The Cryptopolitan Podcast 45 - Markus Levin from XYO Network

Added: 2026-05-30

[00:00:02][music] >> All right.

[00:00:07]Hello everyone. Welcome back to another Cryptopolitan podcast. Good to have you with us again. Today's guest has been in crypto longer than most of us have had Twitter accounts. He mined his first Bitcoin in 2013, co-founded XYO in 2017, and XYO is now one of the largest deep in networks on the planet with over 10 million nodes, $8.8 million in revenue last year, a Revolut listing in December, and their own layer one that went live in September. So, we've got a lot to get through. Joining us today is Markus Levin, co-founder of XYO and head of operations at XY Labs. Markus, welcome to the Cryptopolitan podcast. It's an absolute pleasure to have >> Hey, nice thanks for inviting me. I'm very excited to be here and speak with you with all the cool things XYO.

[00:01:09]>> Uh Markus, before we actually get into the weeds, uh I want to start where most people would like to uh start with. Um basically, what is XYO? Uh what does the project actually do? Uh if you could give us an intro in your own words, uh that would be great for the audience.

[00:01:28]>> Yeah, for sure. So, we started in 2018.

[00:01:31]So, XYO has been a a lot of things since then.

[00:01:35]And and but in the core of it, it's like a reality engine connecting real-world data and validating it with our XYO network, and then uh uh we are approving that data as well.

[00:01:49]And uh and we started out with uh location data uh for and did that for a few years, and and now we validate any available data I collected and validated it. So, and we also do some online stuff, so we we can prove AI agent work now and and AI output work as well.

[00:02:13]>> Amazing.

[00:02:16]>> In a gist like to make it even simpler, I think you can call us a a data company in the in the blockchain space.

[00:02:23]And which proves what's going on.

[00:02:27]>> Mhm.

[00:02:28]Amazing. Yeah, and I think something that I was also really fascinated about when I was going through your project today. And it always sticks with me when I read about XYO is the framing of it, right? As as a bridge almost between the physical world and the on-chain world.

[00:02:45]Um because what I see mostly in crypto even now is uh it's still kind of self-referential in the sense that it's tokens about tokens about tokens. And what you are really describing and what I've been seeing is a different category of problem entirely, which is, you know, how do you get real-world data on-chain in a way you can actually trust. And I think the trust layer is super important. Um and that's a much harder question than it sounds, right? Uh >> Really agree.

[00:03:19]>> So >> It's it's a it's a problem if I may elaborate. The problem for a lot of crypto projects are crypto for crypto, right? And so they understand the market is small and you know, if crypto goes down, their product goes down kind of kind of a way, you know? And and we are avoiding that by building a a real product, you know, which is useful outside of the blockchain and this is what we >> Yeah.

[00:03:41]Yeah, so I think um like my next question that I have to be honest about it as well.

[00:03:47]Uh XYO launched in 2018. And 2018 of course was um when I basically got into crypto as well and I I still remember that day like it was not too far away. But uh if we just do a quick mental scan of the projects from that era, most of them are are no longer with us right now. Most of them are gone. Either they have quietly faded out or they've pivoted into something pretty unrecognizable.

[00:04:17]Um but you're still building like you're you're you haven't really left. You've got the real revenue. You've got 10 million nodes. So, what actually kept you alive when so much of the 2018 class just uh you know, disappeared?

[00:04:34]>> Yeah, it's when we started, uh you know, it was after uh right after the ICO boom. We launched it into a bear market. So, it was a tough launch by itself. And so, we realized, you know, we need to build something real. And I'm lucky I started some other companies before and so, I I uh I I I know what's needed there, you know, and so we built a real business model for XY O. And a lot of other crypto projects, you know, they lived off their token treasury at the time. Yeah, they had they had some ideas and they were building and building and building, but then at the end they didn't have customers.

[00:05:12]And uh because their customers went away because they built for other crypto projects as well. The exception there is stuff that happened in DeFi, right? Uh DeFi obviously is around and does very very well. If you look at like Hyperledger or Uniswap and others.

[00:05:28]And uh they do well because they generate fees and those fees go back into the ecosystem. It's the same for us. We generate revenue from the data and uh and that's uh and that revenue, you know, uh powers the XY O and XO1 uh ecosystem. And so, to generate revenue and profits in some years, you know, then the then the when the tokens go down, you know, you're still around, you know, you still have your customers and and and life goes on, you know, it's Yeah, some years are harder than others, you know, but it's not always connected with the token cycle. And and so, that led us operate now for more than 8 years. And we have been doing pretty well, I think.

[00:06:15]Mhm.

[00:06:15]>> Yeah, I think the reason I I I also think that question matters is that I think survival in this industry is genuinely an undervalued signal in itself.

[00:06:25]I think every cycle there's a new wave of projects showing up making bigger claims than the last one. And the test, I think, of those claims is just are you still here in the next 4 years or so.

[00:06:39]And most of them Yeah, most of them aren't, so I think whatever you've done over the past 8 years or so is probably more interesting than whatever the next hot pitch might be or uh uh is right now.

[00:06:55]So, I think again, just bringing it back to 2018.

[00:06:59]Back then, I noticed that the XY pitch deck was talking about GPS spoofing, self-driving cars, smart cities.

[00:07:08]Uh That was the world you were prepping for.

[00:07:12]But now, 8 years later, how much of that original thesis actually played out the way you expected? And where did the world go somewhere you didn't see coming?

[00:07:25]>> That's a really good question. Uh Yeah, uh we it actually played out a lot as we expected, you know, we we built our proof of location technology, you know, so that's uh how we started, you know, we realized location gets easily spoofed or hacked, right? You can download a GPS spoofing app and pretend that you're in Chennai right now while you're actually in, you know, San Diego. And that's a problem for most apps on your phone, but also for, you know, self-driving cars, and smart cities, and automated supply chains, and and and so on.

[00:07:58]And so, we built this network of connected devices. Those devices do cryptographic handshakes to verify each other's proximity. And as you have more devices, the more interactions you get, the higher degree of of certainty you can have. Relational maps and and other things. And obviously, on top of that, you can build a lot of things, like another data network, or a supply chain, or or whatever you need. And And so, we we built that, you know, because we realized GPS has this problem with certainty, you know? Accuracy is is pretty good, but it has a certainty problem.

[00:08:37]And so, but once we build it, and and once we understood what was going on, then we started to build out our whole data network because we understood data so much better. And now we can validate and verify any real-world data. So, another sensory data from temperature sensor, let's say you need cloud data, weather data, what whatever you need, we can collect it and validate and verify it through our network, and and then, you know, provide it to web2 and web3 companies.

[00:09:07]>> Right. Yeah, that's that's fascinating anything. Um I also wanted to touch upon um a subject matter about specifically about DePIN itself.

[00:09:17]And uh I wanted a founder's take on it.

[00:09:20]Uh You've now survived, you know, multiple crypto cycles, like you mentioned, with real revenue.

[00:09:27]Um I saw that there was 8.8 million last year, which is more than I'd say most public crypto companies, frankly.

[00:09:37]Uh what do most DePIN projects get wrong about building a sustainable business? And um I I suppose the next part of that question is where's the trap that you've watched other people walk into?

[00:09:53]>> Yeah. Uh good question. The the revenue was from the year before that, uh the number you you mentioned. But, yes, we we're doing well.

[00:10:02]And I think the the the DePIN is awesome because it creates real business models, you know, for those who don't know DePIN is decentralized physical infrastructure networks, which allow you to to build something uh with outside uh investors and builders, like in in your communities, for example. Our case, it's a data network, but in other cases, you might have a a cell network or a solar energy network or a Wi-Fi network, and you name it, you know. Uh and so, people will buy devices or pay other people to to buy the devices, and then they earn tokens uh as as the data, for example, in our case, becomes useful.

[00:10:47]And and this way, they're able to decentralize the data collection and and the people who participate are able to earn assets, you know? So, most people live paycheck to paycheck in the world, you know? And if you're a part of the the network, like XYO, you're able to earn uh in our case, XYO tokens. Those XYO tokens might be able to appreciate in value in the future, right? And suddenly, you have an asset working for you, you know, which is special. In our case, you can also use the XYO to stake it in our XYO Layer 1 blockchain and earn XYO tokens, as well.

[00:11:26]And and so, uh the it's a awesome model, but the there there have been some traps in in the past in deep in and others, you know, like some people used the easy revenue route and deep in their business more about selling their devices than than, you know, making revenue from partnerships, you know, and those deep in are not doing so well today anymore. And then overall, the big traps in in crypto are that crypto built for crypto, like I said in the beginning. And doesn't touch the outside world, you know, but crypto when we started out and I'm based in in California, you know, only 2% of US people had touched crypto.

[00:12:10]And you know, now it's about 25% but the the outside market is very small. And so if you've got from from crypto nerds to crypto nerds, you know, then the market is small. So you need to find a way to to broaden access. And in our case, we built a app for something called Coin. You can download it on coinapp.co and with it you can just facilitate data collection and and verification and you earn points which you can exchange for XYO.

[00:12:43]And and you can download the app for free.

[00:12:47]And today more than 80% of the people who download the Coin app are non-crypto users.

[00:12:53]And about 95% of them exchange their their points for XYO or they could also exchange them for BTC or ETH or gold bars or PlayStations and and stuff like that. And and so we make it make it easy for anyone to to participate. And you can also redeem your XYO tokens directly to your Coinbase email address so you don't even need to worry about a wallet. And so our addressable market for data collection is is just huge. And at the same side thing want to be side, when we started out.

[00:13:28]We wanted everyone to pay us with XYM.

[00:13:30]Oh, we only wanted to work with crypto companies, and and that was just difficult, you know? And so over time, obviously, we realized that we should take any type of money. It doesn't matter if it's rupees or US dollars, Bitcoin, or or XYM, and then use the proceeds and and put some back into our ecosystem. And And and work with anyone. And and that increased our addressable market uh a lot, and and that has helped us to survive and and survive for all these years. And, you know, 8 years later, that's a long time not only for a crypto company, but also for regular tech company. You know, most tech companies uh are folding after 2 or 3 years. And and so, we're lucky to be around, you know, after more than 8 years.

[00:14:16]>> Yeah, that's really fascinating. I think as a crypto nerd, like >> Your feedback came back.

[00:14:21]>> Oh, it's come back in. Just give me a second.

[00:14:23]>> Something's It's like rubbing on the >> the mic or something, maybe.

[00:14:27]How about now? Still?

[00:14:29]>> No, no, it's good.

[00:14:32]I I a little uh like uh >> Ah, okay. Maybe probably now?

[00:14:38]>> Yeah, now it's good.

[00:14:40]>> Better? Okay.

[00:14:41]Um yeah, I was just saying I think it was That's a really fascinating answer because as a crypto nerd myself and a research analyst, this is something I know for a fact that I'm going to be diving into right after this podcast.

[00:14:53]So, uh I I definitely see where the project is accruing value.

[00:14:58]Uh and the reason why I think I asked that question as well is because from the outside, I I do see certain traps uh within the deep in projects' design and the supply side.

[00:15:10]Uh you know, they've got token incentives, they've got a node sale.

[00:15:15]Uh they've got people running hardware.

[00:15:17]And then, all of a sudden, there's nobody on the demand side actually paying for the data or the bandwidth or even the compute. So, I think that that entire sort of dynamic is something that I've been trying to really get into my research as well. So, thank you for that answer. It is very really really well put. Um something that I've also been noticing within the deep in space uh is a debate that's quietly happening right now and it's between consumer deep in versus enterprise deep in, right? Um on one side you've got network like yours where the nodes are everyday people with their phones. Uh but on the other side you've got network selling directly to enterprise as the user.

[00:16:06]So, I suppose my question would be where do you think the real value is actually being built? Um or is it or is you know that even the right way to frame it?

[00:16:19]>> The value is built it like it's circular, I think. Yeah.

[00:16:24]It's because it's B2C, so business to consumer to business in a way because you have the the consumers being the nodes and benefiting from sharing data and collecting data.

[00:16:38]And so they earn earn economic rewards for that like XYM tokens. And and then on the enterprise side you know, like there's demand for the data which is which is generated. And so, if you this value all all around I think being being created.

[00:16:59]And it's it's a dependency in a way, you know, like the world is very data hungry, especially the AI age. You know, as as you know, like AI needs three things like data, compute, and and storage. And we provide the data part. So, data will be more valuable over time and and the AI models need more and more data to to learn. And so, now is the opportunity for someone to participate, right? And and be part of of the solution for that.

[00:17:33]And there's a lot of value being created. Our top 500 users, they made together $14 million over the years.

[00:17:42]Amazing because, you know, they they got tokens. You know, then then our token goes up and down as as most crypto tokens do. And and you know, from like 1.1 billion market cap to 50 to 500 and does these things, you know?

[00:17:57]And so, if you cash out at the right time, right? You you might be lucky and and and earn money at the right time. And that's we have some incredible stories.

[00:18:11]And and so, uh it's basically is the opportunity for everyday people like you and me to to participate in the AI economy in an easy way by just collecting and and verifying data and taking control of your own data as well.

[00:18:31]>> Yeah. It's amazing. I was just also looking at your website and I think there's a number that really stands out to me and it's it's right here. The 10 million plus network nodes and I just wanted to touch on that as well.

[00:18:48]Uh So, yeah, you've got over 10 million consumer nodes. That's a massive number, but I want to get specific, right? How much of that network is actually relevant to what you're now building with data? Because I see that the Coin app nodes are mostly providing location and device data, right? Mhm.

[00:19:11]Theta is doing something quite different. They're running AI agents for sports orgs, e-sport teams, and and so on. So, I think it would be useful to understand the bridge there.

[00:19:24]What is the consumer node network actually contribute to the AI agent verification problem?

[00:19:33]>> Mhm. Yeah, let let me go back and explain a few things. So, we have we just launched our XY O layer one blockchain. It's the first data focused layer one and it's for you know, data hungry industries like AI, autonomous vehicles, gaming, you know, you name it. And so, the XY O layer one runs with the XYO token, which is the gas of that thing. And then we have the deep in network and the deep in network generates a lot of data through all of its nodes.

[00:20:06]And the deep in network has it's run by the XY O token, which is for the governance, security, you know, staking, rewards, and so on. You can stake the XY O token in the XY O layer one to earn XYO.

[00:20:21]And so, they are in in a in a way codependent because we generate a lot of data and to make the data immutable and provide provenance and and so on for the data, right? We we put it through the XY O layer one system. And so, with Hadron Labs, we have a super exciting partnership. They are in Edge Cloud, which has more than 30,000 nodes for compute and AI agent building and so on. They work with big sports teams like the Houston Rockets and Olympique de Marseille.

[00:20:57]And they create AI agents for them. And but a a problem is that a lot of of networks as as you mentioned don't they they pretend things, you know, they pretend that they have nodes, they pretend that they have AI compute network and these kinds of things.

[00:21:19]So, what we are doing actually, we use our newly released XY O AI SDK which we just released the last week and we are verifying that the AI compute infrastructure is actually real and life and and that it's performant in the in the way consumers expect. So, basically we we are like AI infrastructure verification piece there and then we take that and write the data onto the XY O layer one to make it immutable and and searchable. And uh that's one side and the other side is we also interact this through our XY O AI SDK with the the AI agents and the underlying model and we are verifying that they're doing what they're supposed to do and we we we we record the answers on the XY O layer one and the data lake connected to XY O layer one and so that they become immutable. And so if you see an AI agent does a a wrong thing, for example, you know, you can check like okay, is the underlying data for example wrong and why does hallucinate and doesn't do the right actions. And so if we verify and the super cool thing is we built it with this XY O AI SDK which allows any white coder to build a product on on the XY O layer one. So, it's uh it's it will take you days or hours depending on on your skill level to build a product, even a game or accounting software, AI managing your robot fleet, you know, whatever you need, uh you can do it there and we expect millions and millions of products to be built there. And and uh and uh bunch of Feder Labs is is a first proof point, you know, that they're able to do that and we do something really cool proving and verifying their AI infrastructure and AI agents. And cuz the deep in networks that deep in network in this case is is not really involved, you know, uh in but uh we connect another deep in network Feder Cloud is cuz Feder Labs is is uh deep in network and that deep in nodes we verify basically their activity as we do for for our own.

[00:23:45]>> Amazing. Um I think there's one thing that I I've just been chewing on since I read it on I believe one of your interviews.

[00:23:54]Uh I think in the past you've argued and I'm paraphrasing here, but that the more value you attach to data, the more people are incentivized to cheat, right?

[00:24:07]Uh does that same logic apply to AI infrastructure performance metrics? Because if I'm running compute or say I'm running an agent and my payment is tied to performance metrics, I'm also reporting.

[00:24:24]That's a pretty obvious set of uh incentives right there. So >> Absolutely. Yeah, you're absolutely right. You know, like as soon as there's money attached there's cheaters. And the more money is, the more cheaters, you know, and more hackers, you know, we see what's happening in the DeFi space. You know, there's so much money flowing around and so there's a lot of hackers around. But it's the same thing for for other categories like computer infrastructure. And what we are doing is we allow our Feder to say, "No, you know, a third party has has verified that our infrastructure works the way it is intended and they record it here in an immutable way. Nobody can change the data and and that's it, you know, and that makes it much easier for them to to sell their infrastructure, of course.

[00:25:16]>> Right. Yeah.

[00:25:18]Um I also want to make this a little concrete. Um So, Theta already has AI agents handling live customer interactions before different teams, right? At scale as well.

[00:25:32]Uh now, when those agents run without you know, human oversight and without any independent record of how the infrastructure actually performed, uh what's the actual risk like there?

[00:25:46]Uh if you could perhaps walk me through a scenario where this goes wrong and somebody actually actually cares.

[00:25:55]>> Yeah. Um I could AI agents, you know, need Someone needs to watch them, you know?

[00:26:03]Uh because, as you know, there's a lot of hallucination in in models, you know, and the hallucination from the models translates into the agents, which are run on top of the models. And and [cough and clears throat] they it can be a a big problem, you know, if you have a customer support issue in in health care, for example, and the agent might make a wrong recommendation, right? And then suddenly you don't go to the doctor, but you should have gone to the doctor uh because the underlying model hallucinated.

[00:26:32]That's a big problem, you know? And and so, uh what we are doing is verifying that the model is running the way it's supposed to run and we we provide immutability and all that answer and check the answer. But, more than that, we can also uh make sure that uh underlying data is correct. We built this technology or this algorithm called a proof of origin, which can prove where data came from. So, let's say data came from the Theta Cloud, we can we can prove that it really came there and then it ended up in a smart contract, for example.

[00:27:13]So, we can prove the path of of data and that makes it highly verifiable. So, it that and that will lead to less hallucination of AI models because they are you know, they run more on verified data. And so, Deepen our Deepen Network plays into there because if a model, for example, lacks some data, we can we can say, "Okay, I need data here. I don't have a answer with this a high degree of certainty." You know, our Deepen Network can go and and collect the data in the real world. For example, it was someone wants to know, you know, does it rain in in China today, you know, and then take a picture of of the clouds.

[00:28:00]Then then you know, a bunch of pictures are taken and this proof of location and other other algorithms and proofs, you know, we can prove that it was actually generated there.

[00:28:10]And then, you know, the data was put into the model and run through our XY O layer one to make it immutable. And that is highly powerful for the world of the future.

[00:28:24]>> Amazing.

[00:28:25]There's something that I just want to drop on the screen as well, which really caught my eye.

[00:28:31]So, this is something that I read from a report from McKinsey.

[00:28:36]And they say that basically 23% of companies are now scaling agentic AI systems, meaning AI that can independently execute tasks like customer service, procurement, data analysis, right? Without any sort of human approval in the process. Uh now for me, 23% does not seem like a small number. And the verification infrastructure for those uh sort of deployments basically does not exist yet.

[00:29:05]So, my question, I suppose, is is that technical gap is it a commercial gap or does the industry just not see it as a as a real problem just yet?

[00:29:20]>> Yeah, I I think bad things need to happen before people change, you know? It's uh like uh when big systems get hacked, for example, then uh uh with agents, for example, then, you know, like uh people understand, you know, that that the problem [clears throat] existed and they do something about it.

[00:29:42]And it's the same here with the verified data. I think we're all just we're just so excited about everything AI and agents, you know?

[00:29:52]Uh because the potential is just huge, you know? Like the world in in 5 years, how is it going to look like, you know?

[00:29:58]We've got quantum computing is going to launch next year, you know? AI is going to use it the year after, probably. And and then, 5 years, who knows, you know? It's going to be going to be crazy, I'm sure. And uh And but someone needs to to verify uh that at that the at machine speed that uh uh that world uh is and the models and the agents run the way they're supposed to run. And And humans won't be able to do that. And you need a third-party verifier, a decentralized one, like XYO, uh to be able to verify that everything's running the way it needs to run. And And 23%, you know, that happened in the last 2 years, basically.

[00:30:41]Agents went from 0 to 23% you know, we got to be at 80% uh in another 2 years, you >> Yeah.

[00:30:51]Um I also think that there might be certain pushback here, right? I'd expect, for example, from a CTO listening to this uh when an AI agent fails, say in production right now, enterprises might tend to fall back on um say contractual agreements or guarantees for their infrastructure provider, right? There's There's an SLA, there's a vendor relationship, or there could be a lawyer even on speed dial, right? Uh what does cryptographic attestation actually give them that the contract doesn't?

[00:31:28]>> It gives you a high degree of certainty. First, it makes a contract It should make the contract cheaper, right? Because you don't need these insurances, you know, like like lawyers, insurance, SLAs. You need to keep a team on on staff to deal with things, yeah.

[00:31:45]So, the overall contract the cost should go down if you are able to to verify. And then it will allow the the the infrastructure builders, you know, which need the actual verification they are able to provide a better and more secure product. They become more competitive.

[00:32:09]And so, uh I I think it's going to be essential in the future cuz someone needs to verify, you know, what AI infrastructure and what AI itself is doing. Uh it's the right thing.

[00:32:23]>> Uh just the last one then um and then we wrap up. Uh most of the industry has framed AI's data problem as a a sort of model problem, right? Better models, better fine-tuning, more parameters.

[00:32:43]You've spent the better part of a decade arguing that it's actually a sort of in- infrastructure problem. The data going into the models isn't is, you know, actually what's broken, not the models themselves. Uh do you think that that conversation has shifted? Are people finally hearing the infrastructure argument, or are we still in the, you know, just the just scale the model sort of era?

[00:33:11]>> [laughter] >> I think it's it's slowly shifting. I think scale the models is is definitely still the leading voice, I think. But if you think about let's just say data is not that rich, you know, like in China, of course, you know, it's a high density, it's a well-connected city, but if you think about a more rural place, you know, then they might not have the data and the sensors and everything that needs to be there. So an AI model might rely on one sensor or two sensors there to give its answers. So it's it's highly uncertain.

[00:33:47]With a deep network, you are first able to provide more data points with a much higher degree of certainty.

[00:33:57]And then you're able to to verify that data not only through the deep network, but also through the explainable AI one.

[00:34:05]And and so the data is verified data is ultra important because then, you know, the model should hallucinate much much less. And if it does that, you know, then in high-stakes moments, you know, like health care, for example, you know, uh you don't run into problems.

[00:34:26]>> Absolutely. Uh before we wrap up, Marcus, I actually want to give you the floor one more time. Any Any closing thoughts you want to leave the audience with, whether that's something about XYO something about Deep in as a category as a whole?

[00:34:41]Um or honestly just something you've been thinking about lately that you think more people in the space should be paying attention to you.

[00:34:50]>> Yeah.

[00:34:51]I think it's an exciting time and because AI and and blockchain is going to connect. And I think anybody can be an entrepreneur today. You know, it's like I encourage go to xyo.network and interact with XYO AI SDK. You know, even my mom can can write code you know, a game there and and connect it with blockchain. You know, and once you do that, you know, you you're part of the story. But if you don't want to go that far, you know, maybe follow us on on X or it's the official XYO or you know, sign up for our newsletter.

[00:35:27]And and just become a part of the story, you know, and and the world's going to be going to be very very very exciting.

[00:35:38]>> Fantastic. Marcus, thank you so much.

[00:35:41]This has genuinely been one of the more substantive conversations that I've had on here.

[00:35:50]There's a lot of crypto founders who can talk for an hour without saying anything and you are very much not one of them.

[00:35:57]So, thank you for the time and for going deep on this topic with us.

[00:36:02]For everyone listening, go check out xyo.network and follow their Twitter as well.

[00:36:09]Uh Yeah, awesome. I think that's it for today's Cryptopolitan podcast. If you got something out of this one, remember as always smash the like button, subscribe to the channel and we'll see you all in the next one. Marcus, once again, thank you so so much.

[00:36:26]>> Thank you and nourish your spirit.