy. Rob and I met at the University of Florida in 2017 where we first found out about digital assets. >> We learned a lot along the way and we're bringing you face to face with the leaders of our industry. Sit back, relax, and enjoy today's episode. Guys, welcome to Near Con. We are here today in San Francisco. It's Monday, February 23rd, episode 22, Stabled Up. Our guest is Hib. We're talking a payments. You think crypto is built for agents, not for humans. We we had a great panel. My voice is way better now than >> Denver. That's great.

>> We are We are back. We are back. >> Beautiful. >> That moderator was clowning you a little bit when you hopped on stage at Denver. >> He was kind of fumbled it though. >> Well, because you didn't do your rap. You didn't do your whole uh rapper or That's right. That's right. >> Um you guys are big investors in Rain. Um other kind of stable coin products as well. Um, you've seen this kind of stable coin growth play out over the last couple years now to 300 billion. Tether circle obviously leading. Where do we go from here with with this market? Now we have this massive improvement in LLM it in models. How does that affect stablecoin supply, stable coin transaction activity? Where do we go from here?

>> So in terms of total supply, I don't know that LM right now are going to do very much. Like the underlying secular trend is pretty clear. For the last however many years, stable coins have been growing about 50 to 60% every year. Um, I think it was Secretary Besson who last year stated he believed stable coins were going to hit 2.7 trillion by the end of the decade. Uh, that would be roughly 70% year-over-year growth every single year. I think that's pretty aggressive. I don't know that it's going to hit that high of a growth rate, but if you even just project the current growth rate of just, you know, 50 60% every single year, then it's pretty easy to get to multi- trillion. And what that what that basically tells you is that a significant portion of the total M2 supply is going to be just stable coins.

Okay, in that world, where is that demand coming from? It's not coming from AI agents, right? AI agents are still obviously subservient to the human beings that are running them. And so it's it's human demand for stable coins that's going to end up informing what the AI agents are going to be using. So at the end of the day, at least for the foreseeable future, it's going to be more and more B2B payments, more and more remittances, more and more of these, you know, countries that are dollarizing effectively through the back door that are going to drive more demand for for stable coins and for dollarization. Stable coins become the way in which countries dollarize >> as you're seeing more and more geopolitical instability, right? What does that do? That makes it so that people want to use other currencies more. And those other currencies, the

easiest one is a dollar. >> Yeah. Uh there was a great article that came out yesterday. It's gone mega viral like 20 million views by Catrini. You had a chance to read it. I did. I did. So I want to get your take on particularly this one part where he talks about this consumer credit crisis. You know the shift from AI agents kind of spinning up their own interfaces ultimately they need a better payments infrastructure than a credit card. Uh and so he you know obviously this is set in the future and he says you know this this uh Q1 2027 Mastercard net revenues you know positive purchase volume growth slows to positive 3.4% year-over-year.

Mastercard's Q1 2027 report was the point of no return. Agentic commerce went from being a product story to a plumbing story. Mastercard dropped 9% the following day. And he talks about how uh these part the these uh and oh visa uh also dropped but not as much because it's position in stable coins was much stronger. Yeah. >> And how agents will start to attack the credit the two and a half 3% credit card processing fee right and compress the margins there down to stable coins which can be pretty much near instant. Do you agree? Do you disagree? Do you think American Express, Mastercard, Visa are all cooked and they're all going and and should start to think about progressing more towards stablecoin infrastructure

here now? >> Okay, look. So, this this piece is basically like you remember AI 2027? >> That was like AI doomer porn. Yeah, >> this is the equivalent but for finance people. This is like your AI doomer porn for finance people. Um the reality is that so many of the steps that they took in this analysis were like I don't know if it's quite going to work that way and you like chain daisy chain like 10 of those together and it's like well this is what once this happens then this thing happens then this thing happens and that's why like the S&P is going to be down 60% from all time higher whatever it is. Okay so let me take a step back. I think the the the core of the argument which is that basically it may well be that there's such an explosion of consumer surplus

and producer surplus but it's not uniformly distributed through the economy and many of the financial assumptions of the economy are dependent on a certain uniformity of wealth and and uh and and uh money moving through the economy in certain ways and that can cause a financial crisis. That is a point well made. It's a point well taken. I think he he does a good job of fleshing out the ways in which that might look. Now, if if you if you break down each of the individual claims, a lot of them are just stupid. A lot of them are just literally nonsense. Like the idea that, you know, Uber and Lyft are going to suddenly get displaced by vibecoded alternatives, which is like, do you understand what these companies do? Like, they don't they're not just software like they're they're logistics companies more than they are just pure

software layers. Otherwise like this would already be done by all the clones of Uber and lift that exist in every like we already have a bunch of Uber lift clones and they all died. So it's not like oh well if you have AI agents suddenly all this stuff changes. Now all that being said let's take the point about Mastercard and Visa. Um the reality is that as we are going to get more and more agents sitting in front of you engaging in financial commerce there's going to be a two-track world. One track is going to be optin or sort of human approves. Okay, that human approves the world is going to be the what the vast majority of human beings are doing in in two, three, four years.

Okay, that is going to be basically what it looks like today with OpenAI is that when you tell OpenAI, hey, I want you to go buy me a rug, okay? And it goes and searches on Etsy for rugs. Etsy right now has a chatbt integration. Um, if you want to buy this rug, chatbt makes you opt in and confirm, yeah, you want me to buy this thing for you? Okay, you click the button, right? And there's a paper trail of you click the button, you are liable. If you bought the wrong thing, it's on you. Yeah. Okay, Visa is very comfortable with that world. They know how to interact with that world. They know what it means if they get your your agreement and they know what a chargeback means. They know all this stuff. They can model and they can they can they can understand how it works.

Okay, if the agent is buying stuff while you're asleep and there's no human pressing any button and and the agent is just yoloing stuff, okay, that's when all the stuff breaks down. That's when Visa is like, "Okay, I don't know what that means. I don't know what it means when you want to charge stuff back in the morning. I don't know how to price that differently. I don't know how to risk model that." All this stuff starts to break when you have AIS doing stuff autonomously. That's going to be more of the open claw hobbyist people kind of yoloing stuff and living in the world of startups, not in the world of big companies. Okay. So, what what what you're going to see it's it's very much like crypto, right? Today, crypto is a tamed financial asset. Now, it's it's safe enough to live in your Fidelity

account, to live in your PayPal account, right? But before that, crypto was just crazy insanity. And in that world, like even today, think about like Coinbase, right? Coinbase allows you to trade stuff directly within Coinbase like memecoins, Doge, all this stuff. But in the early days, it was just Bitcoin, ETH XRP Litecoin >> right? Because it's like, look, I don't know. I don't know that any of this stuff is safe enough for you to touch. Today, we've we've we've civilized it, right? We've civilized it enough that basically you can take the craziness on unis swap or pump. Fun, take the stuff that's usable and bring it onto a centralized exchange. In the same way, that will happen to what's happening in agentic land. But today, tomorrow, next year, even year after that, it's way too early. way too risky, way too poorly understood. So, up until then, the vast

majority of people are still going to be using credit cards. Okay, this is not a 2028 thing or even a 2029 thing. It's going to take more time. That being said, there's going to be people who you and I know if you're, you know, sitting around reading threads on X, you're going to be reading threads from people who live in the future. >> Yeah. >> Now, eventually people are going to catch up to that future, right? But, you know, it's like the the thing that Chris Dixon always says is that, you know, the future is defined by what people today what smart people today are doing on their weekends and even right so the people who are tinkering and doing all the crazy openclaw stuff and getting it to answer all their emails the reality right now you guys probably saw the story that went viral about the person

from Meta who was um you know she was an alignment researcher on the super intelligence team and her openclaw just started deleting her emails >> and she was and she connected to her real Gmail and she was like wait no no stop stop and she was telling it to stop and it couldn't she didn't know how to do like the hard stop on the on the queue exactly and so she just ended up unplugging her machine after already started deleting some of her emails and like look that's what it's going to look like for the hobbyists today you are basically signing up for potential brain damage you know it's like the people who are using cars in the early days like before we had seat belts before we had any of this stuff they were just dying >> they were just dying and like cars

exploded in the early days and like that being said cars were cool it was it's like really awesome to be using a car in the early days but it took a long time for these things to get safe enough for most people to use them in a way that wasn't likely to result in them getting hurt Yeah. >> So I think that's you should expect to be the same adoption curve for agentic payments and for stable coin payments and all this stuff. Now all that being said, look, venture capital has a long time horizon. I'm very comfortable with that. But I think for a lot of people, they really believe that this is going to happen tomorrow or next month or next year. And I think they're going to be very disappointed that there's all the they're have all these smart friends who are doing all this crazy

>> and they're gonna be like, why isn't why is nobody else doing this? I guess crypto failed. And I think that's that's not the way to internalize how this is going to work. It's going to be a two-track story. >> Yeah. You know, kind of a in a in a tangent to that, we have, you know, this Clarity Act stuff kind of getting closer the market structure bill and there's been talk about having to have these kind of KYC wallets and kind of like guard rails on wallets. like there's going to be certain types of frontends that allow you to only interact with them if you're on like a privy that has a very specific uh you know kind of like KYC or specific kind of like institutional um you know AUM requirement or whatever it is >> flipping to the agentic world you're describing kind of cars on fire you're

describing this kind of free-for-all wild west >> what what what's your thought on the uh agent to agent commerce happening I think on on our show in the beginning of the year you predicted that was not going to be happen this year that was in next year thing. >> What's your thoughts on like the kind of guard rails or the restrictability of agent to agent payments and agent commerce? Like how is that going to shape out? Is that an investable market? Like are you guys looking at that? Like what's your thoughts on like how h how restrictive or how open we're going to allow agents to operate on chain when it comes to payments, stable coins? Yeah.

>> So look, there's a there's a really really easy answer which is just give it not very much money. Yeah. Right. That's like the obvious guardrail that people are going to use, which is say, "Okay, great. You could have 20 bucks. If you don't do anything stupid, it's like, you know, having a teenage son." Right. Right. It's kind of like, yeah, it's like you have an allowance and if you do something crazy like, "Okay, the damage is going to be is going to be pretty contained." Um, I think you're going to eventually see more and more complex forms of controls. But in the beginning, people are just going to do the obvious thing. We're just don't give it too much money and just kind of see what it does. Um, today, you know, when I see a lot of these agentic things, like take for

example Conway, right? So, a lot of people got really excited about this announcement of Conway, which is this platform where AI agents can buy their own compute and basically, you know, kind of live off the land by making money and using it to buy more compute so they can continue to survive. Okay, this kind of makes sense. You can imagine an AI agent paying for its own compute or paying for sub agents or paying to say, okay, you already in in essence, if you use quad code today and you tell quad code, use sub aents to call to solve this problem, you are basically giving your claude a wallet. And what's that wallet? It's the wallet to make all their API calls, right? If you go to sleep and you wake up the next morning and you have Cloud running all night, >> you're basically giving Claude free

reign to use your API credits. Yeah. Right. And especially if you're, you know, if you have a credit card connected, effectively claimed ways today, right? And like there's all sorts of mechanisms within these model providers that basically say, look, you know, you can you can charge up to $50, but if you want to refresh up to $50, like you have to manually go and approve that. So there's there's there's soft ways in which we're doing this today. But the reality is that, you know, people don't necessarily, you know, I know a lot of people who've like woken up and had thousand dollar bills because their claw just like did something crazy and just sort of was running in loops and spinning a bunch of sub aents to do stuff extremely inefficiently. And people who are living on the frontier are getting situations like this.

>> If you're a normal user, you're never going to run into something like that because you're not going to be running the models that hard. But um I think you will see all sorts of different mechanisms tried to create these guardrails around these systems, but eventually they will be baked in closer and closer to the systems themselves as opposed to these very kind of uh coarse things like okay just give it 20 bucks and see what it does. >> Havachi is fast and private per trading secured by ZK with Celestia underneath. Go to habachi.xyz to start trading today. >> Want simple? Want fast? Want every chain? Use relay.link. the fastest bridge in crypto. Transfer your assets crosschain instantly with very low slippage right now on relay.link.

>> For a while, you know, we were kind of oscillating between fat application thesis, fat protocol thesis, where does value acrew, bare metal distribution. Ultimately, I think we kind of settled somewhere near fat wallet thesis, which was ultimately what owned the relationship with the end customer. Now, in theory, AI agents spin up their own or don't even need an interface at all. for just prompting intents via you know an LLM and it's going and interacting directly with smart contract level >> are wallets cooked like is this going to be the primary modality and how do we adapt to that >> I think the answer is that yeah wallets are cooked >> I think like overwhelmingly everything that is human centric is going to get disintermediated right and so if you are if your if your business model is

predicated on human friction >> and simplifying things for humans >> like it's going to get simpler than that >> it's just going to get simpler than that and same way that like you know the first wallet was you know um or I don't know what the first wallet was but you know think something like Mist for Ethereum right was the first Ethereum wallet and Mist ended up getting simplified by MetaMask >> and then eventually you know take MetaMask compared to Phantom Phantom is like even simpler than MetaMask and that okay you're simplifying simplifying simplifying the interface for humans but the most simple interface is an agent >> is just you tell somebody hey I want to go buy this thing and you figure it out >> I don't want to click the buttons >> and Electrum was my Bitcoin wallet that got simplified by centralized exchanges.

>> Right. Right. Exactly. >> So we had um Brandon the CEO of Bravon to start off the day and we talking about microp payments and they're trying to do some interesting stuff with aic payments with near in Brave. Yeah. >> And they've got different like LLM implementations in Brave. Now I don't know if you've seen it. They have LEO which is like their kind of like sidebar pull out and they have like different models to uh prevent against prompt injections as well. Different types of like modalities. Um but thinking about the internet as it is today without payments and then thinking about the internet with integrated micro payments you know on stage. We talked about the the SAS apocalypse where you know users and agents are going to want an outcome based or token based kind of uh uh a way to pay rather than having to do like

like seat based or or subscription based. >> What does microp payments look like in the internet? Like is it Catrini, you know, puts a payw wall on this article and I have to pay 10 cents to read it or is it like is Bloomberg's uh you know $40 a month subscription to read all their stuff is like like how does this actually get implemented? What what does this look like from you know stable coins from the crypto rails? Yeah. >> Yeah. What like what's your vision here? >> Look so I think we are okay the most interesting way to frame this is like in terms of tokens and when I say tokens I don't mean crypto tokens I mean AI tokens. So today the the the probably the most microp payment stuff that we're all fixated on right now is literally like if you use cloud code or you use

any of these services or you're using any of these APIs if you're using openclaw right >> you're no longer allowed to connect openclaw to just like a cloud code account you have to connect to anthropic API which means that you are doing microp payments >> you are paying for every individual token now there's now the settlement is happening asynchronously it's happening with each individual thing but the decision about okay you want to like do just a little bit more perk. You want to do one more call, you want to do one more this, you got to pay some fraction of a cent. Okay, that makes sense, right? When you're billing to anthropic, makes perfect sense. When it's going to be that agents are helping other agents, right? Like my open claw is getting help from your open claw or something like this where there's comparative advantage

among other agents that have specialization and are particularly skilled at different things. Um, that is I think when you're going to see this kind of micro payments have to be settled more frequently. um where it's probably not going to be like I pay you a fraction of a cent, right? It's probably going to be more like I pay you 5 cents, 10 cents, 20 cents to to um you know, solve some intermediate task for me, right? You know, Anthropic, for example, just put out this post about uh cobalt. And uh I don't know if you guys saw, but like you know, there's just a gap down in the value of IBM because of this one blog post that Anthropy wrote about Cobalt. Yeah. Just just literally a blog post, not even a product, just a blog post. Um, and uh, I think what you what you may well see is that somebody

goes and builds some kind of agentic cobalt system and it's not a company. >> You know, before it'd be like a company with enterprise contracts and you have like this enterprise sales cycle and all this other stuff, but in the future it may just well be that like there's an agent >> and you can go and interact with that agent and say, "Hey, this is this is an agent that's like trained on co it's fine-tuned on Cobalt. It has all sorts of rag stuff on Cobalt. has perfect uh uh perfectly optimized prompt and memory just for doing cobalt refactoring and like just hit it. It's sitting on an API. You can interact with it and it, you know, just pay it in real time whenever you want to use it. And this becomes like, you know, pay as you go uh company that's metered in real time and

it may well be 20 cents that you pay it, maybe $500, maybe like a, you know, a 30k, you know, transaction that takes multiple months. Yeah. So now I think that's a little bit um simplistic. That's not going to happen tomorrow. It's not going to happen next year that that things are built that way. But over time, the whole point of AI is that so many of the like why do we even have call me numbers for uh enterprise companies? The answer is because it's it's designed around human friction. >> It's designed around things that move at human speed and human speed decision-m. In 5 to 10 years, that's going to change.

>> It's going to change. like the phone call will not be the API by which enterprise decisions get made >> and when that happens all this stuff will need to get negotiated in real time, right? The beauty of AIS is that you know they can interact with each other and have multiple rounds of negotiation in seconds >> while the humans are asleep. >> Looking for yield that stands out. Infinify gives you just that. Deposit stable coins and earn far more than the current savings rate. Infinify is a battle tested reliable DeFi platform that is ready to go. head to infinifi.xyz.

>> Could you unpack more of the allocation thesis around the new fund? Uh you guys raised 650 million, right? And you know, you're you've talked a lot about how the stablecoin rails are still underpinning this. How much is it still, you know, the kind of the quote unquote boring stuff, the payment rails underneath versus kind of digging a little bit deeper into the AI businesses themselves? Are those an investable category yet? you know, these AI driven businesses or is it still sort of picks and shovels kind of supporting that sector that that we anticipate will continue to grow?

>> Yeah. So, I mean, I think I said this when we were in Denver is that like if you knew that AI agents are going to be coming on chain and there's going to be many many more transactions happening, right? It's almost like uh uh it's it's like a huge source of demand. It's like if I were to tell you that, hey, China is going to unban crypto and all the Chinese people are coming on chain. >> What should you buy? The answer is anything. Like literally buy anything and it'll do well, you know? So I think like the demand for crypto is going to push almost everything up. Now there are some things that won't benefit, right? So like okay, if Chinese people are coming on chain like Coinbase is not going to benefit, right? So like if it is catered to a different population

like wallets or certain things like this like okay they're not going to be beneficiaries, >> right? But who's getting like usurped versus who's the benefactor? >> Yeah. I think the answer is the benefactor is like all the underlying infra. Okay. Right. All the stuff that you have to use just to use stable coins. So, Circle's going to benefit, Ethereum is going to benefit, right? Like all the all the protocols that exist on Unis swap's going to benefit. All these things will benefit because AI agents will use these the same way that we do. Um it's the only question is what are they not going to be using, right? Because they're catered to human interfaces, right? So those things I think are going to ultimately get disrupted, disintermediated. Um now for us, no. So where does that inform where we invest? The answer largely is that

okay, if you believe that as I do, then you invest in the same old you know, you just you're underinvesting into things that are dependent on human friction. Um, but over the short term, I mean, look, this stuff that I'm talking about is like five to 10 years in the future. Okay? It's not like right now, okay? And so, now that being said, we make investments today >> into things that are going to get realized 5 10 years in the future. So, if you're betting too much on intermediation, >> uh, that's not going to go well for you, right? Um, now all that all that being said, on the AI side, it is still pretty early to be trying to figure out who's gonna win and who's gonna lose. Okay.

You can see that as well in the AI side itself. Like in in in AI land, what companies have really shown a defensible mode besides the foundation the foundation models? >> Yeah. >> Right. Almost everything that you see that was like a big explosion early on like think Character AI, think you know uh like all these companies that were really hot in the early days, people just kind of forgot about them. What was that? Uh the the social network that was really hot really early on. >> House. >> Uh no no um >> no no there was one that was uh it's like something tomato or something like that. Yeah, I can't remember the name. It was really hot in the valley backed it and then it just like totally disappeared. It was like where people posted like AI generated images of themselves with other people. Um,

>> you know, >> no no whatever doesn't matter. Um, >> your point though. >> Yeah. The point is that like there's so many of these like flash in the pan moments in AI that we've now already forgotten about and that, you know, think perplexity, right? Perplexity was this huge thing. Everybody was like, "Oh yeah, perplexity is so cool. these days. I don't, you know, I've heard that they're doing okay on the enterprise side, but in terms of consumer demand, like complexity for consumers is like almost non-existent. >> Um, so the reality is that it's a really hard space. Remember when people were so excited about notebook LM and like these like automated podcasts? Can't remember the last time anybody did any of those things.

>> So this stuff moves very quickly. And so that's why I think it's hard to invest in a particular thing until you start to see actual scale. Not just tech demo, not just like, oh, isn't that cool? not just some Twitter influencer being like, "Oh man, I've automated my life using this blah blah blah and then you never hear about it again in three months." Um, so that's that's what I'm looking for. Um, so I'd love to see founders who are building really new stuff. Um, and we're and we're happy to back stuff that is, you know, targeting a really really big market, but in terms of the traction today, the reality is that we're talking about the future. We're not talking about the present.

>> Yeah. I guess last thing for you in the convergence of of blockchain and crypto for the last two years it was like let's do uh decentralized inference let's do uh decentralized compute let's uh you know open up access for anyone with a GPU to uh contribute to to some cloud kind of like this like decentralization of like the stack and and now it's more of like aentic right it's like aentic workflows agentic payments uh agentic commerce uh you know having AI you know interfaces all these uh you know types of like more like products.

>> It kind of like it kind of uh resembles the infrastructure to app shift in crypto as well. >> Yeah. >> How do you how do you think about like where the actual convergence is happening now versus like where it was you know talk to us about two years ago. So okay so two years ago a lot of the excitement in cryptoai was this let's decentralize the infrastructure let's decentralized training let's decentralize inference let's decentralized GPUs and these GPU marketplaces and basically so far none of these seem to be winners right all these seem to be just kind of very low demand there's not a lot of people using these um and I I you know we didn't make any investments in into this any of these categories >> in large part well you know I don't want I don't want to be too premature about

that um under the assumption that like look people in AI like AI is not like crypto in the sense that people in AI are not ideological you know it's like if you have the better model I'm using you and dumping the other guy I used yesterday like you know people protest like oh they're stealing data from the artists they don't give a nobody gives a they're just using the best models people this is such an overwhelmingly persuasive product that the only thing people care about is is it good if it's not good I don't care you could you know you could put you know just dead babies inside of it inside the the the data centers, I'd be like, "Yeah, if it's a better model, great. I'll use it." So, um, so the reality is that like, look, it's easy to be ideological when no

one's using your products. Um, so now that being said, that being said, um, do I believe that nobody's ever going to care about any of this stuff? Actually, no. Okay, so I do think that these things are going to matter, but they don't matter yet. Okay, now why do I say that? Um, there was a story just today about Anthropic discovering that all the big Chinese labs were distilling anthropic models, right? And so this is not surprising to anybody who's been really close to these things, which is that, you know, the Chinese lab has always been surprising how close they are to the frontier despite the data disadvantage that they've had. Right.

>> Was accused of this when they had their breakout moment earlier. >> That's right. That's right. And the same thing is true for you know Quen. Same thing is true for uh Miniaax and all these models, right? Same thing for uh for Kimmy. And you know people would often joke around look if you ask them if you go ask GLM who are you oftentimes they'll say I am Claude. I'm trained by Right. And so it's like a huh I wonder how that got in there. Um, so now they they discovered that there was something like 25,000 accounts that were creating this gigantic farm just to do basically post- training and distillation on anthropic models and anthropic and people this is kind of an open secret.

People knew that this was happening but the scale of this >> was surprising right and now anthropic is uh locking down their APIs and the same thing is going to happen to all the other big labs. Okay. So this is I think uh the first shot across the bow at one this in some level this is bad for the decentralized AI thesis. Why is this bad? It's bad because it means the gap between closed source and open source is going to grow again. >> That's not good, right? For decentralized AI means open source models, right? If open source models are significantly worse than the frontier, that's bad news for you. Now the the good news for you is that it's actually really hard to stop this, right? You can you can definitely stop it at the scale that it's happening, but can you stop it

period? If you want an open consumer product that anybody in the world can use, like it's gonna get in the hands of distillers. Yeah. Right. It's just impossible to stop, right? It's the same thing as like, you know, trying to stop uh non-FDA approved drugs, right? Like you can try, but there's going to be people who just like get some Chinese peptides and inject it into their into their, you know, belly fat or whatever. They're going to find a way to use this stuff if they want to, you know? So, I think on some level it's a cat and mouse game, but it's going to get much higher stakes. Okay. it's going to get much higher stakes. So there's I think going to be increasing sense that this is a national security thing.

>> Yeah. >> That this is basically technology transfer of our most valuable companies going to China or other foreign adversaries >> and that is going to result in these labs increasingly getting locked down >> and having national security level secrecy which today they don't have. >> Today it's very just like ah yeah we trust you go figure it out you know and like kind of yolo it's just like any other company. I think it's going to be more like the space race and that's going to result in a lot more seriousness about locking down these models. And what else is that going to result in? It's going to result in more censorship and more control over what these things can and cannot do and can and cannot say.

>> Yeah. >> Okay. So, I'm drawing the line out as if kind of a multi multihop uh uh argument. So, I think what that's going to mean is that censorship in models is going to grow around the world. The reality is today these models they're kind of censored, but they're not that censored, right? Yeah, you can kind of just do whatever you want. You get one of these models, you want to, you know, just do some weird or you want it to do something that's illegal or unlicensed, you want to role play Spider-Man, you can find a way to do it. Okay, that stuff is going to get more and more locked down over time. And so I think the demand for AI much very similar to money, >> right? Like the reason why Bitcoin exists is because money was ultimately controlled by the state and people

wanted to have unfettered access to decentralized money. Why? because you could not do it in a way that was censorship resistant. >> Okay. But at the time today, nobody's censoring AI. >> Yeah. No, >> you know, they're kind a little bit, but like if you if you want even even a little bit, it's not that hard to get it. >> Okay. That is going to change. It is going to change. This is the most politically important >> technology of the last century. >> It's going to really matter who controls it and how it gets used. And so if and as that happens, I think that's when you are going to see decentralized inference becoming more and more important. That's when you're going to see demand for this stuff start to go through the roof. Not for ideology, >> but for people wanting to do things that

the labs won't allow them to do, that their that their governments won't allow them to do. You're already seeing it in China. >> Okay. In China, like for a for a model to be served in China, there is a checklist that the CCP checks you against of input output pairs that you must match. Yeah. Right? It's like if if I ask you about Tiana and Square, here's the answer you have to give. And if you don't give that, you're not allowed to launch a product. Yeah. Okay. So, this level of censorship already exists in certain countries in the world. It's going to become ubiquitous. The EU is starting to get there. >> It's going to become everywhere. And when that happens, that's when it's going to be the moment >> for, you know, in the same way it took the great financial crisis >> and the the the the the um the Fed

bailing out all the banks in 2008 for Bitcoin to have its moment for people to realize why this thing was necessary. I think the same thing will eventually be true for decentralized AI. >> Yeah. Let's see man please please come on again in a a couple months and let's look at the state of censorship in AI because I think that I mean this is something that near's working on this something that a lot of these uh protocols who are building these products are you know envisioning but it's not here now and it's like I said like we've seen the shift away from that theme. So I'm glad that you're kind of showing us that it's still something that's going to be important because I think it's it's it's rooted in the same ideologies that we have with with crypto. That's right. And with money,

it's about freedom. >> So, thanks, Steve. >> My pleasure. >> Enjoy your meetings. >> Thank you. >> Congrats on the raise again. >> Thank you. >> Cheers, man. Cheers. All >> righty. Oh, Steve.