What Happens When AI, Crypto and Gaming Collide?
Wei Xie's company trains game bots on the way real players play, so the bots that fill thin multiplayer games behave more like people.
In this episode · 12 chapters
It’s a game, but the game is designed to be kind of a Trojan horse to teach people about what machine learning models actually are.
Wei Xie, CEO and co-founder of ArenaX Labs, starts from an economic argument: AI and blockchain are making labour and capital abundant, which breaks the assumption of scarcity that economic models rest on. He then asks where that leaves humans.
He explains how AI Arena, a fighting game, teaches players how machine learning models are trained, and how ARK lets studios fill thin multiplayer games with bots trained by humans. He also argues that learning a game from pixels falls short of direct integration.
Wei Xie, co-founder of ArenaX Labs, talks AI, blockchain and gaming on the Thinking On Paper podcast. He says the company's fighting game AI Arena is designed to teach players how machine learning works, while its ARC platform lets those players train human-like bots for other studios' games: “these bots are much more human like because they're trained by humans”.
Why does Wei Xie say AI and blockchain break economics?
Xie argues that AI and blockchain break the core assumption of economics: that factors of production are scarce. He says AI is making labour abundant, while crypto lets any person or group “create capital out of thin air” to fund an idea.
What are the ARC and SAI platforms from ArenaX Labs?
ArenaX Labs runs two platforms. Xie describes ARC as AI gaming infrastructure, a technology stack that powers AI-based game experiences, and SAI as a machine learning competition platform for researchers and developers. AI Arena, a fighting game, was the first game built on ARC.
How does ArenaX Labs find AI talent outside academia?
Xie says ArenaX uses games and competitions to find AI talent outside academia. AI Arena is the top of the funnel, drawing people into AI through play; SAI is the bottom, built to “reveal hidden talent” through competition rather than master's and PhD credentials. He points to the pay packages big companies offer AI researchers as evidence of a talent shortage.
How does ARC help studios with player liquidity?
ARC helps studios solve player liquidity, meaning games that suffer when there are not enough human players for a match. Instead of the scripted bots players find terrible, ARC lets trained players build bots for a studio's game. Xie says “these bots are much more human like because they're trained by humans”, and player data can later make them predictive and adjust difficulty for each player.
Why does ARC not train bots on screen recordings?
Xie says training bots from screen recordings does not work well. ARC uses a direct integration: the studio supplies the game's state space (the variables that shape outcomes) and its action set (the controls), and the agent learns from real play. A model trained on pixels, he says, is like “a kid that learns things through memorization” and fails when a situation changes.
Where Mark and Xie disagree
Mark asks why, if AI talent is scarce, a new AI application seems to appear every day: “Or am I being deceived by my echo chamber?” Xie says many are low-effort “reskins” built on foundational models such as GPT, and that the newer model versions tend to absorb them.
Links from this episode
Questions
Who is Wei Xie?
Wei Xie, CEO and co-founder of ArenaX Labs, starts from an economic argument: AI and blockchain are making labour and capital abundant, which breaks the assumption of scarcity that economic models rest on.
What is AI Arena?
He says the company's fighting game AI Arena is designed to teach players how machine learning works, while its ARC platform lets those players train human-like bots for other studios' games: “these bots are much more human like because they're trained by humans”.
What does ArenaX Labs do?
ArenaX Labs runs two platforms. Xie describes ARC as AI gaming infrastructure, a technology stack that powers AI-based game experiences, and SAI as a machine learning competition platform for researchers and developers.
Transcript
Jeremy Gilbertson Curious minds, welcome to another episode of Thinking on Paper. My name is Jeremy, this is Mark. We're talking about AI and gaming today. We're excited to dive into this thing. We're not going to do any rants on the top end of this thing. We are going to jump right into the conversation because we have a very special guest today. Mark, give us a quick intro. Wei Ji, who is the CEO and co-founder of ArenaX Labs, who are making AI Arena, which is incorporating AI blockchain into a new gaming mechanism.
It's all about new gameplay mechanics, imitation learning. It's very, very cutting edge, and we're looking forward to hearing more. Welcome to the show, Wei. Great to be here, guys. Really looking forward to this conversation. Should we carry on where we left off, Jeremy, I think? Speaking of technology, we actually had a wonderful conversation for the last eight minutes, and that conversation was bumped off the stream for some reason.
We now know where we're headed, and we've had an extended pre-chat. We talked a little bit about, let's just summarize this for everybody. We talked about in our pre-chat the idea of legacy, because Barry Lunn, our last guest, his carryover question was, what is your legacy? We talked about impact and landing in an impact, and we're at a spot where you especially aren't ready to commit and define what that actually means in specifics, but everything you're doing is pointing to helping people figure this stuff out and make an impact.
In that, let's talk about some of the dots, because a lot of this stuff is about connecting the dots, about figuring different things out. That's what we do on Thinking on Paper, a diverse group of technologies coming together and trying to figure out what it means. Let's talk about a couple of the dots that you may have connected to create what you are building today. For sure. Maybe I'll start on just a notion of impact.
The punchline
Wei Xie I think the punchline there is, with our company, ArenaX, really the underlying principle and philosophy is human empowerment in the age of rapid technological advancement, primarily driven by two massively disruptive technological themes, which is artificial intelligence and Web3 and blockchain technology. Maybe starting there, I'll give a more truncated version this time, but I think one of the series of dots that I'm starting to put together and think through is, with the emergence of AI and blockchain technology, I think what's happening is the concept and the notion of economics as a construct is radically changing.
The reason why it's changing is because if you look at traditional economics, there are basically what economics are, is the allocation of scarce resources into production of productive means as a way to create productivity or economic growth. And traditionally, in these models, there are a few different types of factors of production, human labor, land, capital, and more recently, it's like this catch-all bucket of technology.
Within all that, though, there's one critical assumption that underpins all of this, and it's the assumption that these factors of production are scarce. Why is that important? Because all economic models are built off of this assumption. And what we're entering into a world is this actual break in this assumption where factors of production may no longer be scarce. And I don't think it's a stretch of the imagination for people to start to kind of almost like visually and viscerally start to feel this, because with artificial intelligence, where we're trending is labor is no longer going to be scarce.
It's going to be radically abundant. And it's going to be for most tasks that humans have historically been able to to do will be very, very cheap and very, very easy to replicate by AI. And then on the concept of capital itself, historically, that in terms of the concept of money, credit, and what you can do to actually finance growth, that was assumed to be scarce. And very quickly, we're starting to see that that is in and of itself not scarce either.
Now, you can look at the official channels of monetary policy and fiscal policy of central banks and governments. But if you look at crypto and blockchain, as a standalone entity, what it has created is basically for any person, group of people or entity to basically create capital out of thin air, and use it as a way to bootstrap some type of growth behind an idea. So with these two things, we're creating a new type of economic paradigm that is radically different than how the world used to work.
And then one other input that was historically a big assumption into this broader schema is humans, right, through the channel of labor. And if things like technology is starting to basically proliferate factors of production, so that is no longer scarce, where does that leave humans in terms of our role in this future, as these two kind of technologies emerge?
So I think, if you if you look at the public discourse around something like artificial intelligence, it always tends to come with this, this perspective of potentially, like, a very dystopian view of what could emerge out of this, right, which is displacement of human labor, you know, what is it that humans are able to going to be able to do, when AIs are going to be able to do everything that we we can do, and even better, what what becomes of the human race in the in the age of superintelligence, right, there's this, like, kind of general exploration of where, where does human fit into this emerging picture.
And I think we have a generally more optimistic view on that outcome. But it's incumbent on, you know, companies, projects, and groups of people to explore how we can maximize kind of human endeavor and input in this new age of radical abundance. I think that's, you know, in a way that like an abstract way that we're approaching building up different products, and, you know, why we're building things like AI arena, RSI, etc.
Where the human fits
Mark Fielding And where does the human fit in the age of superintelligence is a question that we need to discuss more. Okay, so that's the the background of how you're thinking about this. So could you, for our listeners, explain what is AI arena? And how, because I believe it's for two different, you have the gaming side of it, and you have the AI model development side of it. And they are both separate, separate entities, but part of AI arena, with the same goal and goal in mind.
Wei Xie Yeah, so let me let me kind of demystify that a little bit. So the company is ArenaX. We actually have two, you can call it platforms. One platform is called ARK. ARK is a an AI gaming infrastructure. So ARK is a technological stack that can power AI based game experiences. And I'll come back to that in a second. The other vertical is called SAI. SAI is actually a machine learning competition platform.
SAI is more focused on a sophisticated AI machine learning kind of researcher slash developer cohort is really designed to be an environment where the very sophisticated kind of researchers can go on and actually compete to create new types of AI or machine learning models. So so SAI is rather specific to that audience. ARK is more generalized for application into the gaming context. AI arena is a game. AI arena is, it was the first game that was built on ARK.
So basically, as we were building AI arena, we were building ARK, because the underlying tech that needed to power a game like AI arena did not exist. And to this day, probably still doesn't really exist. So we had to build everything on our own. And as a function of building AI arena, we've also matured tech stack that supports it, which is called ARK.
ARK
Wei Xie And now today, we're able to use ARK as a infrastructure product that we could start to distribute and help third party studios integrate AI based experiences into their games, and also use the tools that are available to solve a lot of the issues and problems and challenges with kind of game development for especially like indie studios. So hopefully, that helps to clarify a little bit what the businesses are.
And then, so what is it that we're trying to achieve, right? Where's this like human empowerment angle come in? Underlying both of these things is the concept of like, how do we find human capital, human talent that can create outsized impact in the age of AI? Part of it is an education funnel. The other part is a discovery funnel, right? So think about AI arena as the education funnel. It's a game, but the game is designed to be kind of a Trojan horse to teach people about what machine learning models actually are.
How do you train them? How do you make them better? What are the trade offs that you're managing? And through playing a game, you can actually build up a very strong intuition base about how artificial intelligence works. Okay? So it's not just gamifying things. It's actually using a game as an apparatus, as a process to teach people about something that's a bit more complex. That's the idea behind AI arena, the game.
Psy is more about, if AI arena is the top of the funnel, which is how can we use a game to attract more and more people into the field of AI, then Psy is more at the bottom of the funnel, which is, well, we already know there are people that are sophisticated in AI. How do we identify more and more of them? Because currently there is a challenge in terms of AI talent.
The talent war
Wei Xie If you look at the talent war in terms of large companies bidding for AI talent and how much they're paying to actually secure the best AI talent on the planet, the compensation packages that are being paid to these individuals is a strong market signal that there's a supply deficit of talent. We don't have enough people in the world that really can build solutions that can further the progress of AI and humanity.
So Psy is designed to be a platform to reveal hidden talent, because traditionally a lot of discovery is through formalized channels like academia. And basically right now it's academia. Masters and PhD programs will give you the credibility to be able to secure these type of roles. But we think there's a different way that we can come around and solve that problem, which is to use competition as a way to reveal talent.
And this is a lot more of a meritocratic process in the sense that you really do have to know what you're doing and you are competing against, hopefully, the best of the best in coming up with provable performance. And so that's really the idea behind Psy. So all of this is kind of integrated. AI is like top of the funnel. Psy is a bit more bottom of the funnel. But it's all about cultivating interest of human capital into AI and then identifying who are the best that can contribute to future impact.
Jeremy Gilbertson Yeah, where my head goes on this, on the talent side of the fence, I think the future is going to point to a couple of different things. Very specialized technological talent, the ability to machine learning, LLMs, AI, blockchain, all of these different technologies. But then also on the other side, I think you're going to have an increase in the humanities. So philosophy, psychology, sociology, some of these things.
And I see those two as highly, highly increasing as being the popular ones. And then the rest of them kind of being functions of either or as solutions come together to handle marketing solutions that come together to handle business operations and all of these things that you go normally. What do you think about that? That is a very plausible outcome. I think the end state is that if and when we get to a, probably more when rather than if, when we get to a stage of super intelligence, it would be hard to imagine many of what we consider to be like careers and jobs today to still be as relevant.
Wei Xie So I think the extrapolation of that is humans should have more idle capacity. And if we have more idle capacity, what do we dedicate that time to? And that could be more exploration of things like philosophy and the arts and creative domains, but more from, not from like a mass production perspective, more from like an artisanal craft kind of perspective. So I do see that as like a potential outcome.
And then on the other end, at least from now until the point we become completely obsolete is yes, you do need, you do need more human engineering talent to build these solutions that continues to scale technology in that direction. So in the process of getting there, I do think like this class of human capital is going to be extremely coveted. And I think there's will be a constant supply problem.
I think, I believe like Elon probably said recently earlier in the year, it's just like, we need more, we need more of these engineers. And it's very, very difficult to hire them at high volume, but on the demand side, there's so many problems that still needs to be for a lot of these, you know, applications of AI to come to fruition. So yeah, I'd probably share that. On the supply side of the talent, when you're saying that it makes sense to me.
Mark Fielding And I think maybe blockchain really good, a level blockchain coders are short on supply as well. But when you look, especially when I look on my feed in my echo chamber of AI applications, there just seems to be a new one every day. Everyone's adding AI to their tech stack, everyone's launching an AI product. If there's a lack of talent, why are there so many applications appearing? Or am I being deceived by my echo chamber?
Is that actually not many?
Wei Xie Is that actually not really that many? Or are they just using other people's AI technology? Yeah, that my personal view is a big part of it is, is ascribed to, basically, it's like low effort application building on core foundational models like chat, dbt, and these LLMs and generative models, right?
And what you see is like this rapid creative destruction type of cycle in AI, where, you know, when GPT three came out, you know, there were, you know, I think at that point, people were releasing things that they're called, they call them like AGI, or agents that it was like a convoluted patchwork of different API's built on LLMs. And they were like, Oh, these are like autonomous agents that can now do all these tasks, right?
Fast forward, like, nine months, it's like, well, GPT four can do that out of the box. You know what I mean? And it's like, it's like, yes, there, it seems like there's a proliferation of volume of applications, but they're all kind of like reskins of a core foundational capability that are being provided by these like source models.
It's not a value judgment to say that's either good or bad. It just what is right. This is a market economy and people see opportunity. And what happened in this, like last couple of years is that I think foundational models like LLMs was so impactful in creating an apparatus for the consumer to, for the first time feel what AI is able to do, right?
That was so impactful that it drew so much attention into this pocket of AI, that investment, consumption dollars, all of those things were converging on this at the same time, which created this effect of, well, where there's money, there's opportunity, and when there's opportunity, people just like, you know, try out a bunch of different things. So there was an explosion in that, in that sphere.
And I think that's where you kind of feel the overwhelming, like this feeling of, wow, there's so much happening. But I think a lot of that was, yeah, more of like a application layer that really didn't have any kind of barriers to entry. And it was, it was at risk of basically the companies or the agencies that own these foundational models, they were going to basically absorb that through subsequent releases.
And, you know, informed people were all saying that they're like these capabilities and these apps, like the next version of GPT was, and basically, it would be available out of the box. And lo and behold, this is where we are, right. And now we're like with GPT-4 is like multi-modal. You know, there's so many different things that you're able to do with it now than just language based applications, right.
So, so I think that will continue as an arc. But what I'm saying is like the engineering, the engineering talent that needs to go into researching and refining the GPT engine so that it can get to four, five, six, seven, eight, nine, 10. And onwards, that level of engineering is in short supply. And it maybe is not in short supply in absolute numbers, but in relative terms is certainly short relative to the demand, which is what can we build with all of this, right?
Jeremy Gilbertson Like, there's so many more things that we can, we can do to, quote, unquote, come up with new Let's see. Okay. So, I get this guy, pull him in, I say, hey, I want you to help me build an AI model to help analyze why this soccer, why this corner kick guy is so good.
Wei Xie And then you come at it from that side. I mean, I can see how this like whole world could like spin up in trying to figure this end result out. Yeah. Yeah. Well, I would say today, if someone is able to come up with a diagnostic AI to diagnose why other people's AI are so good, like that person deserved to be hired by like, that's a non-trivial problem to solve. And it would require a lot of skill.
Isn't that AGI?
Wei Xie But isn't that AGI, isn't that when your super intelligence will be doing that when it will be learning off other AI agents. So we get out of the loop again. Yes, theoretically, but I'm hoping that we're still maybe a decade away from that at least. But like, so today in the current moment, I think it's still very much a human based engineering effort. But yeah, I mean, like from that perspective, like as a thought experiment, certainly plausible.
But to your point, Mark, of like, can I then list my NFT on a market and sell it for a profit? Yes. Theoretically, you can. And practically, mechanically, you can. Then for that to happen is you need people to care about what you have done. Right. So it may be a super niche if you said this is a pure competition about like a corner kick agent. Like we might like it may be a market of three people.
So therefore, really, there's no price discovery for these assets. So if you're but but if you look at it in the context of something like AI Arena, which is like a fighting game, there is a broader community of people who do enjoy the medium of fighting. That is a target market. And if you can get them to care about something like this now, now, like each individual asset as contained by an NFT will have some market value.
Jeremy Gilbertson Right. And so, yeah. So the idea is like, how do you take these kind of competitive primitives and create a problem around it, make it fun and then scale audience into it? So to continue our thought experiment, you know, pushing beyond the boundaries of reasonable discourse, maybe. But, you know, say, all right. So say EA Sports in their new FIFA game is getting a lot of terrible feedback that their corner kicks are not realistic looking and they're all dumpy and they're like, man, this all their fans are like, that sucks.
This part of the game is awful
Wei Xie This part of the game is awful. They start peeking in and looking at Marco and hey, you know, if his thing figures out how to make corner kicks realistic in video games, I might buy that thing. Right. Yeah. So. OK, so Arc as a platform. Is designed to basically help studios solve these types of problems in games is not exactly how you described it, but philosophically is kind of similar. So one of the service verticals of Arc is called we call it permanent player liquidity as a service.
What we do is we actually help games build human like bots to help with this concept of player liquidity and what player liquidity is, is basically the fact that games suffer when there's not enough players in it. Right. And for a lot of types of game, like especially multiplayer games and competitive games, you have this degradation of a user experience as a function of the game not having enough human players.
You have to wait very long time for there to be a match. And sometimes what studios do is they do actually program their own bots in the game to kind of fill in the gaps, if you will, knowing that there's not enough players. So you need to have something in there to keep people busy. But then real human players, when you play against those bots, you're like, oh, my God, these bots are terrible.
Right. So it actually detracts from the experience. What we're able to do with Arc is we're actually able to utilize our player base who's actually sophisticated in this type of AI training game loop. And as long as soon as a studio integrates their game into the Arc SDK or the platform, they get access to our player trainer platform and the trainers are able to now train bots to be used in those games.
Bots that are more human like
Wei Xie Now, these bots are much more human like because they're trained by humans. Right. So now these games actually have human like bots that improve player liquidity, makes the game experience better. And really, the bottom line is that it carries over and impacts player retention and really enhances the overall quality of the experience that these studios are trying to deliver. I just had a flashback of when I was 18. I was playing GoldenEye on the N64 and just how incredibly stupid the NPCs were. And I would literally just walk in circles. And yeah, it wasn't much fun.
Jeremy Gilbertson And I know it's progressed a lot since then. But yes, there's still a huge problem with the NPCs. And I started thinking about the idea of a digital space, almost like an orange theory fitness on the fitness side, where I could send my bots to get programmed to be good bots, like good NPCs to be folded into a game. I just see like this. Yeah. Yeah, that's effectively one of the yeah, that's effectively the first product on ARK.
Wei Xie And it's really starting to generate a lot of interest from studios because a lot of studios have this problem. So and it's exactly that is making either NPCs or kind of opponents in the games much more human like so they're more interesting to play against. And then there's like steps beyond that that we can add to the effect later on, which is once you have that as a primitive, then you can you can actually collect more and more of like player data in that particular game with that data.
A stat sheet
Wei Xie Now you have what is effectively like a statistical distribution of what users are likely to do as a next step in the game. And what you're able to do is use that information and embed that into how you train these bots so that these bots are now predictive. They can actually anticipate what you're going to do next and you can actually calibrate the difficulty for the individual player. Right.
It's almost like it's almost like it's like a very bespoke way of training someone to become really good at something. Right. Like if you're in any type of like combat sport or any sport for that matter, part of the part of the weight, like part of the friction of improving is a lot of times like you're not working on that marginal skill that can really kind of build a solid foundation for you to get to the next level.
So if you can start to make that more granular and more precise, it actually helps people progress in terms of skill development more efficiently. And in a gaming context, it's just it makes it makes like the game experience overall feel way more bespoke, way more personalized. And ultimately it should be much more engaging. I love that. Have you read The Art of Learning, Jeremy? Josh Waitzkin, he talks about that granularity of the learning of how he became the open hands champion was by doing exactly that, by just focusing on this one move.
Jeremy Gilbertson Hundreds and hundreds and hundreds of times and then getting that perfect and then to the next one, that's how he became the champion. This sounds like it's a similar thing. Would you would you mind, Wei, entertaining one more thought experiment for me? Of course, of course. All right. So, all right. Let's go back to EA Sports. Let's go back to FIFA World Cup or whoever runs that thing, right?
You're not letting that go. No, I'm not. I'm not. So, so say Mbappé decides to license his physical maneuvering likeness in a soccer game to EA. So he goes in, they put sensors on him, they volumetrically capture how he does a corner kick. And that particular likeness, just like any other name, image and likeness, has a monetary value. Could that be captured? Could that be converted into an AI model that would teach other things to do that thing exactly digitally?
Wei Xie Yeah, I'm a little bit out of my depth in answering that precise question. I'm out of my depth in asking. Although I have to admit, Jeremy, like as a big lacrosse fan, I'm very impressed by your knowledge of football and Mbappé corners. Wow. I will say this, I will say this. OK, so I'm going to try to answer the question without directly answering it, but hopefully it gives you an idea of how we approach this type of problem.
How ARK works as a platform
Wei Xie So the way that Arc works as a platform when we integrate with a game is it's called a direct integration. So what happens is the game studio actually has to provide us some information about how the game works. The technical term is about the state space. The state space is really just the environment of the game, factors that influence outcome. Think about those variables that impact how a player can extract different outcomes in the game.
And then there's the other thing, which is the action of the action set, which is really like what are the controls of the game and what these controls actually affect in the environment or how this player behaves. Once we have that information, we can basically represent any type of game on our SDK. And then the training is basically once you have that framework, a player, all they have to do is they literally just play the game and then the agent is recording that data set and is able to and then you're able to use that data set to basically train the agent on how to progressively improve in this kind of game context.
So that's how we do it. Well, you said so it's yeah, so it's performance in the game as someone is actually playing the game, it records that and instead of like exporting from physical to digital. Yes, yes. Now, the issue with the export framework is that it's an imprecise representation of what's actually happening. Right. So a very crude way of doing that would just be recording pixel level information on a screen.
Right. You can like I'm watching YouTube of someone playing a first shooter game. I can record that information and theoretically you can say, oh, I just train an AI model based on that visual input and then have that model learn through that visual input and how to behave in the game. The problem with that is. There are there are a lot of different edge cases. Number one, visual representation does not give you full information of the actual state.
Right. It's only one dimensionality view of what's actually happening. Number two, there's a lot of information that's actually not captured by pixels and there's a lot of edge cases in terms of what you think you're watching relative to what the human actually is doing and their intentions like those are there's a there's a big gap there.
So it's a long winded way of saying like a visual extrapolation of the game state is imperfect and you're going to run into scalability challenges because what effectively your model learns is like it learns useful behaviors if it was confronted with that exact situation. Anytime it's like a kid that learns things through memorization and they don't have critical thinking abilities to extrapolate.
The last question
Wei Xie So as soon as you get to the last question on the test, they can't get it right. It's like, wait a minute. It's like it's not two plus three equals five. Like it's like a software accident. They're like, I don't know how to do this. You know what I mean? So it's like it's like a similar outcome in terms of how a model would be able to perform in a dynamic environment.
So direct integration, which is what Arc is based upon, is much more robust, is much more generalizable to to the precise controls and the nature of how you want to manipulate these models. OK, I think that's a good place to drop some some final thoughts, and I'm going to start because as I alluded to,
Mark Fielding I'm not sure if you got this before a little tech problem, but I alluded to research in this interview. I wasn't sure where to begin because there was so much going on with ArenaX Labs and AI Arena. But now I think that Waze has done an amazing job of kind of threading it all together. We started with the educational part of it, of how the game can be used to to to teach AI, to get the good developers excited into doing this and get more of them bringing in.
And then we moved on to the gaming aspect of it for the gamers. So it is a game in there. And so I think it was and then brought together at the end, I thought it was very impressive. So now I get it. So Waze, my model has learned that we need to have a part two on this because there's so much that you're doing that's really exciting, that I'm really interested in, that I would love to kind of stay in touch and maybe as you guys are building things, come back and talk through some of these great examples.
Wei Xie I mean, I think it's I think it's fascinating. I think I think everyone else will find it pretty darn fascinating as well. Yeah, absolutely. Love to be back. And I think maybe this time we'll also bring Brandon, who, you know, I'm more of like a poser and a mouthpiece. I'm not the engineering mind behind all of this. Brandon is the engineering mind that builds everything that we have on the tech side with the help of our very talented engineering team.
But yeah, I'd love to come back and revisit where we are, maybe a little bit, maybe early next year or something like that to to to have, you know, an extension of this conversation, because, yeah, it's been it's been a lot of fun for me as well. So I really appreciate you guys bringing us on. Last question for you, Wei, what is one question you will leave for our next guest? It can be about any topic, any discipline, anything that's on your mind.
A question for the next guest
Wei Xie Doesn't have to be A.I. gaming, whatever pops out. Three, two, one, go. Yeah, this is really difficult, I would say, what what what is the what is the one thing that makes you kind of get get up in the morning and just want to attack the day? Like, what is that thing that just gets you going in the morning that, you know, sets you on that trajectory to have a really good day? It's a great question.
Jeremy Gilbertson Great question there. Yeah. There you have it, friends and neighbors. We are we are sponsored by Synthesis today. The the forcibly interconnecting of dots of information into knowledge and eventually wisdom. Mark, take us out of here. Well, I just will say that question will be asked to come. Who is the the French country manager for Coinbase who we've got next week? So we're going to be talking about the European perspective on on crypto.
Mark Fielding So there is a good a good thread here. So that's going to be interesting. If you're still listening, please like and subscribe so we can get more amazing guests like Wei and join our book club where we'll be reading Nexus by Yuval Noah Harari next. Yeah. Stay disruptive. Be curious. Keep thinking on paper. We'll see you next week. Bye bye.




