AI, Crypto, and the Rise of Autonomous Capital - TOKEN2049 Singapore 2025

Thank you all so much for coming this morning. We are super lucky to have this fantastic lineup. Tun will also be joining us a little bit later. I this morning we're going to be talking about AI crypto and the rise of autonomous capital and some old friends and new friends here. Just to introduce our lineup here.

Alex Pac who is the founder and managing partner of hack VC also one of the founding partners of Dragonfly. We have Jordi who is the founder and CIO of Seleni which is one of the most active quant firms and LP firms in the space. And we also have Amjad who is the CEO and founder of Replet which uh is one of the fastest growing AI coding platforms with more than 30 million developers on the platform. I've actually used it. I over the summer I was hacking around building some applications.

Maybe some of you guys in the audience have used it as well. Uh my name is Irene. I started in the space after dropping out of college joining layer zero where I was head of strategy. afterwards was a venture partner at Bang Capital Crypto and I'm now an angel investor investing across traditional AI in companies like Riplet and I also DeFi. So to jump right in I want to ask you guys to actually define what autonomous capital is.

So um you I think it's the same as AI itself where there's just different levels that you can start imagining from the past into the future. Obviously autonomous capital it could just mean having a strategy that just trades for you and you just you know put money that's been happening for many years in quant trading. Um, at some point you start getting into the fully autonomous capital where the AI just thinks for itself, deploys capital, um, potentially is paying people. So like we're kind of being incentivized by the AI, and that's probably a world that's that's coming soon.

What are the minimum primitives that make that term meaningful? If you want to zoom in on that a little bit more. And Alex, feel free to weigh in here since I know that you have invested across the whole spectrum of different crypto and AI project. Yeah, I think the uh you know even like 10 years ago when I was when I was getting into crypto the sort of fantasy use case was that one day we would have like Uber or Grab running completely onchain and there would just be cars that would come pick you up and you would pay with your you know stablecoin wallet or something and the cars would be owned by some like shadowy DAO that's like collective ownership and it would optimize ize things on its own and it would figure out pricing curves and uh and s and uh you know price charges and uh and the entire company like would be run by this like DAO that is basically managed by an AI um and we're still a little little ways away from that but I think the concept of autonomous capital is like gets us closer towards that right so uh but there's a lot to do like there's Like in the last 10 years, we needed like stable coins, we needed wallets. Uh the agents have to improve significantly.

Uh there's an idea that you can just sort of trust an AI agent to like go make money for you or run your business. And uh that catastrophically will fail if you did it today. Like if you ran the AI for more than an hour or two uh on a on an important task, it will it will mess up uh usually or it will go off the rails. Um, so there's more on like the improving intelligence of models and durability. Uh, and then I think there's the whole layer of like how do you connect uh your your agents to like commerce and payment rails.

Although the stuff that like Cloudflare and Coinbase and Ethereum are doing now is is is a step towards it I think.

Yeah. I mean autonomous capital you know the ultimate form is probably not stable coins unless they're maybe you know algorithmic stable coins. I think right now having USDC or USDT theoretically like these types of things that can get whitelisted blacklisted it's like pulling the plug if you really need to as an emergency and I think it's important that we have those rails for now but it's inevitable that crypto just brings this like new layer where you know if someone's sending you know ETH or Bitcoin or something decentralized you can't really stop it and if it's doing activity that it just decides to do you know I think some of the um you you do a lot of AI AI investments and I'm sure that a lot of people care about safety. Obviously, autonomous capital adds a lot of it's like the autoban in Germany like you get a lot of benefit by going fast but you got to be really really careful. So that's one layer and then of course you know what's actually happening with the execution environment.

Crypto's been experimenting with trusted execution environments and

um those those components are quite mature now. So increasingly we will start seeing those pieces come together. But um I fully agree with Alex that just giving you know an AI money and just telling it to go off and and multiply it for you that's a little bit of a pipe dream because that's not how markets work. Like there's always efficient markets and if something is that easy it'll just get competed down.

I think it's really important to understand what's actually novel with autonomous capital and what has existed previously. So preai in tradfi we we've had high frequency trading firms we've had quant strategies and automated marketmaking in in crypto as well as in tradi and then preai in crypto we've had trading bots we've had uh me um and now with AI onchain in crypto we have agents which can trade without human intervention agents which operate wallets on chain chain and can interface with multiple protocols at once and that introduces a whole different surface area of risk. Um, this is actually an area that I think I definitely want to press on with Tun when he gets here. He's certainly the expert. But um Amjad something that I'm really curious about uh with with your unique vantage point

you engage deeply with AI devs and I want to understand what you think the maybe conceptual and cultural differences are from AI developers and crypto developers their point of view on AI agents and from your time here. I know you're also going to the network state conference. What you think the differences are? What are the curiosities as well?

Yeah. Well, just to give you a bit of backstory about how I I came to sort of thinking about crypto and how it relates to AI and software. Uh, anyone here read the book the sovereign individual? Ah, disappointed. It's not that money.

Well, it should be. It's it's like the crypto cannon.

Do you want to give the 101?

Yeah. So basically in the late 80s early 90s uh this two two kind of British uh intellectuals uh talked about how computers change uh the landscape of politics the landscape of of sovereignty and how the individual is going to be a lot more empowered in the future. And they go into specific details. It's one of the books that actually predicted Bitcoin and the rise of crypto, the idea of digital assets. But the really cool thing there that really captured my imagination is the idea of being able to spin up companies uh on the fly and be able to coordinate people and agents and be able to coordinate large infrastructure and software all seamlessly.

Right now, if you think about making a piece of software, think about how many services you'd have to sign up for. If you want a piece of database, you go somewhere. I mean, at Replet, we're trying to centralize everything, but in reality, we don't have to do that if the market was uh a lot more autonomous. So, if I'm prompting an agent to go build a piece of software for me, the agent should have capital, should have a wallet to be able to pro go a database for me, go to, you know, anthropic, get an API key for me, uh go hire someone for me. If the agent is getting stuck, it should be able to go hire a human or go hire another agent.

If I want to do a security audit, it should be able to go get a security agent from the market and be able to hire it. Now, if you imagine that kind of future, you could sort of start to imagine this this idea of a a billion dollar single person startup which everyone in AI talks about. So you know to the mindset of people in AI everyone knows that this technology is is going to create amazing amount of entrepreneurship new wealth creation new startups it's going to make companies leaner it's going to make it so that um there's a lot more freedom in terms of what how you want to work and what you want to work but crucially there needs to be a very uh low friction uh commerce primitives right uh and chatfi has just not been able to get there. And I think that's been the opportunity for crypto since I started looking on it. Programmable money, that's that's really the the promise.

Honestly, it's been disappointing. And my challenge for you guys today, cuz I'm like not from the crypto world, per se. I don't really live and breathe in it, but why hasn't it happened? Where are the primitives that allow a company like Replet to be able to have our agents uh h have like a seamless wallet be able to do trusted execution, be able to have all the security primitives and the uh underlying controls and the audits and all of that stuff. I I don't think it, you know, there's like base and other things that are coming online now, but really now is the time to for you guys to build the primitives for us so we can go build the future of work essentially.

Can I ask what do you think the maybe cultural gaps are that lead some of the developers you work with to be hesitant to experiment with some of these new primitives? Because I know for example that Jordy, Alex, I I certainly have seen pitch decks maybe not ones that I ended up engaging with, but founders in the crypto space who are trying to create these primitives

building these sorts of, you know, one-click wallet seamless onboarding types. Are there cultural gaps that you see or a different understanding of autonomy than the way that people encrypted about it? uh like the understanding of autonomy in AI is about action and execution. Autonomy in crypto is about custody, trust and things like that. And I think crypto from the outside for people tends to be a little bit inscrutable and really hard to understand because the the the the language and the jargon is so heavy and often times people have this feeling that you know there's so much scam.

I don't know what's real and what's not. I think I think there's a there's a place for a developer first company that's coming in and simplifying everything, creating these really clean interfaces and going out and showing uh showing AI developers what's what's really possible. Think about also the the set of incentives and software today. You know, I came up in open source software and one of the things that I realized almost immediately by building a piece of infrastructure that really big companies were dependent on that I I I just can't make a living as an open source developer, right? like, you know, I could get GitHub stars, but I can't eat GitHub stars.

You know, you need you need money. And I think u incentives is another thing that the crypto community understands that I I don't think the AI people really think about all that much. We think about, you know, all the nerdy stuff of how to make agents run for seven hours, 30 hours, whatever. But um how how do these incentives work? And so why isn't there a version of GitHub where open source developers uh have the underlying commerce uh infrastructure to be able to monetize that work easily?

You can imagine a package manager that uh that allows you to you know pay for packages on demand. We just saw npm had a huge uh supply chain attack where someone took over a GitHub package um and every time you install that npm package it will create a warm that takes over your other packages. Huge issue. Now the problem is that because you know there's no friction there's no money transfer it's all kind of you know you can have bots and attacks and all sorts of things like that. The thing that crypto really understands is is that also money can be a form of protection.

So if if you add a transaction cost uh not only creates the right incentives but also protects against the wrong incentives. There's so much opportunity. So I think it just like new companies that are really focused on those.

That's what I thought that you know Elon was doing with Twitter. you just added that cost uh to get the check marks and then that I thought that would stop the spam and then all these spammers just bought golden check marks for $1,000 and still managed to make that business model work. So unfortunately like even adding some some cost doesn't fully eliminate the problem but it does reduce a lot.

Well there are so many crypto primitives that I think could be could be good like like staking and and challenging and I I don't think it's been fully explored and um again I don't think that traditional finance infrastructure has the right primitives.

This is a fantastic uh opportunity for me to welcome and introduce Tun who just arrived. I I've had the pleasure of working with Trun in different capacities. Trun is the founder of Gauntlet ERA and also co-founder of robot ventures and on this panel certainly the expert of risk. Uh we've defined autonomous capital. I think one thing that would be very interesting to talk about is risk in autonomous capital.

So autonomous strategies compound governance model and operational risk. I'd love to hear from you how exactly these three risk factors compound in that way.

Yeah, I think one interesting thing about agents is how do you attribute risk versus like humans where there is like kind of a natural attribution? This is like kind of the thing people always talk about with insurance and AI, right? Like I have a self-driving car. Do I attribute an accident to the software? Do I attribute the accident to the LAR breaking?

Do I attribute the accident to the open source package that was imported that happened to have the bug that caused the crash, right? Like there's like many different layers and it's not clear any single one has full attribution. I think agents have the same thing for finance. Now one kind of interesting thing is in the last two days actually there have been a bunch of exploits against trust execution enclaves which a lot of agents that are interacting with crypto right now tend to use mainly for computational capacity constraints uh as well as the fact that you having to customize your model to the cryptography you're using is just too expensive and TE's allow you to kind of have a a stop gap but there are some really dangerous zero days there right so what happens S if an agent is running on a decentralized network and one of the node operators who happened to run the enclave that your agent's code was run in didn't upgrade their software. They have the exploitable TE.

Do I blame the node operator or do I blame the network? So I think like there's this attribution problem that is kind of fundamentally different in the agent world because I don't actually know whether I blame the agent software writer who might not have any idea what hardware it was run on. So they don't know that the TE that was used was exploitable, right? But they did choose to use a TE as their form of execution. So I I I guess the the main thing I'd point out is like humans are easy to blame, right?

I can like say you up like you made a choice that was bad or you made or governance voted on something that was like bad

and there's sort of some notion of attribution. And I think the agent world actually is much harder. And the more complicated exploits get, the harder you're going to have a time of attributing risk. Um, so may maybe that's not really an answer as much as just a like why it's a more complicated question I think than people give it credit for. Especially in a world where like there's constant zero days against hardware, software, supply chain attacks that like impact agents.

There have been agents who are impacted by some of the TypeScript exploits and package managers. There have been tons of things like that. And I think crypto at least gives you an incentivized bug bounty to find them. So I think that maybe is the only positive.

Well, the primary work that your team at Gauntlet does is advising protocols and and writing publicly on uh how to price risk. So I'm curious to understand your point of view on the premium you uh would put on the the the risk here and also in your perspective who who bears the primary risk uh in autonomous capital. This might not seem like the correct analogy to smart contract risk, but I think a lot of the edge use cases of agents right now such as lean theorem proving like companies like Harmonic and smart contract risk assessment are actually very similar in that the AI agent itself is not the actual thing that determines the final state. It's actually this execution environment that is formally checking like whether something was correct or not. And that process is very hard and rigid.

Whereas AI agent is much more fluid, right? It's like searching through the space of reasoning uh prompts and like giving you a final output. And so I think one important thing to remember is there are some things that are worst case problems like they're the hard they're kind of like you need to be resistant under every possible attack. Um and so that's the type of stuff I think that you see in DeFi. That's the type of stuff you see in lean theorem proving.

Whereas a lot of AI is built around average case behavior, right? Like I ask the same prompt 10 times, I might get 10 different answers. It's not deterministic, but those answers are close enough and I'm I'm fine with that. And in general, finance stuff doesn't work that way because that tiny gap is usually what you're exploiting, right? The tiny differences between the the same question being asked twice.

And I think effectively to make AI the more deterministic you make a AI agents the easier it is to do this type of risk attribution. And I think like realistically right now the only people I know as a lab who are really focused on this seems to be thinking machines where they're like trying to reduce all non-determinism as much as possible. And I think if you can get to that state then you can kind of be a lot more clear on like the finance how to avoid finance exploits. So that's I I guess what I'm saying is like there are things that are hard problems, there's things that are easy on average problems, and like how to connect those two is like what is still being figured out right now.

So there are two threads to what you said that I I want to lean into. The first one is you mentioned thinking machines. Thinking machines is not a cryptonative team. And backstage we were talking about a talent problem in in the industry. It's something that I think about a lot as an angel investor.

when you're investing at the early stages. I certainly think that it's a lot more about investing the founder. I know uh Terrun robot typically invest at at preede and seed if I understand correctly. And Alex, I think that's true for you as well. Um, and and then the the second uh thread there that I'm curious about is for you, Alex, and for you, Jordy, how you guys diligence uh teams and and also just generally strategies that your own teams are are building that are automated.

Um, so on on the talent side, uh, what are you guys seeing there? And also Amjad you are in SF

you see I think some of the best talent in the space I and oftentimes young people who I I want to pull into crypto

I talk to me about that all of you guys

I mean I have to just you know we're going to agree to disagree I think that the talent building in crypto AI is quite good and is on par with

I I invest in both sides traditional AI as well the people that I talk to in the crypto space are just as smart and talented. It's a harder problem because it's like, you know, it's a newer, less defined space, but I think some massive things will come out of it. And yeah, yes, a lot of the teams do have to be based in SF. You know, a lot of the teams we invested in, Prime Intellect, Eternis, um few a few others that are of course around all the AI labs and they're talking to anthropic kids and they're, you know, they're interacting together.

So, they're not any dumber than than the other kids.

Of course. Yeah. Um I think it's it's very difficult to find the right product market fit in AI crypto but I think in two five 10 years they'll be massive winners right now obviously we had the the wave last year with you know the more like memecoin agent kind of trading this kind of stuff which you it always starts with speculation I think in the next few years the categories like in two years I would say like we'll start having some very interesting autonomous capital things in like the wealth management side um you know helping people manage their their finances and I think crypto is like a much more international way where people can get exposure to I mean these days you can tokenize there's a ton of like XAT gold there's a you know everything's coming on chain you can start having intelligence and autonomous capital in your hands in an app someone's going to do this super well it'll take a couple years um I think in five years time you start having actual communities um obviously we're starting seeing like you know AI I agents basically like potentially be parts of governments.

U you know they're you can't corrupt them and so like there there's there's a way where they're more good for representative democracy. Um people just kind of give their preferences and intelligence actually acts on their behalf instead of humans that are more corruptible. So I think a lot of that stuff will start to appear in about 5 years and then in like 10 years I'm expecting that you know what I love Bitcoin. Bitcoin's great, but I do think that there is going to be space for some new forms of currency that are, you know, more designed for the AI age, whether that's, you know, we're seeing people trying to tokenize compute or in, you know, intelligence units or whatever it's going to be. It'll be interesting.

And I think at some point that that will start to appear as well.

Yeah, I think there's two points uh I want to make about the the talent thing. Uh, one, I think the adverse selection bias of being a crypto founder has actually always been the case. Uh, it it's it is a it's a fact that the best technical founders by and large are are in America. Uh, or they don't they're not necessarily Americans, but they find their way to America eventually. And up until like six months ago, there was a very real chance that if you were a successful crypto founder in America, you had like a over 50% chance of going to jail uh in the near future.

And that was been the case for a decade, right? So if you could build a multi-billion dollar business uh like job did without the risk of going to jail or a very small risk maybe, you know, not I I don't know. I don't know what you do in your personal life. Uh if you could do that, yeah, no brainer. You you would do that all day, right?

if you're a rational actor. Um, and so it takes a certain kind of person who's just like very ideologically interested or whatever they're a misfit or something like that uh to go into crypto and be a crypto founder. Um, and I think the crypto AI intersection is is seeing this situation as well. Uh, where there's still a lot of regulatory risk by doing crypto AI. Um, and you have to understand both things, right?

So the so the and I I found this too like I I've invested in like 50 AI companies going back to 2017 and I've seated like 10 multi-billion dollar even deca billion dollar totally non-crypto just straight AI companies right so I've seen I've seen both stages um and uh and yeah like the the amount of founders that can do that but then can also like understand tokconomics and incentive design is is is very rare right and it's also the amount of investors like very few uh traditional VCA AI investors are able to invest in the crypto AI category. Uh they just know that they'll just they're like they'll get fleeced by like crypto native funds or something. Um and then the second thing is that the to go to like to sort of the takeaway my takeaway from what uh Terrun was saying uh which is still a little unscrutable to me is um

is there are a lot of

it's it's okay I try to bring it down to the material plane is that there are still a ton of like very wonky infrastructural problems needed to be solved to sort of blend AI agents and and and uh and thinking machines to like commerce and finance, right? And so, uh, yeah, you need like wonky, nerdy founders to go figure out like how do you create these like decentralized markets for GPUs and compute and inference on chain and so it's not as sexy. Like there's not as really I'm pretty bearish like I'm pretty bearish a lot of like the AI consumer plays, you know, and I'm more bullish on the AI infrastructure plays that sort of make this stuff a reality and allow like AI trading to actually work for more than like 2 to three hours. Uh, but that's not as sexy. It's not as easy to underwrite, you know.

So, it takes time for us to get more to the consumer phase like we're doing in in normal crypto right now.

I I I think crypto has has a potential advantage in in talent in that you can make money much earlier in life and you you make both loads of money like so, you know, I grew up in Jordan and I started my career in like just building uh open source software. I actually had the idea for Replet like you maybe 15 years ago when I was in school maybe more. Uh and it was like there should be a place on the internet where I can go with nothing but an idea and be be able to build a business. Uh and I had a lot of business ideas but it was really hard to get started. Like just installing the development environment was hard.

You know kind of solved a lot of that stuff right now but like at the time uh I just had to build a lot of that infrastructure components and I didn't make a lot of much money on it at all actually. uh and despite that like the really large companies in the US and other places were were using my software. So again it goes back to this idea of incentives if I'm a kid today and you know wherever in the world uh and I can go you know pick up one of Tun's uh you know bug bounties and like make $100,000 in one shot. That's amazing. That's like never happened before.

And you have a talent pipeline that AI doesn't have because there's no AI buck bounties, right? and topic is not putting out like you know challenges like that and so I think leaning into that is is very good. There are also political forces that at play that I think is is potentially advantageous as an American. It hurts me to say this but like America is not very welcoming to immigrants as much anymore, right? And so I I went to the network school uh in Johor uh yesterday and I found a lot of kids there that would have otherwise been in America but actually they're going there.

They're spending, you know, fraction of a fraction of how much they would spend in America and they're starting to build businesses there from the internet and like no one knows that they're not in San Francisco. I know a lot of kids that are, you know, starting uh companies on Stripe, Atlas, they're Delaware based. There is there is from VCs via Zoom. No one knows where they are. So, it doesn't really matter anymore.

And I think uh crypto leaning into that uh leaning into that model will will create a pipeline of talent way earlier than someone becoming established in their career and having enough money uh to go and move to to San Francisco. So I I I think again I I I would lean back that the incentive structures in crypto could be very advantageous for hiring developers. I would actually add one thing which is I think it's actually a little bit of like a lack of imagination in the sense that in AI it's very easy to see like sci-fi the touring test all of these notions of like what it means to be more intelligent right there's like a clear notion of like my intelligence has gotten better in crypto and AI there's no like Cohen of wisdom there's no like oh this is the problem that you must solve there's no grand problem that crypto AI could solve like the best I could come up with is make an AI AI agent AI agent that like hedges its GPU cost and like costs much less than like all the other prevailing models that's not like very inspirational to you know like a

I don't know I I mentioned the book sovereign individual earlier when I read it I got super fired up like we can build a world where individuals can build multi-billion dollar businesses and they're they can be they can have political power and all that stuff which I think I think there's a version I think biology does a good job with this there's a version of crypto that that is incred incredibly inspiring and exciting for for the future, but not many people articulate that. I I agree.

Yeah. I just feel like the AI narrative like like super intelligence is very like easy to gro. If I'm like a 15-year-old who's really smart somewhere in the world,

it's very easy to imagine this idea that this machine is going and doing all this stuff for me. It's very hard to imagine this machine is doing things with money for me as a 15-year-old. That's like really, you know what I mean? Like so I don't feel like we've crossed the chasm of there's this great grand problem that that crypto AI solves and like this is the obvious way that you're making progress towards it which maybe will happen in in a few years. I think that's true with DeFi too though when going back to like 2017 18 when you started Gauntlet um these categories are just early you know it'll we still call it just like crypto x AI like what is it's just like this weird primordial soup where like various use cases and weird founders operate in you know um I think it's I I like that I mean I I've I've sort of made most money like investing in categories before they have names

you like the cryptoxai

no I like that it's it's so poorly understood. That's like where alpha comes from.

Oh, I see.

You know, like it's sort of it's it's sort of hard to figure out what the there's no easy narratives around it yet. You know, you want to invest a little bit before the narratives. Uh hopefully hopefully they form eventually. Um but that's okay. I I don't know.

I just think it takes a few years. Like it would be really cool if like businesses could be run by one to zero people, you know, and it was just all like intelligent designed over Dows. But yeah, man. It's like way it does seem way further away than even super intelligence, which is odd. Like

that's super years feels much closer looking up an AI to like Yeah. to like trusted execution environment. That's like 5 10 years, you know.

I mean, yeah. I mean, what you guys are saying certainly is true. it's a lot easier to follow follow like the you know what has the last 5 years of AI progression done and let's just add another 5 years and start imagining what super intelligence looks like. I think the cryptoi I think you know I've gone down this rabbit hole with a lot of inspiring people and um I'll say this I'll give some of the alpha it's hard for me to even try to I'm not publicly communicated on this topic but basically I think the the magic where it happens and um the alpha is in essence we have the 3D space where we're all here this is the human space and then you have the machine intelligence space which is obviously happening you know in silicone and you this disconnect between imagine like these are parallel realities that you've seen these movies like uh you know in space Matthew McConna you know he's he's there but then there's like these different dimensions just imagine two simple dimensions there's the human dimension we're all here there's this you know digital AI dimension what is the role of crypto and and it actually is the bridge it's the money that combines the two worlds because you know What is anything in the world? It's it's basically actions and actions are driven by incentives.

And how do we communicate with this like AI world? It's through, you know, something valuable scarce money in essence. So if you imagine that I'm giving money to the AI, that's a way for me to influence its actions. And if it's giving money back, you know, it's it's giving money to people, that's or that's like a communication from that dimension. So, I know it sounds a little bit woowoo, but if you start actually thinking about, you know, what does it actually mean for um AI crypto?

I think it's it's the bridge between these two worlds. Hopefully, that made some sense. I I I was just like add like a small thing here which is like um when you interact with even before chat GPT GPT3 anyone could see the power like think about how much you have to wait through in order to see the power of crypto like you know where's the prompt box of crypto so I think a lot of it yes narrative is important all of that but like just like show people the power instead of like just talking about it

so Amjad you mentioned the sovereign individual UAL and Jordy, you you referenced movies. What I so appreciate about what you said, Jun, is you are talking about storytelling effectively. And so I think that what we're all gesturing towards is actually beauty. I think about beauty often. And I I think the call to action here to founders or actually people who are just simply believers in crypto is not just you know leaning on people like Bali who have done so much for our industry or authors who who wrote uh books years ago but for people who are building new things to think a lot about simplicity and elegance.

When I am meeting with founders, I'm often looking for the removal. Like if you're building products, like etch away at the at the ugly, leave the the simplicity. I think that's something that you guys have done very well at at Replet. And I know when you guys are diligent in companies and looking at at products, that's something that we're all looking at. Um and so this leads to my last question which is uh if you were to construct a thesis around AI uh or a wish list uh of a a product or a company.

Talk to me about what that is. Or maybe it's an archetype of a founder. I share that with all of us. And it doesn't have to be in the next six months. We're long-term thinkers.

It could be over the next four years. uh but certainly something that is uh pushing forward autonomy. We're all people who believe in agency. Otherwise, why would we be here? Uh crypto is uh fundamentally I think about the ideology of of human agency.

Whoever wants to go first.

Yeah, happy to to share. The way I I think about investing in crypto AI right now is um it's very difficult to do actual AI inference. let al let alone training like onchain in a fully decentralized way. Um so but I like like but it's still generative AI this transition that happened with you know uh GBT 3.0 is like the biggest sort of tailwind uh in technology and in in the economy right now and so there's tons of those like tailwinds to pick up that blow blow things in crypto.

So I like investing in like preai on chain. So for instance like what are the ingredients that go into making uh intelligence? you have like data, compute, models. You kind of mix those three things in a bowl and you get artificial intelligence. And so uh all of those are like digital resources uh that are very scarce and are run by massive monopolies that are like de like trillion dollar companies and um uh and those companies like guard them religiously and it's very hard to access like Nvidia chips and Microsoft cloud and stuff like that.

Um, and so, but there's all these interesting marketplaces and financial tools that crypto has come up with to allow you to like uh tokenize and trade and share GPUs on chain and things like that. And this is all infrastructure, right? It's all like tooling that allows AI inference to happen on the edge and on on on um on chain over time. Uh, but it's essential, right? And the market sizes of these are really massive.

And I try not to get hung up on on the founder stuff too much because like I don't know Jack Maul was like a school teacher and Abe Lincoln like lived in a log cabin and didn't go to college. I don't know you know like it's it's uh early early eras like to call for the people like they make great founders you know who who go into these spaces who are like bold and stuff. Um uh but those are the areas that are exciting to me right now. And I'm I would generally be short like AI trading like man that I mean like I don't it's like like Selini does AI trading you know with like quant models and stuff like that and jump and Renaissance does AI trading like we should not we mere mortals should not get too involved with it you know unless we lose our shirts. So

yeah like I said by definition if there was if everybody could just have a chat GBT make money for them that edge would just disappear instantly. So it can't exist. Um for me like I said next two years the the massive use case I see is is just around you know personal wealth management. I think crypto's biggest problem out of everything and we have a lot of problems but the the biggest one in my view is the misallocation of capital. We really have a lot of capital and we just put it in all kinds of misallocated places.

Um and so in essence using abstraction which is such a powerful word because it kind of takes something very complex and gives it to like you know uh people that don't need to study it. So uh using abstraction using AI and starting to actually allocate capital properly hopefully some founder builds it. I someie guys are interested in in this idea. If not you know I I'll have to incubate it but somebody somebody's got to do this. Um I I would love to see a an attempt at building a market driven developer infrastructure.

Crypto is a detail of that. I think it's an important detail of that, but I would I wouldn't say it's a cryptonative package manager. just build like a package manager with incentives and protocols built in so that like AI agents such as Replet's coding agent could be able to provision a database by paying via crypto to some developer somewhere be able to pay for a proprietary package so that open source developers are able to to make money to to make it so that commerce is built built into the act of software creation. Uh I I think this is such a powerful idea and there hasn't been an attempt at it that is pure. A lot of it just has a lot of complications around it.

So protocol for provisioning and creating software could be very very powerful with with incentives aligned.

I think um crypto is very good at making contests. Mining was a contest. Staking to some extent is a contest. DeFi is a contest somehow. There's never been very good contests in AI in some sense.

Uh whether it's for bug bounties or whether it's for energy reduction like how do I get an algorithm that how do I incentivize the person who has the open source algorithm with the lowest energy usage and so figuring out the right set of contests that map well to crypto incentives that can be built today which might be things like the energy usage one. That's the one I think is like the easiest.

You're giving examples like like Frasa basically where people trying to compete to hack the the private key from a LLM, that kind of stuff.

Yeah, exactly. I feel like crypto is good at contests. So like lean into that.

My wish list is a beautiful mobile application that is built on blockchain rails. A very simple ask. We are at the end. Thank you so much for this fantastic panel and thank you all for coming this morning.

Thank you.