Pod Insights The Pomp Podcast #1348 Tom Shaughnessy | Crypto Investor Reveals His Top Investment Ideas About this episode Tom Shaughnessy is the Co-Founder of Delphi Digital and Founding Partner at Delphi Ventures. In this conversation, we talk about the intersection between AI & crypto, OpenAI & ChatGPT, open-source, where value is accruing, monetization, risks, and what he Is doing at Delphi. ======================= Introducing Espresso - the world’s most interactive portable display. They have a portable screen that is incredibly light, comes with a nice stand, and the user interface is very easy. Anyone who listens to this podcast can go to us.espres.so/pomp. They have a brand new offer waiting for you. ======================= BetOnline.ag is a proud sponsor of the the Pomp Podcast. Use crypto to bet on sports, play poker and enjoy casino games at BetOnline. Visit https://promotions.betonline.ag/pomp and use promo code POMP100 to receive a 100% matching bonus on any crypto deposit. BetOnline boasts no crypto transaction fees, and processing is anonymous, instantaneous and secure. ======================= Pomp writes a daily letter to over 265,000+ investors about business, technology, and finance. He breaks down complex topics into easy-to-understand language while sharing opinions on various aspects of each industry. You can subscribe at https://pomp.substack.com/ ======================= View 10k+ open startup jobs: https://dreamstartupjob.com/ Enroll in my Crypto Academy: https://www.thecryptoacademy.io/
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:02 Anthony Pompliano? What's up everyone. This is Anthony Pompliano. Many of you know me as Pomp. You're listening to the Pomp Podcast, which is my effort to find the most interesting people in the world and sit with them for hours while I ask questions in an effort to learn. So it would mean the world to me if you would subscribe to the show on your favorite audio platform, watch episodes on YouTube, and tell your friends and family about the podcast. My goal is to help millions learn from the world's most interesting people. So let's get into today's episode. Tom Shaughnessy is the co-founder of Delphi Digital and the founding partner at Delphi Ventures. In this conversation, we talk about the intersection between AI and crypto, OpenAI and ChatGPT, open source, where value is accruing, monetization of these different
0:44 Anthony Pompliano? technologies, where the risks are, and what he is investing in at Delphi. I always enjoy talking to Tom, and he has tons of insights in this episode, so I'm super excited for you all to listen to it. Here is my conversation with Tom Shaughnessy.
0:58 Speaker 01 Anthony Pompliano runs Pomp Investments. All views of him and the guests on his podcast are solely their opinions and do not reflect the opinions of Pomp Investments. You should not treat any opinion expressed by Pomp or his guests as a specific inducement to make a particular investment or follow a particular strategy, but only as an expression of his personal opinion. This podcast is for informational purposes only.
1:22 Anthony Pompliano? Today's episode is brought to you by Espresso, the maker of the world's thinnest portable display. Now, listen up. If you're like me, you feel like you are at a command center when you sit down at your desk. I got a gazillion tabs open and different windows for different activities. There's my web browser, my text messages, I have Slack open, and I got a Notes app. I normally work on a desktop and can be very, very productive. But everything falls apart the second I leave my desk. If I'm traveling, if I go to a coffee shop to do some work, or just want to work from the kitchen table, my laptop doesn't have enough screen space. I lose my command center and my productivity falls off a cliff. It's a major problem. But this is where Espresso comes in. They have a portable screen that is so beautiful that you think Steve Jobs came back from the dead to create
2:05 Anthony Pompliano? it. The thing is incredibly light, it comes with a nice stand, and the user interface So easy that I figured it out how to do it in less than three minutes. If you listen to this podcast, you know that's not an easy feat. So the espresso team and I, we became friends. I got to know them because I really like the product. And those screens, they now want to offer them to any fan of the podcast. So we struck a little deal. Here's how it works. Anyone who listens to this podcast can go to us dot E S P R E S dot S O. Or that's too confusing. Just go click the link in the description. If you go to espresso's website, they've got a brand new offer there sitting for you. You get a little discount and you'll get a beautiful screen.
2:46 Anthony Pompliano? Trust me, I use mine every day. You'll love the espresso screen and I think it'll make you more productive. Go check them out today by clicking on the link in the description.
2:56 Anthony Pompliano? This episode is brought to you by BetOnline. Do you like making a profit from sports betting? Well, set yourself up to take home the most profit possible using crypto to fund your sports betting. Casino and poker account at betonline.ag. You can avoid costly transaction fees, get your payouts lightning fast, and do it all securely and anonymously with the highest deposit and withdrawal limits in the industry. If you want to get in on the action, do it the smart way with crypto at betonline.ag. Head to the website, sign up with promo code POMP100 to get 100% bonus on your crypto deposit today. If you go and you deposit, they'll give you 100% bonus if you use promo code POMP100. BetOnline, the game starts here. BetOnline.ag. Go check them out today.
3:44 Anthony Pompliano? All right, guys. Bang, bang. I've got Tom here. Tom, artificial intelligence seems to have stolen the spotlight from crypto. But now there's an intersection of AI, crypto. You guys are making tons of investments here. Help me understand like, what is the intersection between these two technologies? And why are you guys so excited about some of the investments in that area?
4:03 Speaker 02 Yeah, Pomp, it's awesome to be here, man. Thanks for having me. And it's always great to chat with you. So I'm really excited. I think just to back up, I mean, a lot of the time when we invest, we want a cornerstone or sort of a foundational reason on why we're interested in the sector and then why we aren't, right? The thing about Crypto AI that really drew me in, you know, over a year ago, was just seeing how powerful ChatGPT was, right? You can enter in some text, you get a response, it's really good at predicting the next word. It gives you this feeling of sentience, this feeling of life, right? And when you start to extrapolate that, and you start to figure out that, hey, OpenAI has the lock on hardware, on talent.
4:47 Speaker 02 They have the lead, they have the largest parameter models. And you start to look at, anybody can start to build an application on top of OpenAI, right? There's already millions of ChatGPTs, all based on this single foundational model that will touch billions of people. And it gets really scary because this model is a black box and it's private. And if they want to tweak the parameters, if they want to, you know, bias against a certain group of people or a certain technology, or a president, or whatever they want to do, that will then flow through every single app built on top and alter society. And that shit starts to get really scary, right? So the Crypto AI world is the complete opposite extreme of that, right?
5:31 Speaker 02 It's open models, it's knowing the weights, or keeping them private, but being able to validate what's going on, right? It's being transparent, it's being able to build an entire world where it's transparent and we know exactly what's going on with the most powerful technology of our generation.
5:48 Anthony Pompliano? And so how does this technically work in terms of, you know, there's like open source? What is the difference between maybe just like plain open source versus the intersection with actual crypto blockchains and these protocols?
6:01 Speaker 02 Yeah, it's a great question. The crypto is the underlying incentive mechanism for the entire AI or Crypto AI area to run, right? If you start at the absolute bottom of the stack, you have meat space deep in projects, right? Projects like, you know, full disclosure, I'll mention somewhere invested in, but projects like Ionet, right? 500,000 GPUs connected around the world, right? Being able to do that and being able to cluster that. To do that, you need a token, right, to incentivize people to bring on their latent hardware. So you give them a token and in return, you get decentralized training, decentralized inference, whatever sort of, you know, mechanism you're looking for, you can get. So at the end
6:44 Speaker 02 of the day, the tokens are just incentive mechanisms for the specific AI apps. We can work all the way up the stack too, if we have time and if you're interested.
6:54 Anthony Pompliano? Yeah, I would love to hear kind of, so the deep in space is interesting, but like, when you start thinking of these individual technologies, where do you see maybe value accruing that isn't going to accrue to, you know, the Open AIs, the Microsofts, the Googles? Like, one of the things in AI that seems to be happening is like, the big winners of the past decade are just having their advantages and their moats kind of extended, right? They have the data, they have the distribution, they have the engineers, they have the resources. And so is it something where like, the intersection of crypto and AI will allow for these new upstarts to be able to potentially disrupt some of those incumbents?
7:36 Speaker 02 Yeah, I mean, phenomenal question. And I love the phrasing and sort of how you're looking at it. On a short-term basis, I think ChatGPT, Facebook, you know, OpenAI, Meta, Microsoft, I think they de facto have the winning lead, right? Meta has the most H100s in history, right? Thank God that they open sourced their models, right? So we could fine tune and edit them, right? OpenAI clearly has the lead and GPT-5 will blow probably everything out of the water this summer, right? But when you start to look at a medium to long-term sort of horizon, right? And you start to think through just the creative
8:18 Speaker 02 nature of technologies, it starts to get a little worrisome and starts to lead into the crypto AI space because of that, right? And I'll give you, I'll give you an example here. If you're a killer engineer, right? You want to create a new model and you work at OpenAI, like, good luck, right? Like, they're not going to let you just, you know, edit their model. They're not going to let you just have a field day and be super creative and change whatever you want, right? If you come over to crypto, you can fine tune any model you want, but in the best sense, you can also get paid for it and incentivize people to help you do that creativity, right? And I'll give you a couple of examples here.
8:57 Speaker 02 Myshell.ai is a project we invested in and basically they lower the barriers to entry to create an AI app to zero, right? You log on to their website, you choose any foundational model you want, right? Closed source, open source, doesn't matter, right? You add the data that you want, right? Let's say all of Pom's podcasts throughout history, right? You integrate it with Telegram or Discord and then you could add a crypto flywheel to that AI app so that people can own it. It incentivizes people to use it, et cetera, right? So in a matter of five minutes, you have an app that you otherwise could have never built. That has an incentive, that's global and that you can use. The same goes for building foundational models, the same goes for building fine-tuned models.
9:41 Speaker 02 So I think over the longest arc, crypto will have the most creativity and the most shots on goal on the AI space because of how open it is.
9:52 Anthony Pompliano? And when we see kind of this approach, I think it feels like, one, it's an approach to building kind of open source, the ability for anyone to edit, et cetera. Is there also a different technical stack as well? That, you know, maybe people that are kind of building in this crypto world are building with different blockchain protocols, databases, you know, scripting languages. Like what is the differences on the technical side?
10:18 Speaker 02 Yeah, another great question. I mean, generally to me, it seems like the creating is much the same, other than it being completely open and having much different incentive mechanisms, right? So you can either, you know, on the far left of the spectrum, you could build a foundational model, right? That's like what ChatGPT is. In the middle, you could fine tune an existing model, right? Take a foundational model like Llama or whatever you'd like and fine tune that for your specific use case. Or on the right side, you know, the far right. You have a model that already exists, you're using it, and it pulls in realtime relevant data from the internet, right? Like, what's the ETH price? What's the Bitcoin price? That's all realtime data. You can't get that out of, you know, zeros and ones in the model, right?
11:02 Speaker 02 So, three different spectrums here. But when you add crypto into the mix, that whole value chain changes a lot, right? And just to walk through an example here, if you want to train that foundational model on the left, right? Or fine tune an existing model right in that middle, you need a lot of data, right? Or very specific data. And there's a project out there called GetGrass.io. They have a million, maybe more at this point since we're recording, a million downloads of a Chrome extension that scrapes the internet for specific data, right? It's insane. It's one of the biggest deep in projects ever, right? Helium had a million IOT nodes, physical nodes. I think over the course of a year or two, Gras has done it in a couple of months, right? It's kind of insane, but it's software.
11:48 Speaker 02 What they do is, they scrape the entire internet's web data, historical and on a go-forward. They segment that data into folders that can be very specifically used to train certain AI models, right? And the obvious question is, oh, you know, can't centralized companies just scrape the internet? The answer is kind of no, right? Because if you're scraping the internet from a centralized data center, you immediately get blocked, right? They know that this is one single IP address repinging over and over. They'll give you the wrong info. They'll give you a flight price that's an extra 50 bucks because you live in New York, right? Like, they'll do all these things to mess with you. So, you know, Gras is just one example of augmenting that AI value chain using crypto, right? And it is pretty interesting when you get into it.
12:33 Anthony Pompliano? What about like monetization for these? Obviously, you know, crypto, one of the areas of innovation has been around economic incentives. How do you feel that's changing when it comes to AI?
12:45 Speaker 02 Yeah, it's tough, right? Because there's a lot of strong views on how how different parts of this stack make money. But at the end of the day, people have to be paying for a service that's actually useful, right? So, once consumers or applications on their behalf are paying for these AI models, then I think value starts to flow through the system, right? And there's a lot of people that you want to take care of, right? Like, you want to give some money to those foundational models, right? You want to give some money to the people who provided the data, right? You want to give money to the people who labeled that data,
13:25 Speaker 02 right? Label data to train is super important. You want to give money to the coordination mechanism, like a BitTensor, right? That that made this possible. So, there are a lot of ways that you want to, a lot of people you want to take care of. But crypto is one of the only places where you can raise money so quickly for a use case to have that creativity, right? You could raise money for a foundational model and everyone could own it, right? That's like a hat tip to Prime Intellect and Akash who have done that, or Akash has so far. But it is a, it is a complex value flow, for sure.
14:01 Anthony Pompliano? Now, what are some of the other projects that you guys have invested in that you guys are excited about? Or you think, maybe, would people understand how these technologies or products are working, they may better understand this intersection of crypto and AI?
14:14 Speaker 02 Yeah, I mean, one project that is pretty interesting, that I think everyone could sort of relate to is Test Machine, right? So, TestMachine.ai, you sort of, you know how everyone waits like six to eight months for a smart contract audit, and there's bugs anyway, and everyone gets hacked? These guys built a proprietary algorithm for literally just hacking your smart contracts. So, you link your GitHub, you drop a file, and bam, like this thing goes to work trying to drain your tokens, right? So, it's more of a developer use case, but it's one that really kind of makes sense and really gets it. The other ones we've invested in that that I'm pretty excited about are the coordination networks,
14:59 Speaker 02 right? So BitTensor is worth, you know, 10 or 15 billion dollars. I have my concerns, but I think it's a phenomenal experiment and, you know, cornerstone of the space. But there are other coordination networks, right? There are other projects that are the meeting point or the marketplace between model creators and end users, right? And those are like Sentient, those are like Alora and others that I'm really excited about. I'm also personally, not through the fund, but pretty excited about Morpheus, which is a pretty cypherpunk version of an AI network that has around 350 million ETH
15:43 Speaker 02 in its stake ETH pool, two weeks into launching, a month and a half ago, which is kind of ridiculous.
15:51 Anthony Pompliano? And what about from like a user standpoint? How many of the users of these products or services are going to realize that they're using crypto and/or AI versus they're just going to have a product that like solves the problem or does the job they need?
16:03 Speaker 02 I'm hoping that they never, that they never know, right? But it is sort of interesting because you're starting to see ways in which people are using crypto AI very differently from ChatGPT, right? I'll give you another example. When you enter a prompt into ChatGPT, right? You say, who were the founding fathers? You say, you know, tell me about Palm's history, whatever you want to say. There is a two-page list of all the rules and regulations it has to go through before it gives you your answer, right? Some of them are fine, like some like be diverse, you know, don't be racist, like things like that. But as you and I both know, that list is going to get long, right? It's going to get even longer.
16:47 Speaker 02 There'll be more restrictions, there'll be more rules, there'll be more things that it can and can't do, right? If you go on a crypto AI large language model like NQI, like E-N-Q-A-I, and you ask a question, you get a totally different answer because there's no rules, right? There's no regulations, right?
17:05 Speaker 02 So thinking through the future of who decides what those prompt guardrails are and what they aren't, I have a funny feeling it's not going to go well. So people are already seeing the difference between crypto AI and normal LLMs.
17:19 Anthony Pompliano? And when we start to see these, I'm going to call them crypto entrepreneurs or kind of crypto developers. How many of them have the technical skills to build the AI side, right? Like, it almost feels like in the AI world, there's such a competition for talent and these researchers. And people are just spending immense amounts of money, et cetera. And those researchers have a specific skill set, experience, you know, technical ability, et cetera. Do people in the crypto world have those skill sets and vice versa? Like, do the people in the AI world that maybe like don't know as much about crypto? And what that really begs the question of is like, when you're evaluating a team, what are the, you know, is it like, hey, we need some AI people, we need some crypto people?
18:04 Anthony Pompliano? Or do you find like, oh, we can find two or three engineers and they could do both?
18:08 Speaker 02 Yeah, no, I mean, it's sort of like your age old battle against Wall Street, right? It's very, there's a pretty big rift between crypto people and AI people, right? But if you talk to the AI people who have come over to crypto, it is hilarious and ironic how much they agree on things, right? So, once you're able to convince them, it goes a long way. And there, there's obviously more crypto people, crypto AI people than there are AI crypto people right now. But if you're able to find a team that has the AI part in it, you know, you start to get blown away, right? One great example of that is News Research, right?
18:51 Speaker 02 I know you asked there, they built a BitTensor subnet. These are just traditional, they're untraditional how good they are, but they're AI folks that met together on a Discord and have now built an incredible project as AI people, right? And it's crazy how fast AI people could run in crypto, because these folks ran into an issue, right? They saw Hugging Face, which is the largest traditional website for open source models, right? There's half a million models on there. You could just take whichever one you want, use it and, you know, have fun. But the leaderboard for which model is good and bad is flawed, right? You can All the specifics on how to gain that leaderboard are known, right?
19:33 Speaker 02 So people just make models overfit to the leaderboard and they're not commercially viable. Right? The people started to take those models and deploy them on BitTensor for different use cases. Right? But it sucked because they're not commercially viable models. These guys at News created a subnet, so one of the BitTensor sort of networks, to basically test any model against real time synthetic data in real time. So you don't know what the data is. You don't know what the parameters are going to be in real time. But hey, submit your model, we'll give you a ranking that's actually real. Right? So it's crazy when a group of AI people come to crypto and could do things like that
20:15 Speaker 02 in such a short amount of time because it just speaks to like no barriers to entry, real money flow. And I'm sure they're making a ton of money on the town incentives just from, just from that.
20:25 Anthony Pompliano? Got it. And then what are the areas that maybe you all are looking for opportunities, but you haven't yet found a company, you know, or kind of like, almost like a request for startup, if you will?
20:35 Speaker 02 Yeah, I'm really scared and looking for a really good proof of personhood project.
20:44 Speaker 02 WorldCoin has obviously done really well, but just massive FTV and sort of looking for some alternatives. It's already pretty clear you can't tell what is real or what is not real on the internet. That's only going to get way worse. So I think finding something like that is, is hard. Even if you find something, it still gets really hard, right? Like if trump.eth, you know, signs or stamps of video that he recorded something, you know, that that's great, right? But how do we know that he owns trump.eth, right? Like, so it gets really tough, but I'm really looking forward to some sort of proof of personhood
21:29 Speaker 02 or some way to validate that our, our world is, is real before it slips away.
21:34 Anthony Pompliano? So how do you balance like in that world, right? If you were to go talk to the critics of WorldCoin as an example, they'd be like, oh my God, this is like a surveillance tool. You could put your biometrics and all this kind of stuff. Like, is there a world where we can do proof of personhood without giving up some of that information or like, how are you all thinking about maybe what the right solution there could be?
21:53 Speaker 02 Yeah, I mean, I think the WorldCoin team gets some flack from just being so ahead of their time, right? It's usually hard to understand projects that sort of visionary that this early on. They've begun or started or almost done open sourcing their stack, right? They're using really intense CK technology to make sure that your data isn't being released, right? But you have to find an orb by you, right? You have to scan your eye. Culturally, people probably, long-tail people are not going to be okay with that. And there's also like a lot of weird stuff going on, right? Like people scanning their eye and then selling
22:36 Speaker 02 their identity, right? Like, it's going to be hard to get your identity back after you sell it, but it does get a little wacky.
22:45 Anthony Pompliano? And then what are the areas that maybe are too hyped up? People are too excited about when it comes to AI and crypto, but you guys are maybe not as excited or think that it may not actually accrue as much value as people think?
22:59 Speaker 02 I've had public concerns on BitTensor for quite a while. I think it's, again, like cornerstone crypto AI project. It is definitely attracting phenomenal builders, but there are some concerns, right? Like massively overpaying for subnets is one, right? Like you don't get that money back if people are starting to milk the network for some reason. The incentive flow is also currently sort of arbitrary, right? Like, you know, those who have a large stake decide where the tile flows and then within a subnet, they decide where it flows, right? So it is sort of a bit of a more manual process. It'll definitely
23:44 Speaker 02 improve, but I think that we need to be a little bit more critical of projects earlier on because as you and I both know, AI and especially crypto AI will hit a lull, right? It will go through a bear market and people are gonna have to build and really deliver. And I don't think we're anywhere near that bear market yet, but I think people should realize that it will come. People will get pretty down on the space until we have new use cases and then, as we know, it'll pop up once again.
24:18 Anthony Pompliano? And when we see kind of Delphi investing in this space, should we expect there to be faster liquidity, right? Because there's kind of crypto involved than maybe a traditional venture capital investment. Should we expect there to be more asymmetry than maybe a traditional venture investment? Like, why is the investment component of it so interesting? You guys could easily just say, hey, we think that space is going to obviously become popular. You know, we pay attention to the market. Sounds interesting. But the fact you guys are actually deploying capital, making a lot of investments here. Is there structural things that you think the crypto component being added to AI makes it more attractive from an investment standpoint?
24:58 Speaker 02 Yeah, absolutely. I mean, first and foremost, there just aren't that many liquid options right now, right? Other than Bitensor and a bunch of meme coins and stuff like that, there really aren't a lot of good liquid plays, right? Bitensor is pretty huge. Olas is pretty huge. Worldcoin is already pretty big. So there is like some market validation there, but a lot of the best deals are currently private. But yeah, I mean, just the crypto incentive to the entire stack here is just, you know, it's just really interesting, right? Like allowing, like just again, zooming back out to that foundational,
25:39 Speaker 02 like first principles level, like giving this stack ownership, control and a say in the most powerful technology of our generation, which is AI, is the only way that this can ever be unbiased at scale and legitimate, right? Like, I don't think OpenAI ever open sources their stack. I think this is the only way that we're going to be able to do it.
26:03 Anthony Pompliano? Do you ever consider AI is more powerful than crypto and kind of blockchains and protocols? Or do you think that crypto blockchains and protocols are more powerful than AI?
26:15 Speaker 02 It's a great question. I generally view them as extremely complimentary, right? I think that if we fast forward to an AGI, they are obviously not opening a Chase Bank account, right? They're not, you know, they're not, they're never going to send, an AGI will never send a Fedwire other than convincing a human to do it for some reason, right? The path of least resistance is for them to use crypto, right? To accrue value for whatever their use case is and to build communities around itself in crypto. And eventually, I think there will be a quote "team" that develops a crypto project and accrues a lot of value. And we later figure out that it is some sort of AI, right?
27:00 Speaker 02 You know, maybe we'll start to see something like that soon.
27:04 Anthony Pompliano? Do you think we're close to AGI?
27:07 Speaker 02 Um, yeah, I would, as a non-technical AI person, I have a less than informed view than a lot of other people. But I think that we are closer than people think. I think the OpenAI drama months ago, where the team went kind of crazy and just seeing the lawsuit from Musk and that discovery, just seeing the co-founders freak out over what they built. I definitely think we're closer than people think.
27:40 Anthony Pompliano? And what changes in the world? Like, one of the things I've always thought is this idea of machine-to-machine transactions. Like, obviously, they're not going to, you know, use two-day settlement time or wires to send economic value back and forth. So like, the quote-unquote "intersection" right now of AI and crypto, is it like they both enable each other? And almost like it's like one plus one equals five type scenario?
28:07 Speaker 02 Yeah, no, absolutely. I mean, it's it's just the the synergy will be crazy. And sorry, your question was on machine to machine stuff, right?
28:15 Anthony Pompliano? Yes.
28:17 Speaker 02 Yeah, absolutely. I mean, I think it's a little bit of a ways out before we have AIs just interacting with AIs directly, but it's definitely going to happen. It it's just like when you have, you know, back to our earlier example, like fine tuning that specific use case, right? Like, I think we get a world where there's millions of highly specific AI models that don't have to generalize about the world. And these models are working together with each other directly, right? Like, one will handle your Uniswap liquidity provision, if we're still doing that in a couple of years. And one will handle your trade on CowSwap, and then one will handle, you know, taking out that loan on AVE for you, right? And and they'll all work together sort of symbiotically. And the only thing you'll do is
29:03 Speaker 02 enter your intent or your desire in an LLM like you do on Gemini, right? Like, hey, I want to trade ETH for USDC, and I want to take out a loan on AVE and borrow Maker, right? Like, and you'll see like a simulated flow, you'll hit go, and that'll go out to a network of AI solvers, human or or machine.
29:22 Anthony Pompliano? Got it. And then in terms of like so much data being available in the blockchain and crypto world, it feels like AI is all about data. It may not be proprietary data because it's on these public ledgers, but do you guys see any opportunities there? Or have you seen anyone really trying to figure out like what to do with all this data and overlay it with AI?
29:42 Speaker 02 Yeah, absolutely. I mean, Gras is one of the best examples of that, just taking like massive amounts of data public and and use it to train specific models. But yeah, the the data side is interesting because models have grown in parameter count, right? They've grown in massive size, right? But they haven't the data used to train those models. Could be argued is not growing commensurate with the parameter count, right? And so when you start to get to really like Chagavati was trained on like Reddit data, right? And stuff like that, like, and obviously a bunch more data. But if you want to train a really good model, it doesn't have to have this massive parameter
30:26 Speaker 02 count anymore, right? It could be five or 10 gigabytes in size, right? Like the myshell.ai team just proved that last night, right? They they they trained a model with $100,000 that's more performant than Meta's small model, right? 8 gigabyte model versus 7 gigabyte model, right? That's crazy. And it's even cheaper to to fine tune that. So, yeah, you don't need these crazy, earth shattering data center models anymore. You just need, to your point, that highly specific data to to train it.
30:56 Anthony Pompliano? Got it. And then when when you guys are making these investments, like, what are some of the technical checklist or like criteria that you kind of go through to evaluate these opportunities, right? So like, we've talked about where the segments that you're interested in. We've talked about kind of the team. But are there other things that as part of the diligence process that you're really paying attention to?
31:17 Speaker 02 Yeah, I mean, the team is the team is really big, right? You need a team that both deeply understands AI and a team that is not a stranger to crypto, right? As we both know, the worst thing in the world would be this killer team that, you know, fails because they don't understand the token aspects, the community building aspects, the open source nature of all of this, like, that that always sucks, right? So the team is is really important. The other one is sort of like a crazy understanding of of why their project works really well and a crazy energy around it, right? Like, when we met the Lightworks founder, Gabe, who did data labeling, who does data
32:00 Speaker 02 labeling, his mastery of just the whole process was was crazy, right? So definitely want that that sort of mastery. And we also just want like a first principles understanding of of why crypto and AI, right? Like, why does this actually need this? And, you know, there are enough risks and venture. We want to make sure that we at least have a good, good foundation on why they're building what they're building.
32:25 Anthony Pompliano? And when you think about Delphi today, help us understand a little bit more just in terms of you guys have the research, you have the investing, you've got labs, like what are some of the updates in terms of, you know, team size, what you guys are doing, how things are going?
32:39 Speaker 02 Yeah, I mean, Pom, you're one of the people that helped us out the earliest and we're super happy to have you as part of Delphi Research right on that cap table. Annoying you early on was was probably one of the best things I ever did. So really appreciate you you being a part of it and helping out. Yeah, we have three, three entities, right? We have a research division which covers sell-side research, talking about the market, projects, things like that. We have the ventures division where we're investing our money into projects that we're really big on. And then the labs division, which sort of incubates from the, from the ground up. So yeah, dozens and dozens of people across these different entities. And we're really excited to to have the bull market back in full swing.
33:25 Anthony Pompliano? Yeah. And then on the investing side, like what are the other things that you're investing in outside of AI and crypto?
33:33 Speaker 02 Yeah, no, I'm glad you asked. I mean, we have a full gaming team that's that's super, super active. My partner Jan and Pierce and Jeremy, you know, crush that that area as well. Jan also spans DeFi and other things there. So we have a pretty active DeFi side. Our in-for team is also very active. So we made sort of foundational investments in Celestia and Eigenlayer and and others. John Girl, who's at Delphi, sort of leads a lot of that stuff for us as well. So I hope I'm not forgetting anybody. But yeah, we're, we've been very active across the spectrums.
34:12 Anthony Pompliano? And then maybe my last question is, for this bull market, are you guys investing in crypto and AI for this market? Or sometimes in crypto, what people will do is they'll make investments today and they know that it's, you know, four to five years away. And so it's actually like you're almost investing for the next cycle. How do you guys think about, you know, the length of time it will take to build this stuff out? And then also when you guys can realize, you know, economic gain on the investments that you're making?
34:42 Speaker 02 Yeah, no, I mean, it's it's tough, right? It it is sort of a sometimes that's risk that it's really hard to to figure out, right? Like, I'm pretty sure that most of these investments are for next cycle, right? And generally, you want to invest in the bear market when everything's boring and there's not tourists and the valves are great, but you know, you can't have your cake and eat it too, right? AI is crypto AI is happening during a bull market, so it is what it is. But I I think most of these will be next cycle investments, right? And I'm trying to instill like within Delphi that obviously this is the most foundational technology linked with the most powerful financial application ever, which is crypto.
35:26 Speaker 02 But look, there will be slowdowns in this stuff, right? We're in the hype right now, but understand there will be slowdowns. There will be times when these teams have to build in the dark for a while, but next cycle and hopefully this cycle, I think these projects will will really deliver.
35:40 Anthony Pompliano? That makes sense to me. Where can we send people to find you on the Internet or learn more about Delphi?
35:45 Speaker 02 Delphi digital.io and Sean C. 119 on on Twitter. And Pom, thanks so much for having me, man. You run one of the best podcasts in the space and it's always super energetic. Really appreciate it.
35:55 Anthony Pompliano? I appreciate you coming on and teaching me. I learned something today. I learned something every single time we talk, so I appreciate it very much.
36:01 Speaker 02 My man. Talk to you soon, Pom.
Previous #1347 Zac Townsend | Bitcoin Life Insurance Is Coming Fast! Next #1349 Adam Sullivan | Core Scientific CEO on Bitcoin Halving & Mining Industry Page updated October 19, 2026 · transcript and summary are machine-generated
