Why We're Headed Into "deAI Summer," w/ CoinFund's Jake Brukhman

Hello and welcome to the People's AI presented by Vanna. I'm your host, Jeff Wilson, and I am delighted to be joined today by one of the smartest investors in the world of Web 3 AI, Jake Brooklyn. And I am also delighted to partner with Vanna to bring this podcast to you. Vanna's vision is user-owned AI through user-owned data. And the mission is to be the world's first open protocol for data sovereignty. And Jake's mission, probably speaking, is to support and invest smartly and shrewdly in this space. Now, if you've been in the crypto world for a minute, Jake Brooklyn needs no introduction. His firm, Coin Fund, was founded in 2015. It was one of the very first crypto-native investment firms. And now Jake is very focused on the intersection of AI in Web 3. His Twitter handle now even says, "Dai summer." And as he explains in their conversation, he has a theory as to why the next few months will see an explosion of activity. This episode honestly is stuffed with alpha. Jake and I cover a lot of ground in the crypto AI space, from the most promising use cases to his investing thesis, to red flags, and green flags he looks for in investing. Think of it as sitting down with a glass of wine with one of the savviest investors in the space, and picking his brain for what he likes, what he doesn't like. Now, hashtag, not financial advice, obviously, but I think it is super helpful context for perspective for anyone. We also explore an inherent paradox in investing in decentralized AI. That much of his tech is open source and transparent, but typically early-stage investors want to see a moat, or basically a way to make money. Jake kind of explores and reconciles this seeming paradox. So without further ado, please enjoy my conversation with Jake Brooklyn, founder and CEO of Coin Fund. Jake, welcome to the People's AI. Hi, Jeff. Thanks for having me. So let's start big picture. When you look at the intersection of AI and crypto right now, what excites you the most? What's the killer use case that actually makes sense? Yeah, it's a good question. I think a lot of the stuff in the space is really focused on. It's very like zoomed in on various features, like people talk about inference, they talk about GPU aggregation, they talk about agents, but you're asking a good question, which is what is the big picture? The big picture for me is can open source AI produce models that are on the frontier and competitive with what big tech is doing? That is a big question. We have some reason to believe that we can aggregate a lot of compute and maybe even more compute than big tech companies. For example, Bitcoin Network is a really, really good example of that, but there's a big question of can open source compete. A lot of people in the traditional AI world don't think so. Ilya Stutskever has said as much publicly, Sam Oltman has said it. They're obviously a bit biased in their opinion there, of course, but we can think about what their biases might be. My bet is that decentralized AI will create a pipeline for the production of AI models that is fantastically more open. In other words, while most people in decentralized AI are thinking about these particular threats, what I'm thinking about in general is the fact that this is a space that is about moving the state of the art of AI itself, such that AI can run on public decentralized networks. In other words, it's a great public goods. Yeah, I think it's a great macro important view to basically us. I want to get into the tech stack and a bit in which area do you think has the most potential and promise in which areas are maybe a little hypey. Before we go there, I want to start this ground this in East Denver. We were both in the last week at East Denver. We saw just side events, after side events of AI agent events and AI crypto events. This space is oozing with energy, excitement, capital, and hype. I've been trying to get my arms around the entire board of all the projects. It's hard to do. There are hundreds or thousands of these decentralized AI projects. Before we get into particular verticals, there's so much out there. What separates as an investor from your perspective, what separates it would digitalize AI project from one that's writing the hype, whether some things you look at to sniff out green flags versus, okay, this is AI flavored wand. They're waiting. That's a really good question. I think profile and ethos, and they'll explain what I mean, but that's a profile that I tend to look for and we tend to invest in is people who are actually in AI or have AI experience coming to Web 3, realizing, hey, we can take some of these primitives and create products, services, research, advancement, etc. That is competitive with the Web 2 world. I want the customers of these products also to be, you know, not just Web 3, ideally Web 2 customers as well, because then we're going to be showing. They'll eventually have to be, right? If the audience only Web 3 were in big trouble. Absolutely. We could take a look at inference as an example, right? So inference is probably one of the major categories of investment in Web 3 AI. But if you start to think about inference, it's a little bit counterintuitive. If you want to sell inference to Web 2 customers, probably one of the main considerations is going to be the cost, right? There's a lot of inference out there. There's a lot of compute out there. There's a lot of services that are providing, you know, open and proprietary inference of LLMs and other AI models. You know, there's the OpenAI API. There's mid-journey. There's all kinds of APIs out there that are competing with each other in a couple of domains. A lot of it is about cost, right? You want to have the biggest model, the best model for the lowest price. You could see that of the competition between open AI, deep research and GROC, deep search. But then when you go back to the kind of inference that happens on a decentralized network, as some people have proposed, you realize that actually this is not poised to be less expensive in most cases. I mean, it's by design, it has massive structural disadvantage in terms of co-comput and co-location compared to centralized training. So it's like the handicap has to be overcome. The only competitive on price. That's right. So in the handicap is essentially verifiability, privacy, maybe even some other infrastructural considerations that happen in Web 3 networks. That generally speaking, do not make Web 3 inference and decentralized networks cheaper than what you have in a centralized way. Now, there's projects that are trying to overcome that very early. There's people working on basically like sharded inference and commodity hardware. There's people that are looking at statistical methods for verifying the fidelity of what comes out of an LLM. And these are all mechanisms that are trying really to break down the cost of inference that could be competitive. But we're not there yet. So anyway, that's an example of, again, looking for founders who have done pretty serious work in the traditional space and are using Web 3 Proofers to create an advantage. I mean, I think I'll you from near comes to mind is exhibit a someone who was working like and the foundational team at Google that helped birth the tech that launched the transformer in LLM as we know it. So clearly he has impeccable AI mainstream AI chops and now it's new in Web 3 projects. So that's the profile. That's the profile component. So you mentioned ethos. Yes. Thank you. So I was getting there. Yeah. So and by ethos, what I mean is kind of like the ethos of problem solving that's happening in this in this proposed company or project, right? I want to understand what are the core very difficult technical problems at the heart of this sort of open pipeline of getting AI onto public decentralized networks. What are the really hard problems that we have to solve? And there's there's problems, you know, inference in and of itself is not a hard problem. It's hey, we set up some models and we set up a service to give their outputs. But in the course of solving something like, you know, shard inference on commodity hardware, you might be working with more difficult algorithms or you might be like advancing some distributed computing scheme. But where this really, the area in which this really is illustrated very well for me is in the area of decentralized training. Right? I've been a pretty early investor. Yeah. The centralized training is something that seems very, very, very hard to do on a decentralized network. And the reason is because, you know, there's a number of limitations that come into play. There's latency being a big one if I'm not mistaken. Latency is yes the fashionable one to sort of talk about. This refers to the latency between the nodes in a decentralized network they're trying to you know fit this neural network model. But there's many other problems too. There's there's Byzantine resistance, there's fault tolerance, there's the question of and I'll just say so so now it's we're in March of 2025. One of our portfolio companies Primate Select has demonstrated a number of models that have been trained on open decentralized networks today. So now we know that at least you know to some degree this is possible. A lot of people have you know in the past have said this is not a problem worth working on. It's not really possible because of those latencies and so on. But it turns out that actually if you sit down and think about the properties of a training process that happens in such a network and you ask yourself like what are these properties and how well research show these properties. What I actually started to see is that there's a lot of greenfield research that is available and low hanging actually can enable this and you know Primate Select is one of the ways that you could do it using data parallel approaches. You know our company Pluralis is now in the process of doing more advanced approaches that can scale to create even you know larger larger models. We talked about green flags and I love that ethos and profile framework. Let's flip a tables talk about red flags. I'm guessing there's a I mean as a journalist Jake I am getting shit loads of pitches left and right and it is just so hard to sniff out what's substantive and what is Johnny cum lately hype train. So I have my I'm working on my own kind of red flag detector. What's yours? What how do you what are some obvious red flags and like what kind of makes your your bullshit meter really you know tick tick upwards. Well I mean I don't know if even that's necessarily the framing that I would go with of some projects to reel and some projects to bullshit. I just think that like some projects are have higher probabilities right and other another projects might be kidding themselves about having a market or just like sure. I don't mean to imply there's it's just 90% bad actors trying to scam that that's not what I'm saying. I guess by it's more okay this is this is projects 417 of the projects all trying to do a similar thing and they have like the weakest team and the least legs of standup on so their tensions might be good but this looks like a pretty I think shaky bet right. So yeah so red flags would be something like solving a very you know high competition kind of area or not very defensible problem you know and thinking that you could still succeed but not actually having you know unfair advantages so maybe it maybe let's make that concrete a little bit. So if you're you know if you are doing a startup for let's say like data data labeling right you have to be really cognizant of where this data is going to flow into it like for data labeling the majority customers of that kind of data today is probably like big deck it's probably web 2 people who are trading larger models and require like specialized data cleaning services we have a scale.ai which is you know multi-billion dollar very successful company you know serving this this niche in web 2 you know so if you're making the proposition that okay we're going to have data labeling but in web 3 then we have to ask the question like who's going to be the customer of that. You know now I just talked a lot about decentralized training right so in the future there could be a market where models are traded decentralized way and decentralized networks are demanding this data but that's not the state of the market today and so you know between a team today in this area where one team can pick up the phone and call Google pick up the phone call Facebook pick up the phone and call like web 2 customers versus a team who's sitting in web 3 and sort of waiting for that market to materialize and mature you know I would highly prefer like the former and so that would be a flag for me you know. Next sense back to the the crux of this what seems at least on the surface like a paradox we alluded to earlier that so many projects are building open open permissionless AI models and classically we see these like defensibility in a mode you mentioned there are a couple of examples of your portfolio companies that are starting to crack the code but you can help me understand a little more about how that's possible you know how is there confidence that a company will eventually have a proper profit model that delivers return for investors in this kind of very open ecosystem that's a great question so let's let's start with the models that we sort of see in the market you have very proprietary models those would be you know open AI, a thropic, croc etc you know users pay for these models and then they get access and their access will be like cut off if they stop paying and also you know their data the data that those three systems is now owned by these companies. Classic world garden model. Classic Web 2 yeah let's say model and that is great but one of the risks in that space is you see like there's like a very vicious race for the frontier model for like huge funding for data centers some people are you know starting nuclear reactors to feed their data centers and so on and so forth then you go into the Web 2 model but now that's focused on open models and that would be a company for example stability AI or maybe maybe mistroll as well and what you see there is that you or actually meta meta the long enough series. Yeah easy, even long enough to fit in that lane I would think right yeah. Now having these models has been absolutely fantastic this is open source it allows developers and users to play with the models it allows there's been a whole lot of innovation where people have modified these models to learn how they worked now you can run kind of image generation on your on your MacBook Pro there's even been training frameworks and fine tuning frameworks that have been developed because you know these things have been open what do we mean by open there we mean that you know Mark Zuckerberg trains a big Lama model and he opens sources the weights of the model this is a giant matrix it's a bunch of numbers and we call that open source now what's definitely true is that Mark Zuckerberg also goes and podcasts and says hey you know if it ever became expedient for us to monetize a close proprietary model we would absolutely not open source at that point you know so this open source is it's very unsustainable it's very provisional and what it also doesn't do is it doesn't open some very important issues about model making which is well what is the data is this thing has been trained what is the hardware setup what is the architecture of the model and so on and so forth so it becomes an interesting thought that if you created a network which is an open network where you could create open models but they were open in the sense of their setup versus in the sense of their weights then I guess like what I'm saying is closed like models with closed weights and open setup are going to be more open than models with open weights and closed setup like we have from meta but it isn't much of the web 3 decentralized AI ethos to open all of it to have the transparency on both the weights and the setup and and I'm still fun and I take your point that the long bucket if you will of like open ish AI that's not fully open I take your point that's not that open really but I'm still struggling to see how the more open and more decentralized and transparent web 3 AI solutions to put it overly simplistically how they make money where where where where is an investor where does the where's the profit coming to the play eventually yeah so so in this model that I'm proposing where you have open setup but closed weights we have a business model because if you if you can have the term the term of art is non-extractability of weights if you could have non-extractability of weights then these models have a business model in inference like some community gets together they form capital they train this model this model is particularly good at a certain thing or maybe it's the best LLM in the world and suddenly people have demand for this model well now they have to go back to this open network the trained it in order to do the inference in order to receive the outputs of this model and they're willing to pay for that just like they're willing to pay for the open AI API and an anthropic and grok and so on and so forth however what this system maintains is the same property that you have on GitHub I can go and one click fork a GitHub repository and it's free sure and the same thing is I can go go in fork, the model setup presumably in such a network, but then I have to get my own resources to train the model, right? So it's the same thing as, you know, if I fork a popular project on GitHub, I own the code now, but I don't have the intent to do it. - And there are plenty of examples, even the web too, where open source tech is a big chunk of what a project is doing, but there's additional layers on top of that that have more proprietary savvy that can be monetized. - So this setup really solves two key problems. It solves the sustainability of open source, historically open source is not very sustainable, it doesn't have good business models. It relies on donations, it's, you know, it's not as, people who work at open source are not necessarily competitively compensated, but it also solves the totally closed and proprietary and almost like dangerous nature of models that are trained in the secrecy, you know, of big companies. So, you know, I think it's a incredibly interesting idea. - Love it, and that's really helpful, thank you. Let's get a little more specific now on the different layers of the stack here and what excites you most. And we talked a bit about this off mic way back when in Bangkok at DevCon, kind of like what, what fires you up and what do you think most folks are undervaluing between the, you know, starting from the most, like whether it's the, we're gonna start with the data layer or the training layer all the way through apps and agents. What right now do you think has the most potential and what excites you the most? - Well, I would say as a general matter, right? I think we very clearly see that in Web 3, the topic of AI has become a huge topic and it takes up like a ton of mindshare. I mean, we had the top crypto influencer this year for many weeks be an AI agent, right? That, which is I'm referring to AIXBT, which has meant to surface like crypto alpha on Twitter or X. So, it's safe to say that AI is a hot topic. - I mean, Jake, if you look, every crypto conference has become a default or a de facto crypto AI conference. And there's so many more events that have an AI changed to them at the very least. - Absolutely, absolutely. And the other thing I'll say about that is, you know, if you look at it from the perspective of a normal person who wants to, who's interested in AI, wants to get financial exposure, wants to invest in some frontier AI projects, well, you notice that, you know, there's not really a ton of stocks on the stock market. In fact, any that give you access to the most up-to-date frontier AI, not a public company and Thropic is not a public company. - And the video isn't exactly cheap these days. The video is not cheap and I guess you could buy Google but then you're buying exposure to, you know, all the other businesses that they have there, not necessarily just AI. And are they, you know, frontier competitive anyway? And maybe that's up for debate. So, no public companies, not really, you know, most people are not gonna access the private rounds of opening AI and now there are hundreds of billions. So how do people, if they wanna invest in space, how do they do it? Well, it turns out the lowest friction as the class is, crypto. - That's an amazing, I hadn't thought about normies coming in this niche for undiluted, investing access to AI. - Yeah. - And token Web 3, Arena, that's interesting. - Absolutely and I think that's why we saw some of this like AI meme coin exuberance. - Yeah. - You know, earlier this year and late last year, unfortunately, this is because there's all this pent up demand for like frontier AI exposure. But unfortunately, like, I think most of those tokens didn't really give exposure to the real Web 3 innovation that's going on. And this is why I have D-A-I summer 2025 in my Twitter handle. - Shall that love it? - Big thanks. Because, you know, I really think that 2025 is gonna be the year when people in traditional AI will start to realize that a lot of the Web 3 AI, like research and problem solving is very, very applicable to AI broadly. And they're gonna get really interested in it. There's gonna be papers published that I see them out. There's gonna be convergence. And there's gonna be a lot of people buying crypto AI tokens. - Let's go there now. - I do wanna get put a bookmark in the difference kind of layers the stack that fire you up. But one question I have a meeting to ask you, and I'm really, I'm curious about is why we don't see more folks in the traditional Slash Centralized AI worlds exploring these decentralized Web 3 solutions. And just maybe to step one little back, you know, I've written plenty about this. I've written for coin ads going to articles on crypto AI intersection, you know, starting from a couple of years ago, and it was still just a, just an early early idea for most folks. And the rallying cries often been, well, we all know crypto needs AI and AI needs crypto. It was your two technologies that fit perfectly, right? So I love that idea. I am wondering though, if that is true, why, can you get me asking? I talk to folks in the more norm core AI world all the time, a very small percentage of them, at least in my experience, are saying, oh yeah, we need to be incorporating blockchain solutions. We need token frameworks. We need to embrace the power of Web 3. That's a pretty tiny pie slice of folks in the larger AI world. I think that why, why didn't it, why aren't there more, how do you reconcile that? I'll give you my take on it for sure. I think the reason that you're seeing that is because there's really what's happening here is a convergence of two different communities. It's a cultural issue. So you have one community, and of course, I'm referring to the AI industry and research world, right? And kind of commercial world. This AI over here has been developing for 80 years or something. And these are some of the most traditional academic people who are coming from extremely centralized environments, like they all come out of universities and big tech companies because historically this is where this research has been done. Peticry is so, so important. What lab you work that out of college is an incredible predictor of future success of the AI, right? Just the various centralized sort of traditionalist mindset. And now, in a phase where they're making tremendous progress, we've invented the transformer, we see scaling, the scaling hypothesis sort of like playing out, they're like extremely focused on, you know, figuring out how to win the AI race, figuring out how to get to AGI. What they're not thinking about is what to them seems like really auxiliary problems with a lot of headaches, by the way. That would occur if you were to try to start doing this in Web 3. This seems totally sub-profless. And on the other hand, you have, you know, the Web 3 community, this technology has been around for 10 to 15 years. It's constantly trying to apply itself to different areas, you know, banking, digital collectibles, now AI. And it's always coming with this like decentralization requirement that's making it kind of difficult to navigate or pick like technical. And again, you know, now all that being said is that there's also not zero people on the intersection of Web 3 who are very credible. Like we always talk about Ilya being one of the original authors. We talk about like a mod who was a crypto founder now, then before starting Stability AI, and now working on other projects, you know, and so on and so forth. But I think, again, what I'd like anyone from those communities to take away from this podcast is that that convergence is happening in a big way. And the overlap of problems is actually very, very large. And that is going to become obvious this year, as I think the centralized training algorithms start to produce larger and larger and larger models. And eventually, I think, compete with what the big tech guys are doing. I totally agree, by the way. Let's go back though to the stack. Do you want to get a little more clarification on what of, you know, when you look at this from compute storage, model training, data marketplaces, AI agents, you know, what, what, what are some corners or niches that you think that excites you and or are not enough folks are talking about people who are sleeping on? Well, I mean, there's a lot of talk about AI agents. If you saw, they're every, every talk is everywhere. Yes. I mean, if you look at the state of the art, such things, you know, if you play around with operator or, you know, deep research, like what is obvious is that these things are very early, right? It's, you know, you can't actually set operator to go and do something for you and not expect to hand hold it a lot, not expect it to get, you know, into some kind of loop or make a mistake. I know I was doing some tax work and it started, I gave it access to my Google sheets and it started opening up like other. Sir, you're, you're a courageous man. Yeah. Yeah, agent, do my taxes. That's a low, low stakes. Let's start with something low stakes. my taxes. Exactly. Well, exactly. I mean, I mean, if you're going to have agents that are your assistant, like they better be capable of doing the actual work, that is the kind of thing that down the road. If this all plays out that way, you want, you're right. That should be in scope. Yeah. So I could tell you, like, personal experience, you know, we're just, we're just not there. And so I think, I think a lot of the web 3 AI agent stuff is very speculative. And what you see is that it's classified into a few different areas. We've seen a lot of AI influencer agents. That's pretty cool. But generally has kind of a crappy business model. You know, they tend to have a meme coin. A lot of meme coins get a zero and so on and so forth. Then we've seen, I mentioned a XBT. We've seen more like domain expert agents agents with a terminal, you know, like a XBT knows about all kinds of crypto sentiment from Twitter. You can ask questions about projects. It could even, you know, pull up some data in as far as that data has been sort of highlighted on the platform. And that's useful. It has a better business model, potentially with, you know, subscriptions or, you know, the way that AXBT is today, you have to hold a certain amount of token to use the terminal. But that's also not like the most exciting business model. And I feel like it's also not the most sustainable one because obviously agents who are ingesting more and better and cleaner data, they're going to like outperform AI XBT if AI XBT isn't, isn't catching up to those things. So there's always this like race and domain experts to be like the best domain expert and to maintain like the value, you know, of the subscription. But I'll tell you the third category, which I'm most excited about, which is on chain useful kind of defy agents. So one example of this is we have a portfolio company called Giza. It's a tech. It's YZ and they've been launching or they just launched ARMA, which is an agent that helps you optimize your yield across different, you know, vaults that are available. And you know, in the future, you could see these agents doing useful things, rolling your perps, you know, making some kind of like risk assessment on financial transactions that you're doing and so on. And I think what's really interesting about the useful agent, there's a few things about Giza that's really interesting. First of all, they're completely zero knowledge verified. Like if your agent is handling your money, you definitely want some security guarantees. You want some correctness guarantees. You want to make sure you're, you know, inferencing the right model. You want to make sure that the inference has been done correctly. You can't just trust people on a Web 3 open network. Sure. You know, naively with that task because you will, you will lose money, right? And so you need, you need to be strong, stronger security guarantees. And the other thing that's really interesting, the business model rocks because the business model is the entire volume of D5 volume, right? And so these agents can take a small key, convenience fee, you know, and running through, you know, some of these operations, like for example, if it's optimizing your yield, it might be taking a little bit of a carry on that yield. But overall, you're getting a great result and a better yield that you would be getting by your side. Totally aligned with that and being bullish on that overall concept of using agents in this kind of D5 land because what's some of the biggest knocks about D5 and smart contracts more broadly is it can be super technical, super manual and mechanical to do all these things. And but, but AI's agents need a lot of, you can train them and they don't care if things are complicated. They don't care. It takes 37 steps to do these things. They can do that in 0.001 seconds, right? So the idea of having AI and thinking a little loosey goosey here, but orchestrate, elegantly orchestrate this world of smart contracts makes total sense to me. And I think that's one, one area where there is a massive advantage in combining the best qualities of Web 3 with what AI's really good at as well. So, yeah, totally great. And look, I think like for me, the, I like that useful agent framework more than all the other AI agents that have been proposed. To me, like an agent is someone who is doing things on my behalf, like really doing them and also doing them in parallel to me, which include increases my productivity. And again, I don't think we're there yet. I don't think most agents are capable, I think they're capable of like niche things right now. Yeah. Hey, give me some data or run a report on real estate in Miami. You know, maybe optimize my yield in DeFi, you know, but what you really want is sort of a generalist assistant, you know, that does whatever you need it to do. And you can trust to do it as well. I'm going to, I'm going to, I just made a very bullish comments. I'm going to now be annoying and make a skeptical comment or such question. I would love to thought sign while I completely see the merit and in use case of agents talking to other agents and you have this kind of blossoming agents to agents economy. The idea of having these constant like high volume micro transactions and using some kind of cryptocurrency, God knows what it will be, but some kind of some kind of token solution in the DeFi universe makes sense. Got it. Where I, where I, what I am a little skeptical about is when this topic first gained a lot of excitement a year, year and a half ago, two years ago, the kind of almost people tended to state almost as a bedrock principle or statement of facts. Well, clearly AI agents will be working on your behalf to spend money. Like maybe it'll book you a flight you'll say, Hey, by me, by me ticket, I'm going to go to Tokyo on these dates, find me a the best flight book me the hotel, do all the work and then execute it on my behalf, make this happen. And the argument or the thought is given AI agents are digitally native, it makes sense they would use the digital native currency, which is cryptocurrency. That's just logical. And almost like a like a like a a priori idea. My question, long wind way of saying my question is in the DeFi world, yes. But in the more web to quote legacy world, but which is frankly still most of the world, if I'm using my agents to buy a tight ticket to Tokyo, is United Airlines when they're setting up their infrastructure to deal with all this, are they more likely to be play ball with the agents that is using some kind of cryptocurrency or an AI agent who is just using a credit card, right? I guess the similar question is, why can't AI agents when interacting with like traditional companies just use credit cards and you sync it up to your banking accounts and whatever all that stuff. Like I'm not sure I see why necessarily, tradify has to be ebbed out if AI agents scale up. I think AI agents will need to support credit cards realistically to interact with things in real world. And also I also think that you know crypto is the most sensible kind of substrate for agente commerce. We should be encouraging agents to to move money this way versus like a bank or something. I can't even imagine the complexity of setting up an account for an agent. But here's where Jeff, I feel pretty confident is, you know, we really like, if you zoom out from agents a little bit onto the crypto world, you know, we're really like in a period of convergence again between tradify and defy. I mean, this is like one of the hottest topics. You're seeing stable coin legislation, the horizon, you're seeing banks starting to adopt or thinking about starting to adopt public blockchains and stable coin payments. And you're seeing a lot of neo banks in the market sort of saying, look, we can have a combination of, you know, traditional banking compliance and traditional wire services. But look, if you want to use stable coin payments, if you want to use like defy for borrow lending, you know how these new primitives. And that bringing defy to market is really exciting. And also, you know, I think accounts like that could be used with agents to cover the entire gamut of payment use cases that agents might run into. That's a great point. There are some broader tailwinds for space at large that will hopefully remove some of that friction, which leads me to my next question about regulation. Obviously, we use stable coin regulation. I feel like the regulatory bodies, or I don't just guesses from the cheap teeth here are so far behind grappling with these, these issues, right? They're just, they're just starting at the arms around finally like stable coins. And now we're talking about this potential billions of AI agents constantly flinging cryptocurrency tokens back and forth. Like I don't think your average congressperson, that's on their radar at all, right? But do you think that where's your head at as far as if and when regulation will, will quote, enter the chat when it comes to crypto plus AI and how do you think that will shake out when it does? So I'm going to go and disagree with some of your assumptions. Please. I actually think that Congress has been getting like very rapidly up to speed over the last three years on frontier technologies like crypto and probably more recently AI as well. And what is so interesting about like these two pieces of legislation that are crypto later that everyone thinks we'll be very likely to happen in the next two-year time frame is that they're all like bipartisan bills, right? They like it didn't. As they should be, they should not be appointed to be. As they should be. Yeah, as they should be. And what I guess what I'm saying is like it wouldn't have mattered who won the election to the extent that I still think these these bills would would be likely as likely to pass eventually. Yeah. So so there's that. That's alright and your broader question was just kind of like what's the regulatory? Yeah well I do I guess my I guess the subject behind it is are we proceeding? Are we excited about this space? Yeah. In a world that without really regulation, looming on the horizon, but in two or three years, in my simplistic view of things, which maybe is misguided, if Congress wakes up, holy shit, like what are you guys talking about billions of agents doing all this? Is that going to cast a shadow in the same way three to five years ago that Web 3 landscape was like oh, it's a big step back. I think you're pointing to you know a an issue that always happens with technology, which is it runs ahead of government and regulators and what's interesting living through our time is that you know we're on this exponential curve of innovation where this technology is moving just exceptionally fast. If you think about how AI models come to market and then three days later, there's a there's a new a better one and five days later there's the new a better one. This is literally the time scales of what we're dealing with so what's going to happen when you know there's a massive step change breakthrough that happens like on this very short time frame and changes everything. So that's definitely a challenge. On the other hand like looking at crypto, I mean we went from in 10 years right we went from just absolute zero on on regs to an administration that's touting crypto our first crypto president if you will. Extremely favorable regulators in the SEC, CFTC, Treasury, you know other regulators. You have two piece of legislation as we keep saying like Lumen in the horizon. You have a crypto task force at the SEC you have a crypto czar in the White House. You have the president literally launching tokens for better or worse and you know like it's really the best sort of regulatory environment that crypto has ever seen and what this is actually going to enable to me anyway is just more institutional enterprise adoption and that's the leg of adoption that hasn't been able to proceed in blockchain for so many years because of lack of regulatory clarity and I think we need consumer adoption and enterprise adoption to sort of play off each other to really bring blockchains mainstream. So I'm really here for that. Love it. Last couple of questions as we wind down. First I know that many listeners of his pod guessing are building in the space or founders are working on projects and crypto plus AI or decentralized AI whatever you want to call it. What advice do you have for them? What advice to teams that are that are that are maybe looking for capital or wondering when to get capital. What do you what's your kind of words of wisdom do you have? You know again I you know I mentioned some of the profiles we're looking for you should have great team composition you should have experience in your field you know I am one of the wisdoms of I guess being like an early stage investor over time is that you see that a lot of companies are really killed by timing like some companies are too early other companies launch it a wrong time and then fall into a bear market that's too long and are unable to raise and go out of business right so you know one of the most important considerations and some of the most successful companies that are in our industry you know have really gotten there by making very very good use of timing so if you're you know like squarely in you know Web 3 inference or Web 3 AI agents you should be asking like how do you see this timing playing out how do you see Web 2 competition in agents let's say interacting with you what are going to be the main issues that are going to make one agent product more successful than another agent product for example you know Google and Facebook and Amazon already have all your zero-party data so if they're making useful agents those agents are already going to position to be a little bit more personalized than something that you get in you know in Web 3 so look we look for the top founders who are like we are thesis and trend driven so we're looking for them to fit into the trends that we're seeing with what to support the best teams there and every once in a while a team comes along who changes how we think about those trends how we think about our assumptions right and that that's that happens more rarely but it's also more rewarding love that it's great advice final question for you give us a couple predictions a little prediction on two different time horizons one on the end of this was it so hot what was the summer your your hashtag your popularizing the hot deep D.A.I summit D.A.I. it five D.A.I summer 2025 yeah okay great so at the end of I mean at the so we finished summer we're in fall 2025 or winter what's what's different give us you know what's changed as a result of this this kind of fruitful summer and then second time horizon for you let's go wild five years out what's the D.A.I yeah I know that's absurd but you know it has to come with it no one's in a whole new this five years from now what's what do you think the what a prediction or two from that that time horizon in the D.A.I. world yeah absolutely well I think so the short term prediction for me would be again I think A.I. will do like really really really well this year and of course I mean web three A.I. I think we're gonna see a lot of papers published that I see them all we're seeing companies like Primantilect and Jensen starting to launch products we're gonna see Frontier Research in AI coming out of web three I think we're gonna we still see a lot of interest in AI kind of exposure through tokens and of course there's all the companies that are doing these other strategies around GPU aggregation inference data all kinds of verticals in AI web three so I think like on a 12 month time frame regardless of whatever else happens in crypto I just think like the web three I world will will see like a nice market there and then on a five year time frame I mean this is this is like a million years in blockchain time but you know that's why it's fun but you know I like predictions that say you know based on our past experiences with things changing quickly I think we're gonna see some assets and let's call it the top 10 that are very very surprising and and and yeah I remember like in in 2017 at the token summit my original co-founder at CoinFond Alex Bulken was asked to make a bold prediction and he said you know in in like five to seven years time you know the top 10 will be completely different and he wasn't right but he was right like seven out of ten yeah I'm not really to expect unexpected is a good note I think to end on Jake thank you so much man really really enjoy this you're always a a a a voice that I love listening to to bring some clarity to very complicated space thanks take your time best of luck to you and if I appreciate it awesome I really appreciate that Jeff thank you for having me well there you have it thanks again to Jake Bruckman founder and CEO of CoinFund and thank you to my partner Vana when you look at Jake's earlier investment test of ethos and profile I personally think Vana passes the flying colors given the very real problem trying to tackle a vision of user owned AI through user owned data and a mission of being the world's first open protocol for data sovereignty and thank you finally to you dear listener the people's AI is brand new we are just launching so please help us spread the word subscribe on the usual podcast platforms including YouTube post on social all that good stuff thanks again and see you next time you