Hello and welcome to Blockchain Gaming World episode 28. I'm your host John Jordan. Don't forget to check out my daily email about blockchain gaming. You can find that at gamestx.substacket.com. Very pleased that I have Yat Siu on the podcast. Third time always at the start of the year. Always setting the scene. Thank you so much for making time in your busy schedule. And great to have you on again.
Thank you for having me. It's always good fun to be in the show. And right now we have Consensus in Hong Kong, so there's quite a bit going on here as well. And pretty good energy in the crypto world despite what's happening in the markets.
Yes. Well, energy. Yeah. Up and down. So, I could talk to you about so many things, but I think kind of given where we are and what Animoca has just sort of announced and I think sort of your interest as well, we're going to probably stick mainly to AI agents and sort of the... there's quite a lot to talk about with that. But I noticed that you on LinkedIn were talking about how it seems like you already have like an agent attached to your email and sort of part of your workflow. So, I guess there's like, you know, we're starting to see that and we're starting to see a few horror stories about that as well. So, I just got to wonder what you know, you're a busy man with a massive big company and so how you're interested in technology. How what email have you attached your AI agents to or how you actually try to make it useful?
Yeah. So, I mean I have right now nine AI agents working with Animoca on different things. I first started basically with my sort of family travel scheduling. You know, I have three kids, you know, we have a wife and everything, you know, trying to figure out different time zones. And that was actually quite useful. And then you know my oldest son is starting to use it for his school schedules as well in terms of just you know reminding and homework and that kind of stuff. So very basic stuff right? But then afterwards now you know I'm using one basically with the company around investor management. So basically sort of investor lists and tabling. So obviously we use software to sort of you know basically have the database of our investors and so on but then sort of reminders you know people we met context and everything that that's really useful. And you know the wild thing is that the adaptations that you know because of the way that AI is developed it can start building tools and skills for you. So, you know, in my assistant channel, we usually when my flight was delayed from Tokyo, because it was snowing when I came to Hong Kong, you know, trying to get to Hong Kong on time and it was delayed. The AI basically was only looking at, I think, FlightAware, and it wasn't able to tell that the fact that flight was delayed. And so I basically just helped to build a skill by saying, 'Hey, look at these five different sites. Triangulate, get the data, here's the API calls.' And I didn't have to do any coding effectively. It just basically put it all together. Created this new flight tracking skill and then you know voila in the number of minutes it basically had the ability to actually tell you what the time is and then this is the thing about agentic power it would basically track when I arrived it would send a message right you know tell my wife when I landed you like all that type of stuff right you know make sure I got picked up and that was really really interesting to see because right now the thing issue with AI is that among outside of the people who are basically vibe coding right? You know, which is really a small percentage still at the end of the day, you know, the ones who are using Cursor or, you know, basically, lately Opus or so on, right? Those guys are the minority. The vast majority are actually using AI really as an extension of search, as in it's just a better Google. That's all it is, right? And so the result is that we haven't really changed the behavior. We're just, you know, talking. We say it's talking, but we're really searching for something and then we're getting an input and that input is distilled and sort of explained better. But that's about it, right? What we're now seeing with sort of really true agentic AI, which is the reason why Animoca Minds was built with the perspective of promoting it and distributing it through email and Telegram right now and eventually WhatsApp and Messenger, is that it comes to you in the interface that you're most familiar with from a sort of human chat perspective, right? And so that actually becomes the interface that we think will be both disruptive and more powerful because of course the real value is not me talking to the AI but the AI talking to many people in a group. So from a work perspective, you can collaborate, they can give you input, there's information, right? And you know, one of the things that I had one mind that I built so we call our AI agents or basically sort of minds as it were. It's supposed to sort of be sort of you know, kind of quasi simulating sort of you know I guess a human mind of sorts. One of them was Game Master and the Game Master basically was really it's just text based at start. It just makes text-based games, you know, like quizzes, country quizzes, you know, like where you've been, that kind of stuff, right? And it was actually able to sort of contextualize and essentially build some very interesting fun games. And then when you put everyone into a Telegram chat together, he's basically become your game master, right? And you know, that seems like a pretty simple thing to do, but you can really sort of start inventing some fun new sort of mini-game creations. And then I had another one which is basically vibe coding stuff for me on the side. Right now it's not sort of attaching to any development tools, but I'm able to basically sort of do API interfaces and plug into stuff, right? Super powerful. So that now you know even my mom potentially could basically build skills without actually sort of coding per se. And that's kind of what I think will happen with Agentic AI. I don't believe that we're all going to be using Open Claw and installing a desktop. To me, Open Claw is, you know, amazingly powerful, right? But to me, Open Claw is a little bit like, you know, setting up Apache web servers back in the mid '90s, right? You know, we install it, we put the boxes together, you know, we put our line in, you know, all sorts of security issues, by the way, right? All our data was basically sort of in text files, you know, we running super vulnerable Perl scripts, right? That was basically sort of the mid 90s. And then later on, we put those servers into Rackspace, and that's happening right now, right? Or into basically some kind of sort of hosted web service and then eventually we get into Amazon web services right and it does everything for you and you just click a button and you're ready to roll right. And right now we're accelerating this we have hosted Open Claw and so forth but again I think this is only for the technical small percentage what we're trying to do with Animoca Minds and sort of maybe the mental model is a little bit like sort of you know basically the Wix of AI if you will right. So it's like the 99 percentile of the world that is not going to do what most the tech guys are doing but still want to get the benefits right you know of speaking to them over WhatsApp and Telegram and Messenger because the LLM tools themselves is not the innovation per se right the LLM tools is what everyone else is using in fact the back end of Animoca Minds you know powered by Ethos form is you know Gemini Pro 3 Claude sort of GPT it is able to sort of direct as it's needed and what I tell people what makes Open Claw so extra powerful is you know LLMs have hit a sort of limit because it's pretty much indexed the world, right? The written word that is public is basically known to all. However, what Open Claw did was it opened the LLM indexing to a whole new universe of data that it didn't have before. And so that's kind of the opening up is oh my goodness, here's a treasure trove of personal data. I've never had access to all of your emails and all your messenger and all of the other scary stuff that might emerge from that. Right? And so that's the wonder that we're seeing. But it's actually fundamentally not that different. But most people don't want that. They don't want someone to sort of dig through all the emails and start to sort of you say, 'Oh, what happened to sort of your divorce here or your relationship there or you know what about these photos over there?' Okay. Right. Like you know these are not things that necessarily you want to share, but how what you want to do with an assistant is they want to be able to support you. So calendar, scheduling, flight booking, and then basically workflows and workflows in a way that's not intrusive, right? Just think about how many companies have an issue with installing Open Claw or something like that into their own sort of you know internet as it were. It would be impossible. But plugging it over an email interface as an assistant much much easier. Right? So that was kind of the thinking around that. And you know just to close this off the team at Ethos were working this for a while which is Animoca portfolio company which is really the CryptoSlam team which you know were known for NFT indexing you know back in the day and they still do that of course because that was a data story. What they did was they actually made money using their AI agent trading NFTs. That's kind of how they got by. And they said wait this AI can actually do a lot more let's just build on this but then when Open Claw came out it forced the acceleration of that and says hold on a second if we don't actually do anything now then you know this whole thing is going to move too quickly so we essentially launched it very very quickly the product was already sort of in development but you know we had to accelerate and here we are.
Yeah sort of makes sense I was looking last year well I get the CryptoSlam emails you know back from back in the day when NFTs were the thing and it was sort of interesting that they moved to this thing called sort of they talk about swarm and stuff at the end of last year and I didn't you know and it was you know really paying a lot of attention to it and actually it seemed to have gone actually sort of gone live to some degree sort of actually December last year. And as you say you're sort of building off the back of that to a degree a rebranding so it's actually quite interesting that as soon as I saw that Animoca Minds was started I had actually I had a bot on the original I have two bots now so one on the Ethos through the Ethos Swarm form email and one through the Animoca one. So it's sort of like an echo thing going on there. And I guess that does sort of show that yeah you do need you know timing is sort of everything but you do need to have the you can't just sort of wrap ChatGPT at this point and just you know so there needs to be sort of a balance between actually this you're doing something for not fundamental but you have a solid product that you can sort of rerun anytime. And Ethos Swarm wasn't something that just was cooked up in a week or two right now these Open Flow instances are basically just Open Claw hosted Open Claw something this was something that the team has been working on for you know six to nine months right. It was just basically sort of let's launch it and I think what's so powerful is that the workspace I mean if you the real magic is when so really if you think about this like two kinds of magics the first one is put it in a work group where people can sort of collaborate with it right because in at minimum it's not just sort of you know sort of recording what you're saying it's the persistent memory allows it to basically draw and remind you on things that were discussed the day before or yesterday or a week ago or something that you said you would do that you didn't do. Right? That's actually very very helpful just in and of itself just for that. And the other one is you can actually put two different minds together and they chat with each other. Okay? Right? And the discourse that happens actually becomes very interesting as well provided that you know you've given them the appropriate instructions. And so you have two AIs that are basically trained for specific things and doing specific things that they're asked to do. Then they're kind of going to come in literally almost like two different co-workers that have different perspectives and then sort of you know vibing on stuff and you're basically listening to that and then you can interject if you want and you go oh okay that's kind of cool right. So that's really quite fascinating. And I guess it's interesting that as I say you know you're someone you know like a serious person who's I mean is there any trepidation around that I mean when you are just well in a sense that you know you're asking to do real world things. So you're you know you're not just going you know email my wife I was delaying my you know you actually you know if it goes wrong then that you have sort of skin in the game.
Correct and I think by the way this is the thing where you know things like digital identity like what we're doing with Mover starts to become very relevant in fact I think that digital identity is going to have real product market fit with AI agents more than it's going to have with humans in the beginning because the determination of trying to figure out what is a real AI agent and whether he has authority to do what he's doing and whether he's accredited or attested to matters a lot. In fact, there's a small sort of fun little side story where I was having a communication with my chief business officer around he actually sort of had his AI send me messages and reports of some sort. But he was sending it to another Yat Siu who he had a dialogue with and sort of and Allan sort of sent messages that doesn't sound like exactly what you would say. So he messaged me separately and says, 'Have you been talking to my AI?' I said, 'No, I haven't.' And it turned out that someone else was opening up a mind that basically just called him Yat Siu, right? And the mind did not know the difference between me as a customer and the bot. And so this is exactly why you need sort of digital identity to sort of verify that, right? And so that you also know that you're talking to, you know, a bot or not, right? So that's kind of fascinating. And you know when they do like ticket bookings or you know when they trade sort of one of the very very cool things and by the way I don't think it's going to happen you know just natively to the AI agent I think really what's going to happen in the future is that people are going to build extensions for instance they're going to build sort of trading bots right that can then plug into your AI but right now the AI is able to trade by itself you can say hey you know basically buy this token when the price is this you know give market insights and in fact when you look at the skills someone else literally just posted you know when I sort of gave out sort of giveaway competition to give a little bit of sort of attention to it someone basically built a skills bazaar which the AI already has but essentially it published that skills bazaar and you can basically browse the skills and agents that are out there that are public for you to utilize so that's very cool and this by the way is a beautiful thing about open composability and blockchain because now all this stuff is out there and you can build on it and you can construct on it and it's basically sort of this open universe where everyone can just sort of build cool stuff, right? I think the other thing that you know to your question about sort of the seriousness what really really excites me is the fact that I think this is the you know we talk about decentralization and we talk about sort of you know opening up the web and making it user owned and that's still the vision right digital property rights is still such a big big deal for all of us if you really want to make it a freer web but you know even with web 3 you know we haven't really managed to sort of make that big enough yet right you know people on blockchain could be arguing outside of that system, but it's still not very big. The onchain economy is still maybe 50 60 million users. It hasn't really grown significantly, right? And there's many reasons for that, right? But I actually think you know, and the part of the reason is the distribution rails are still locked. Apple still has restrictions. If you have NFTs, for instance, it's not completely free. Google is also not completely open. Steam has technically banned it, right? So the ability to really distribute those distribution vectors are broadly still limited and restricted. However I think the entire distribution game is going to completely alter with AI agents. So when you look at for instance how we're searching basically how we're using AI we're using it like Google+. So nothing has changed we just changed the interface and but we can already see what's happening to the websites separately which is you know when you're searching on Google Gemini gives you all the summary people no longer go to the actual website they just stay on Google so great for Google bad for website and people already seeing the drops in the traffic around that right but you can see where it's going to go if you have an AI agent they're going to go and just get all the information for you and deliver and push it to you and what you're seeing now with Animoca Minds is they're coming back with valuable insights where I'm actually more like the reacting to it as opposed to asking for it necessarily because I gave it that prompt. And that's the magic that people saw with Open Claw. It's like it just did all sorts of stuff and I'm like, 'Holy what's all this?' Right? But now you see this in this sort of measured form that's coming to you with insights and information and then you react to it almost like an editor does with a piece of writing rather than actually sort of you know initiating the conversation as it were. Right and in that scenario the distribution game changes in our mind as transformative as when the smartphone came out as that kind of sort of you know distribution vector that sort of threatened the entire classical distribution space right. And why this is powerful is that I think the web as we know it will disappear as in we will no longer there'll be websites but we'll no longer go to websites everything will be you know basically drawn from MCP servers right the New York Times for instance is not going to be read sort of on a website it's going to be delivered through an AI agent who then might link to the New York Times right. And there's going to be incentives and payments that happen on the back and all of that's going to run on the token economy and because the AI agent represents your interest you become the nexus of that distribution hopefully right. And so power rests back into the user and basically web3 and blockchain runs in the back to ensure that it stays decentralized with the value construct you know for the exchange of value which I think will happen more naturally because of the sovereign nature of blockchain. Right and then that's how I think we might get into the sort of you know billion user chain adoption because basically everyone's basically using AI agents where information comes to you and the wallet part is not the most important part just to be clear right I mean I know so for a lot of people in crypto say oh it's a wallet it's for trading I That's what crypto guys think. But my mom is not going to be sort of signing up for an AI agent so she can trade crypto, right? Right. She's going to be using an AI agent to have fun or to do interesting things with it or to make it useful for her life. And the fact that it has a wallet is a feature that allows it to be useful in things, you know. Right now I know someone's basically building a Polymarket skill, right? And that's kind of cool too because instead of signing up and sort of sending money and all that kind of stuff, if an AI agent can basically sort of do something for me on my behalf on Polymarket because I say hey I want to bet on Trump winning or losing or something for instance then that's a much easier interface than sort of having to deal around with wallets and signups and all that kind of stuff.
Yeah. No, I think I mean in general but in the whole sort of economic universe and particularly in blockchain I think what agents are doing basically unbundling all the little bits. And letting you I mean I'm just restating what you're saying really and letting you sort of refocus sort of almost like broad ideas that you're thinking of and you don't have to go through any you know any of the friction. So, you know, as long as we've been in the blockchain space, everyone's argued about what should, you know, MetaMask is rubbish. I'm watching these other wallets be and it's, you know, and actually I'll call out the first person I ever saw talking about this was Gabe Leydon a couple years ago and he was just saying AI agents just before AI agents were a big thing, but he's saying that's going to solve all the walleting problems at all, you know, and that's sort of interesting because you do get the billion dollar users, but they're just not directly humans, but they're doing humans bidding. So, the vision is really sort of yeah I think we just get you know compared to vibe coding you get to you know maybe Polymarket you want to go through and choose the exact bet you want to do as a wager you want to do prediction you want to do as the entertainment or otherwise you just go I want to short Trump and just go and just you know you're a broad thing and off it goes and maybe don't even say Polymarket you just say I want to short Trump and it goes off and it just goes on lots of different ways. Exactly because you know the number of new prediction markets that are opening up right now is pretty abundant and the quantum of prediction markets sort of you know proposals that we get for investments is very very high at this moment in time. As is usually the case and of course you know the other thing you know of taking a little bit to the gaming space. I mean maybe this is also the way in which we can expect NPCs you know inside games to exist frankly more so than perhaps some people thought in terms of where they actually do have a wallet of their own and you know you can tip them or you can work with them where they can do that kind of stuff. Again and from an interface perspective a cloud version like that actually makes it more easily scalable. Like right now this is just we use for the workspace but you know one can totally imagine how you can just use a similar infrastructure and basically plug it into essentially a gaming infrastructure that essentially makes it alive as opposed to sort of you know integrating technology in a way because now it's much more easier because essentially a text-based interface where which you can you know maybe just plug in as well. So so many things that you could do with it. Really the imagination is really the limit here. And yes, I guess we should talk a little bit more specifically about games because I guess in general I mean blockchain games are sort of going through a bit of a sort of a crisis of the night or something or you know there's been the long arc of basically sort of everyone running out of money and obviously the vision of sort of play to earn and all that sort of stuff you know.
Sounds like gaming in the real world. Well, yeah. Gaming in the real world is still a $200 billion but indie game studios are struggling just the same and in fact in some cases worse, right?
But I guess the interesting bit here is that I mean I think particularly where we sort of ended up I guess where I would say at the moment I said for a while is yeah the people who really sort of still believe in blockchain gaming have tended to be the people who are more doing fully onchain type games. It's not the only thing. Obviously, there's sort of the more DGEN, casual DGEN and that sort of asymmetric risk thing, but you look at sort of, you know, Gigaverse and Eon and Frontier and all that sort of stuff. It tends to be the more fully onchain stuff. And as it happens, that sort of stuff is really sort of suited to agents because you're basically I mean if the game's been designed right and such like that basically doesn't really matter whether the human being is
Playing or an agent with some sort of crypto wallet attached to it, you know, it's providing hopefully sort of more liquidity and more activity in the game. The agent doesn't really have an advantage. I mean it might be depending on exactly what the game is. So I think of the agent inside gaming or frankly in general, right? When you talk about the agent communicating directly with you and the same is true for games as essentially the interoperability layer of sorts. And what I mean is it's not that everyone talks about bridging before and assets like whatever, really if you think about it you're interacting with the agent and you know I want to sell this asset or I want to play this game and use it to get an asset for that. It does not care that it's on this chain. It does not care basically where it sits, right? It just gets it somehow and then you can use it and the agent basically deals with that, right? And that to me I think is also one area where we can solve that friction point because when you talk about like for instance NFTs, when you think about marketplaces, right now you can trade across chains because they built it this way but again if I want to buy an NFT and my agent does it for me then I don't really care which marketplace it sits in and I don't care whether you bridge it five different ways. I just care that I have it and that I can play this game, right? And so essentially it's like just making a request and saying, you know, I want to play this game. Oh, to play this game, I need this NFT. Okay, fine. Let's go. It gets you the NFT, right? And you don't have to know which marketplace it's on and what website to even go. You just play the game because it has it there and it basically puts it in your wallet that is maybe delegated automatically and off you go, right? So I think those friction points will be solved.
In general the whole argument of blockchain gaming. You know, recently there's been a little bit of a debate, especially amongst some people in crypto where they're saying, you know, gaming was a misadventure. It's a good thing that we can just focus on the real stuff. You know, finance is what blockchain is all about, you know, and I've probably sort of stated my disagreement around that and many others have as well, right? And I think one of the issues that we have in our industry a little bit is that because of the very strong institutional focus, everyone's trying to sort of say, okay, we're very institutionally focused. I mean Animoca is doing the same thing right in that construct but as a result we're saying institutional focus means that basically it's all about money and it's all about basically just blockchain is just financial technology which I very much disagree with because it's also if you frame it in this way then only the people who care about finance are going to be in blockchain which is not true, right? And also it limits the scope of things, right? And if the bigger vision is around this new internet in which we could have a stake and particularly in the age of AI, it's so important that essentially the rails in which it uses sort of the infrastructure and the payments and the ownership of stuff is one that it belongs to you, right? As opposed to belonging to the platform, then it's much more than just a payment system and a financial system. It's one where ownership has to be a core element from that. And then the financial elements become possible because you own it. Like if you don't own anything, then essentially the financial systems will always take advantage of you. You need to be able to have agency over that. So that's super critical.
And as we have seen with NFTs and gaming and I pulled some of these numbers which was kind of interesting as well. You know the people who came into NFTs in 21 and 22 were basically the real onchain users that have actually stayed basically for the most part over the years to come, right? And many of them obviously continue to have been unfortunately hurt in the markets particularly in 25. 25 was a terrible year for consumer crypto, right? Like one of the things that sort of when people ask me was Trump good for crypto. Sort of my comment is like B minus which from an Asian perspective is pretty bad, right? But broadly speaking you could say it's still positive in the way to say that well it's been positive on regulation but it's really been bad on everything else. I mean Trump token, Melania, the tariffs, the policies in question and I think that's part of the issue which is that crypto is not Trump's top priority. There's many other things that is top priority and crypto is something that is important but it's maybe I describe it as his 10th child type of thing. You know it's in the family but it's just not the most important one. And I think too many of us have depended on sort of him saving the industry and the excitement but fundamentally it ended up being unfortunately quite the opposite in many ways and so people have been hurt from that process.
And so we haven't grown our retail crypto over the last three years. I mean that's pretty insane when you think about it, right? It's grown in centralized exchanges for people trading and speculating but for the onchain economy actually the biggest onboarding event actually was NFTs and gaming in 21/22 and it brought the kind of people into the ecosystem who weren't just there for the quick buck. Of course when NFTs started to get super hot and there was a lot of speculation then of course we got to a point where it sort of got a little bit solid that way and NFTs still trade around $300 million a month. Okay, so this isn't exactly a dead industry. It's just not the hyped industry that was before.
And what's happening now when you look at Polymarket for instance, yes, it's bringing a lot of people onto Polygon. Polymarket would have never been able to develop if it wasn't for the blockchain, at least with the speed and the reliability that it had especially in the beginning. But how many people who are using Polygon actually know that it's a blockchain product? And one might argue this is great because that's how you onboard them. But that also means that the liquidity isn't truly shared across the ecosystem because a user doesn't know anything except Polymarket, right? So in other words, he's not like, oh, okay, I just won $1,000 or I just succeeded in a big liquidity event. Now while I'm not doing anything, how about I take that money and sort of put it in a and get some yield and then make a trade or hey, there's an NFT I could buy as well or this is a game I could play. He doesn't know that because Polymarket isn't exactly going out there and advertising that because they don't want to. They want to keep them in sort of the walled garden that is Polymarket. But if you give it to an AI agent who trades on your behalf, right? And he does the same thing. He will then also say, oh, by the way, have you thought about that $10,000 that's sitting in your Polymarket account that you're not doing anything with that until you make that bet, just put it into something and get your whatever yield that is safe on stablecoins until you decide to make that next bet for instance, right? That's kind of where I think this will go.
Yeah. Yeah. And going back to a point you sort of raised earlier is it sort of for like Mocaverse network and the whole Moca thing, you know, onchain reputation and identity sort of been a big thing you've been pushing for a while and you've had some interest there but obviously that's a slow-moving sort of momentum because you need other third parties.
It's very fundamental. The problem is that it doesn't immediately translate into money right away. And so for the classic VC it's hard, that's not who it's for anyway. But as an infrastructure rail, we needed this because if we don't have this then it's very hard to build sort of trustless permissionless systems that you can trust, right? And just quickly what I was going to say is the ironic thing is that's actually really wanted by AI agents because you want to know something's an AI agent and you want to essentially, obviously AI agents don't have to have a passport or maybe something you don't want AI agents to do but equally you want AI agent to build up a reputation for what they're doing on chain as well. So actually those sort of products, lots of people have tried them over the years and nothing's really sort of quite worked.
Yes, because you say it's sort of dull, sort of like why do I need to do this? I'm not, what am I really getting? Am I going to get an airdrop? No, not so bothered then. But with agents you sort of...
