Inside the Wild World of "AI Agent Traders", and What That Means for the Rest Of Us, w/ PIP CEO Saad Naja

how fast are AI agents moving from a novelty to a reality? When could AI agents be better than humans at certain tasks? What happens when judgment calls become fully outsourced? These are right now burning questions in the world of AI and business. But today we are going deep on this with a concrete, very real, hyper specific case study, day trading, as in day trading stocks and how some intriguing new experiments reveal that AI agents might be better at this than humans in what that means for us. My name is Jeff Wilson. I'm a journalist. I'm an AI strategist in the host of AI curious. So maybe you've traded stocks or crypto before and maybe you haven't. I'm guessing, frankly, most listeners have not and have no interest in spending a lot of time on trading stocks. That is not a requirement for this episode. And I think this topic is worthy of exploring for one key reason. We hear a lot of how AI agents might transform how all of us live in work. I have talked about it quite a bit on the show before, but 99% of the time that is still pretty abstract, hard to visualize, and it's out there in the distant future. Well, finance in the world of day trading falls in that 1%. It is here. It is here now. Thanks to the cold objective reality of market data, stock price history, market trends, trading winners and losers, finance is in many ways the perfect data driven sandbox for AI agents experimentation. This is not widely discussed outside of trading circles, but there is quietly a grand experiment at play. Can AI agents manage money better than a human? Are they better at big picture strategy? Can they sniff out trends before they happen? Can they make fewer mistakes? Can they stick to a plan better because they're less emotional? And if the answer is yes, what does that mean for the rest of us as AI agents infiltrate whatever industry you work in? Now, I am delighted today to invite someone who is at the tip of the spear of this AI trading revolution. Saad Naja, the founder of a company called PIP World. Now, Saad has done something really interesting. He's created a system of trading swarms where you have what amounts to an AI portfolio manager, more or less. And then you have different AI agent traders, a swarm of these traders and analysts who kind of report to the manager and then you, the human, can pick different portfolio managers to oversee your accounts. There are over 300,000 human users and testers of this. And currently it's all still in simulation, but they plan to go live with real money soon. We'll can go wrong, right? So why should listeners care? And why should non traders care? If you don't trade, like so what? And the reason is this same manager agents plus specialist AI swarm framework is eventually perhaps coming to sales, marketing, ops, hiring, accounting, forecasting, and just about every industry accepts good hurting. And here now is where I have a personal confession, even disclosure. I know the world of day trading better than you might expect. For a few years, I was deep in the closet as an aspiring day trader. I spent hours every morning nerding out, analyzing charts and buying and selling stocks, often for as little as two minutes trying to beat the market. I spent countless hours watching educational videos, reading books, crunchy numbers, joining trading groups. And the truth is that I loved it. It was really fun. On Sunday nights, I looked forward to Monday morning for the stock market to open, but I also learned it's really, really, really hard. It's almost impossible to beat the market. Study after study has shown that most day traders fail. And I was one of them. Now I didn't lose a ton of money or anything. These were usually 10 or 20 bucks a trade, mainly to see if I could be profitable as an experiment. And I almost got there. I had some very good stretches, then bad stretches, then good stretches, and so on. But ultimately, there was something that made me quit once and for all. AI. As generative AI got better and better, it hit me that even if I could prove to be profitable, the odds were zero that in the future, I'd be able to compete with these super powered lightning quick AI traders. I would be out guns. Well, that world is now happening. It's here today. Again, you can think of this almost as a petri dish, if you will, for agents in a larger economy. If you want to understand what autonomous agents will look like in the real world, finance is arguably the first place to watch because it keeps score in real time. And now, back to sod. I first met sod in person at the AI summit I produced in a hosted earlier this year in Toronto at the consensus conference. Now, both pip world and consensus have the roots in crypto, web, three, and blockchain. But in many ways, that makes perfect sense. The crypto markets, broadly speaking, have less regulation and red tape in the trading platforms of fidelity and Charles Schwab. It's often called the Wild West. So it totally makes sense. That would be the first place for AI trading experimentation. And I have you this conversation as a glimpse into the future of more mainstream and traditional finance and perhaps a glimpse into how the wider economy can shift for better or worse. So with that, please enjoy my conversation with sod Naja, the founder of pip world. Sod, welcome to AI curious. How are you doing, Jeff? I am doing great, man. Super excited to dive in with you. We spoke briefly before going on stage earlier this year at the AI summit at consensus. I was like, oh, I wish I had an hour to pick his brain on AI and trading. So now I do. So now you are stuck with me. So looking forward to diving into all of this. Not a lot, but it's absolutely pleasure to be here. All right. So let us start with a baseline. Before we get into precisely what you've built in the kind of AI analysts and traders, they're already doing things as we speak. How would you describe the current state of play of how AI has already changed the world of investing and trading? So AI, I mean, really proliferated into the trading and investing world probably in the early 2000s. When you look at kind of HFT funds, Renaissance technologies with the millennium fund, which is probably the best performing fund in the world, they were one of the first to really utilize AI in the world of trading and investing from a high frequency standpoint. By that, do you mean what many conversationally referred to as kind of quant trading? The idea that like if you're, if humans are, if you're saying cool, I'm going to buy and sell shares of Apple. If it hits this price, I'm going to buy this or sell this kind of trying to make a like scalping, make a profit within a few minutes. There are quote-quant traders doing that at like every micro seconds. There's like millions and millions or billions of transactions and they are even doing that for decades. Yeah, absolutely. So I mean, the whole context of AI that you know, has really taken a hold over the last few years. Everyone has seen that as something very new and revolutionary and in the context of chat GPT and being able to actually converse with a model, it's completely new. But in terms of AI technology, it's all predictive modeling and predictive learning, which has been around for yeah, at least two decades. It's not more. All of these big HFT funds, all of the big real algorithmic funds, they've all been utilizing AI in different ways for their predictive models. For the models that are constantly bringing in hundreds of terabytes of data, crunching all those numbers within milliseconds and then outputting some form of an executable trade. So I mean, it's really been exclusive to the big high frequency firms in terms of the millenniums, the jumps and the big boys. Whereas all of the retail traders, all of the mid-sized traders and all of the layman guys, we've never had access to any kind of technology like that. It's really been reserved for the big powerhouses. So I mean, it's been around for a while just using completely different contexts to how the normal guy sees AI and what that looks like. As you point out, AI is nothing new for most of the Wall Street bait trading firms. So what has changed in the past few years as Gen AI has risen to prominence? I'm guessing even these big firms have found new use cases and new applications for generative AI. And to your point also, I'm sure we'll get into this. Retail traders now have access to that same type of technology. So what's shifted in the past few years? Yeah, so I mean, from an AI perspective, AI of great context collectors, right? They're great at pulling in information, they're great at synthesizing information. And now we've really reached an inflection point where agents actually are able to act, they're not able to just give advice and crunch numbers and data. They actually have the capability to be able to execute, which is something that wasn't really happening before. Now, when you look at how, like you said, the big firms, how they utilizing AI now, you don't have to go far, right? You've got countless amounts of investment banks and I know people personally who've been laid off recently from an analyst level because the reality is chat GPT and these different models out there now have the capability of filing S1s, analyzing cash flow reports, breaking down PE statements, at incredible levels. Work that used to take a human weeks or an analyst weeks to do can now be done in the matter of hours with the right kind of tools and the right prompting from different machines. So, it's a completely different ball game in terms of giving you that edge that, you know, what was reserved for kind of a special elite class of people is now being a lot more open source than a lot more democratized because you can now upload a cash flow statement to pretty much any LLM model and ask it to give you what's good, what's bad, what it looks like from a forecasting perspective. That completely changes the game as to how people can analyze reports. You don't have to have the most mathematical mind anymore. It's completely changing that landscape of people being able to actually break down analyzing companies. So, let's get into then what is different and possible now in the retail trader worlds that was not before. And I'm also curious how like AI trader, AI analyst is different from kind of an old school trading bot or quant, right? You mentioned that they could do things but like in some ways like trading bots could do things, right? Like we, to your point, for years and years, you have these massive firms where the, basically, AI traders, we wouldn't call them that, but they were this quants programs are physically executing trades, right? So, it's not saying it's not telling analysts, hey, you should buy this serve Exxon and they keyed it. It's all done, right? So, what are the ways that these new breed of AI traders are different from a kind of more quote unquote traditional trading bot or trading quants? I'll give you the, my number one example here is when you look at algo trading, it's very simplistic, right? You set the variables and the algo will execute. When the RSI crosses this threshold and price hits this, you execute a buy or sell. That algo will do that. I already have that, been around for years and years and we understand completely how that works. Now, what that doesn't do and the problem with algo trading is it's very one-dimensional, right? It doesn't take into account anything else from a wider horizon perspective or anything else from a broader perspective. The what AI allows you to do from a technology standpoint is essentially really build out a coordinated trading desk, whereas one facet of your strategy may be, hey, when the RSI hits this and is overbought, then you should either buy or sell very, very simplistic. Now, by actually being able to utilize AI swarms and multi-agent systems, you're able to consume much more data and actually synthesize it in a completely different way. So, what that really looks like is being able to replicate a trading desk with just one agent. Now, that means you pick your strategy, so whatever it may be, when this particular event happens, you should do X, Y or Z, so you should sell or buy. But, you can also take into account, hey, what does the sentiment also look like? Is the sentiment really overbought? Really oversold? Is there a lot of greed? Is it max fear? So, you have the sentiment being also taken into consideration. Hey, okay, well, let's also look at the macro news. Do you have J-Power speaking? Do you have a big risk off news event coming in where you probably don't want to be in a trade because it's going to be very volatile? All of these different pieces of information are now able to be brought in, and that allows the agent to actually not just be fixated on one particular variable, and this is why algorithms really lose because they don't have the context of what's happening in the world, right? They're not able to come in and say, oh yeah, okay, well, this particular pattern has been hit, but hold on, Jay Powell is speaking in three minutes and the market is selling off, so maybe we shouldn't execute the trade, or sentiment is really in the gutter at the minute, and China's just banned Bitcoin for that medium of time. Maybe we should actually wait before jumping into the market or jumping into a position. That's what it really allows you to do. It's five agents, one coordinated decision, and really have an audit trail as to why you've taken that particular trade from a viable self-respective. Interesting. It's a good segue into this swarm agents that you've built. Minor standing is that you have one kind of AI manager super agents that's like the head of this trading desk, and then you have these, then they almost adhere she, they are kind of coordinating this swarm of agents that report into them in a way, or one might be focused on the micro. Okay, we're looking just at the performance of this one asset. Maybe it's Bitcoin, maybe it is Apple stock, whatever it is, and they are going to meet laser focused on the technical analysis of Bitcoin. Then you have some other agents that I'm making this up, let me know where I'm getting off track, that you have, they are looking more at like the incoming news of global events, and they're looking at a, what's what is maybe even like, are there any weather forecast changes that would impact things? Is there a crash by Boeing that would have some surprising impacts on travel stocks? Maybe that actually sinks the broader S&P, and because the S&P is sinking, there's a sell-off in crypto, and think things like that, right? And then you have some other agents who is just crunching numbers of reports, and there's earnings release, and this agent is just going non-stop, crunching, and so you have the super agents that's kind of approximating human judgments that also I'm guessing has to have visibility and insight to what the actual human wants at the portfolio and what the risk tolerance is, right? Do you want, is an aggressive portfolio as a higher risk appetite, or is it super conservative? Like, whoa, well, it's just like nice steady S&G goes, and then this team of agents is working together to maximize return. Am I, am I warm here? Let me know where I'm, where I'm getting ready when you're on. You're cooking, you're cooking, Jeff, it's genuinely, yeah, spot on. I think you laid it out perfectly, right? Specialization is key, and the beauty of a multi-agent system is you can have specialization with each specific analyst. That's what the analyst has to focus on, it's one particular task of, hey, I want you to pull in as much sentiment data as you can from Twitter, and really tell me whether it's what the sentiment looks like from a fear and greed perspective. Hey, I really want you to analyze the macro policy in China, and tell me what's coming out from a news perspective. Hey, I really want you to analyze on-chain movement. Have there been any big transfers into an exchange? Have there been any big, what does the coin base premium look like? All of these big things are massively executable into smaller tasks, and by having specialization, it means each particular agent knows its role, understands its focus, you have far less hallucination, and that reports back up into a chief investment officer who will synthesize all of those reports and say, hey, we're going to buy cell or hold a particular asset. And again, like you said, based on the risk tolerance perspectives of the actual owner. This is so fascinating. I have so many follow-up questions, but let's give a sense of, for listeners, the Twitter scent you've already deployed this, and how widely is it being used, and I think it's important to this is not just speculation, not a whiteboard idea. You guys are actually doing it. Yeah, absolutely. So I mean, we've really been deep diving into AI and AI trading for the last 18 months now, I would say. The massive, I would say, moat that we have in the USP that we have is we've really been able to train these LLMs on 27 years of historical, great proprietary data, which we've been able to pull from one of our main backers, Exynity. Now, all of that data shows us exactly what the best traders we've actually one have done, and what the worst traders we've actually lost have done. And just from an avoidance, you're welcome. You're welcome, sad. I have personally helped train the worst trader bucket of data. I have, I have, absolutely. There is a single trader going, who hasn't fallen to like bucket, at least some point in their life. So yeah, don't show too bad, but we've, yeah, I mean, we've really had the ability to completely teach these small language models and build up on proprietary models into what's good and what's bad from a trading behaviour perspective, and what kind of edge can that give you? Now, having modeled that, you know, for the last 18 months and really gone into this whole refinement and optimization period, we got to a stage where the models were winning, I would say, close to about 48, 47% of the time, but because they have such big risk to award ratios, even at a 30% win rate, they were still in their profitable, because the beauty of the agents are, their risk is always capped, right? They follow risk tolerance practices. They follow the best practice. And the reality is, most humans don't actually suck as traders. A lot of the ideas are actually right. And again, the data showed this, right? A lot of people's initial instincts were right in regards to buying a particular asset. The problem was their execution. They either went into heavy, or they timed it a little bit too wrong, they didn't have a wide enough stock loss or gap, and that's where they actually ended up fading. It wasn't that, hey, their thesis was completely wrong. It was actually just poor execution on their standpoint, and had they actually executed this practice, people would be significantly more profitable. Sarah Sun, just a quick clarifying note for folks who might not be as immersed in this world. The way trading often works is you might say, hey, I'm going to risk, let's call it $10, but if the trade works, I'm going to make $20, right? And so, and if you're disappointed of that rule, so that'd be a good, well, there's a two to one reward to risk ratio, right? Risk is 10 to get 20, which is why to your point, even if they're only winning 48% of the time, if they are getting 20 bucks for everyone, losing 10 bucks for every loss, as a massive victory. How did you test this? Like, how did you kind of, what was the environment you used to, to, to gauge that performance? Were you testing against real life, were they putting into the actual wilds, or was a kind of simulator? And how, what do you, what do you think, can it just look like there? Because I'm or even, are they legally allowed to trade? Like, how did that all work out? Great question. So, in terms of actually testing it, the first element is always backtesting, but backtesting can only paint you a small picture. We all know, past performance is not indicative of future, future behavior. So, backtesting really gives you a picture of how these models were performed in different scenarios and different kind of areas of market dynamic should say, because the market's got three different phases, as we know, sometimes it's very momentum driven, sometimes it's very choppy, each kind of market phase is different, and behavior needs to actually be able to adapt to that. So, that's one element, then the same way that way, everywhere, everyone likes, you take the agents, you actually give them some capital and you start deploying it. Now, that could be with real world, so we actually started first on running it as simulated. So, all real world pricing, all real world, all the books, and modeling what the performance looked like in real time. And then from there, we've actually tried different elements of actually plugging the agents into different executable platforms to see what that looks like. I mean, if you look at today, we are the fastest growing AI trading platform in the world by a long shot. We've been live for 20 days and I've already broken 300,000 users. And the best part about this is, we launched it in a gamified simulated environment because we want people to actually see the results. We don't want to hide anything, we don't want to promise numbers, we want you to go in, build a trading team, and actually see what these agents would have returned for you, had they been managing your money and your own portfolio. And that's the beauty of a sandbox environment that, hey, I have no capital at risk. I'm speaking to a front team of traders, deploying them to the markets and letting them trade the assets that I want. And then seeing what trades they took, I can see, hey, you know, this particular agent bought Bitcoin at 85K and sold it at 90. His stock loss was INT2K. He made this much. This is why he took the trades. Being able to see and verify all of that information in real time gives everyone, there's, you can't hide behind that. We're not, there is no more, it's all black and white for everyone to see. And the beauty of this, what really blew, I mean, even us away, because the reality is we, you know, as much as we've trained these models, as much as we're going through this process, we're still learning. I mean, these AI is in terms of the self-reinforcement loops and how they pick and engage new data is amazing. And the most recent one was the crash on, when was it? 11, 11, or 10, 11, the big finance crash. We actually had, I think, eight out of ten agents began shooting into that crash and made an absolute huge win through. And again, I was sleeping at the time that I think it was midnight or 1 a.m. in Dubai, but I woke up around 7 a.m. and was just blown away by what the results looked like, because yeah, I personally would have been buying into that crash, knowing what's going on, thinking, hey, everything's on sale along you. The agents, Europe is a crypto, for folks who're not following Bitcoin, Bitcoin crashed hard in November. And you're saying these, and whereas the old school trading bots might have had more of like a very narrowly focused, looking just at price action, they might have made a mistake of like actually scooping up more, or potentially, right? And having a very narrow kind of scope and would have got slaughtered for at least in that time frame, but this, these agents analysts put a little more holistic view of the market, decided instead to short and ended up profiting handsomely while the rest of humans were losing money. Now, was that all, is the three first off, congratulations on the explosive growth out of the gate, 300,000 users in 20 days, or is that all still in the simulator, or are these being deployed for real capital at this point? So that's also still in the simulator, we haven't moved on to actually giving real capital, and that will be early next year, we will move on to real world trading and facilitate uses actually being able to have and customize their own agents, and then actually deploy cash with them. The first stage is just for people to actually gauge the technology, get more customer of it, and we've had people who are now poppy trading, got agents, and then poppy come out on Twitter, and actually done very well for themselves. So yeah, I might have to start charging people commission on some of these things. Absolutely, yeah. So what types of assets are they able to trade now, and then what have you learned so far? What are some ways that these agents, or AI traders, are different from humans? On top of you already mentioned, they can be disciplines, right? Whereas humans might break their rules, AI will follow the rules. On top of that, what is surprised you about the way AI is trading so far? I mean, when they have a particular thesis about assets, it's made up of very, very different ways that we would interpret something. So like, I mean, humans naturally are pretty contrary to you saying the sense that, hey, when the price is going down, we should be buying. We should be buying interviews. Or when the price is massively going up, we should be selling. We should be selling it. Agents don't do that. They actually are far more patient in, hey, once the price has actually gone down to one, let's say, whatever it may be, it's broken 100, that's more of an indicator for us to actually sell into it. They wait, even though they're not going to get as much of a bigger win, they wait for confirmation before actually executing the trade. And that in itself gives them the ability to really be a lot more patient. For us, we want to really maximize any kind of position that we're going to take. They don't do that. They'll happily give up some of the win, but just wait for more confirmation that the trade is going to go their way, which again was massively insightful to see. And then, when I really look at the broader pictures to how agents have actually really learned and really kind of adapted over time, it's been amazing to actually understand that different agents are better in different market cycles. There is no real one-fittal narrative, right? You have certain agents that really focus on the breakouts where, hey, the price has just broken out of a particular range. That's their hotspot. They'll be continuously trend following. They'll just, can you elaborate a bit? I don't think I understand well enough how each agent is different or build different links on imagining. It sounds like it's not as simple as like, if you were to take just off the shelf, chat GPT and say, hey, one conversation on chat GPT, I want you to focus on trading this kind of stock and when do a reversal. So when if the price is going down for a while, you think it's about to go up, buy anything that to go right or a reversal or this other agent's breakout, as you say, has been kind of price has been kind of consolidating a low for a while and we think based on variables x, y and t, it's not primes for this massive breakout to go up. I would think that it's like the same chat GPT or whatever it is would be equally dept to all of these things, but it sounds like they're more kind of like, finally trained and had the most personalities and striking weaknesses. Can you elaborate on kind of how that works and how they're built? Yeah, absolutely. And you're completely right. I mean, everyone goes with the whole, hey, chat GPT, make me a really a percent a year and do it trading these stocks. It's like, yeah, if that, if life was that easy, we'd all be well-buffet, but it doesn't work that way. The way the real speciality here is in the context, right? And like I said before, the markets go through different phases. They never just continuously up, they never continuously down and they all go through different peaks and troughs. We know that for a fact. Different agent, and this is what we learned over the kind of 18 months of different agents specialized in performance for different kind of market dynamics. Now, you mentioned it very much earlier on where you talked about support and resistance. What does that really look like? It means, hey, at a particular price, there seems to be a lot of buying pressure. So the price is really going to break below here. Hey, at this particular price, there seems to be a lot of selling pressure. So price is really going to rarely going to go above here. Now, that's a range, right? And most kind of prices will trend to actually trade within that range. Now, what happens when you break, break out of a range? Hey, we've just broken through a big level. Is it going to go back to that level? Is it going to continue going up? This is where you have different agent personalities and different agent styles that really focus on different trading dynamics, you see? Thumb will really look at, hey, we've just broken out of a very big key range. We should be hitting this stock aggressively because the pullbacks are going to be shallow, which means that once we're out of that particular range, price tends to go one particular way. Those agents do really, really well when the markets are doing that in particular. So big example of that is COVID, for example, when the COVID crash happened twice with one way, the break out was to them was downwards and it kept dropping. Another great example is when President Trump was elected. When President Trump was elected, markets were extremely bullish and we went from, it was around 67K on Bitcoin all the way up to 100. Price broke out of a particular range and carried on going up. Then you have different agents that really focus on trading levels. Now, what does that mean? It means, hey, once we actually hit support or hit resistance, we should take the opposite trade because price isn't most likely going to bounce off that particular level. So when markets are ranging, like Bitcoin kind of is now between 90K, shall we say, 110. It means in those particular ranges, they do very well because they're just trading from key levels to key level. And then you obviously have the ability to really look at very simple stuff like mean reversion. Hey, once we drifted away from a standard deviation of mean by XML, that's going to come back at a certain stage back to mean because we know mathematically it always does. We should be focusing on mean reversion. And what people fail to understand and why so many people are still so far behind in the AI spaces when it comes to trading is there is no one-size-fits-all. There is no one agent, hey, make me a hundred percent a year go do your job good luck. Doesn't work like that. There will always need to be some form of human involvement. It's just going to be at a far more shallow level. And that human involvement actually will help to shift kind of what does that strategy look like? Hey, are we bullish? Are we bearish? Are we more focused on a range? And this is where it's like different agents at different jobs are kind of think of it like you're the football manager, right? And you're picking which particular player or trader you want to put in the market at this particular point? I think a lot of listeners might be wondering, okay this all sounds great, all sounds very cool, but I know from experience that AI makes mistakes. Like every time I use some chatbot to do research for me, it will like make up some fact that is patently false, right? So it's one thing when you're doing something when you're doing some research on civil civil war and it makes up some battle that ever happened, like okay, that's a problem, but not a big problem, perhaps as you think you're losing your money. Exactly. Oh, I mean to risk $100 and it actually risk $10,000 or losing the all the money, right? So how have you worked to mitigate the risk of hallucinations and more broadly AI going sideways? Yeah, so I mean this really falls into that context window of not overloading agents with way too much information and building out complex kind of information systems. It's really about segregating different elements of risk and different data collection points. So that like I said, specialization is really key here and what that whole reinforcement look like are really really key here because it allows the agents themselves specifically to know what their job is and stay focused on their particular job. Now that really, really reduces the impact of hallucinations considerably and one of the biggest challenges we had early on was that, like you said agents do tend to go a bit crazy and start doing crazy stuff on their own, it's like yeah, that can happen when you're doing your course work with research for a presentation. It can't happen once I've trusted you with you know $100,000 of your world money. So no, I mean we have really picked in huge amounts of guard rolls and circuit breakers into the agents and have multiple checkpoints at kind of different levels to ensure that loads are followed. We haven't had any agents go rogue, which is good. Don't get me wrong, there are agents that can lose money for sure, but we haven't had any of that. Yeah, it's all still being pretty controlled and far less exciting actually. One thing I am fascinated by is a question of, okay, what's the end game here? And I mentioned up top that I was personally more than dabbling in day trading for quite some time for over over two years, every morning I do defle, multiple monitors out and looked at all the price charts and all that stuff. Eventually I realized, okay, even if I get to a point where I can do this consistently profitably, my gut was that that there is no guarantee that will continue in the future as AI gets better and better at this stuff, right? And like if I am able to say, if I'm able to look at this price chart and say, oops, my rules are that the when price hits resistance at this level and also the RSI is here and also this variable is here. Even if that makes me profitable, eventually AI can do it too. And it seems like sod your swarm agents are at this point at least proof that yeah, that's happening and that will happen in the future that AI will do a really dang good job with this. So what's the end game here? How does anyone have an edge if my Annel AI trader is doing its thing, your AI trader is doing its thing, every listener of his podcast, their AI traders are doing their thing. Presumably we'll get to a point where all the models will be the model arms race and they're all going to they're all going to be very good. How does anyone have an edge? How is this not just like a just AI on AI and AI and it's all a wash, all a zero sum game? So at the minute, it's PPP, everyone knows that it's always been player versus player in terms of trader versus trader. That will shift from a landscape perspective to agent versus agent. Now like you write what he just said, what does that look like? If my agent versus your agent, who's going to perform better, then essentially you've got one layer deep right, which is what what data is your agent consuming? What data is your agent trained on? Well mine is significantly better. We've got a lot more data. We've got a much richer data set. We've got far better data workflows and she learning patterns that give us the edge and that's where from an agent perspective, it will then fall down to what it currently does, which is who has the better data, who has the better information. And like I said before, it's just because you have an agent, and I have an agent, there are different market dynamics, right, which is, is it your agent that's being put into the market at the right time or this particular agent, which is being put into the market at the right time? Is it the breakout one? Is it the range trading one? Is it the mean reversion one? Each one of these are separate. There is no just one size fits all where, hey, we all deploy our single agent and we sit back in and just watch the magic happen. It's going to go a layer deeper, there's going to be different elements of people being able to compete, but you're right in terms of when it does become agent tick wars, I like to call agent tick wars, what that happened, the margins forever become smaller, which means like I said, the cops aren't as bad. They're not going to be as, because they are now from a human level where humans go and, you know, un proportionately risk all of their capital and end up going on till and, you know, lose all of their money from the trading account perspective. Agents won't do that, but the margins forever will be smaller. It means that each little piece of information, each little piece of data, each agent in terms of what it's trained on, is going to impact that overall performance and not return. Just like any business, right, your margins get smaller and smaller. So, I mean, it will change for sure and the world of finance and trading is so big. It's a, we're a long way away from there for sure, but there's going to be a complete landscape reshuffle. People have no idea where it's coming, right, from Wall Street to the retail trader who's currently trading on Robert Herd or whatever it may be, that entire dynamic is going to change. It doesn't matter about people think AI is chat GPT and publicity and so on. That's the beginning. We've only seen 5% of what that really looks like. What's going to come in the next two, three, four, five years is going to completely reshape finance, trading, investing as we know it because the reality is humans aren't as capable as these lovely autonomous spots which can actually think for themselves, put in information for themselves and synthesize it and come to a decision. So, yeah, I mean, it will be amazing to see different platforms being offering their own different types of agents and what that actually looks like for them. And this is why we've actually had so much uptake, even with people. People think that trading brokerages and exchanges that want users to lose and hey, they're out and they're trying to steal our money. It's not true. The reality is an exchange or brokerage wants you to survive and live as long as you possibly can because if you're trading on that, you're making the fees. That's a great point and you just gave for a final question for you and it's a great segue. You just provided here. Give us a prediction of how well this world you mentioned, people are not prepared for what's coming and how the world is changing. So give us a tease of what the world could look like in your choice of timeframe, two years, five years of how the world's trading could look differently in the future. I can categorically say within five years, I would say 80%, 90% of trading will be agentic. It will not be humans trading and putting out their phones and aping into different positions on different brokerages. It will all move agentic and humans will get removed from that equation more and more. The only reality is that we as humans are not psychologically the most disciplined creatures and that's our biggest fault when it comes to trading. The results show that it's very black and white. It's not devotable. We suck at being able to maintain discipline and we suck at being able to really control our attitude when it comes to dull premiering especially when it comes to making or losing money and you described it yourself in terms of the personal journey you went through and I did the same and right now I would happily wage a hundred thousand dollars against any retail trader to come and over three months go against one of my AI trading agents and let's see who wins. I'm very categorically happy that agents can continuously and dramatically outperform humans. The question is how quickly and how you're going to embrace it at some stage it's just are you going to be the latter part of the audience that does or at the beginning because that's going to make a big difference to your portfolio when it comes to what the results look like. Well, I look forward to my agent traders going against your agent traders in the future. I'll get inside what you guys are building is very interesting. Thanks so much. Really enjoyed it. That's it looks you guys. Take care Jeff. Appreciate the chat. Well, there you have it. Thanks again to Saad Naja, the founder of PIP World and thank you to you dear listener. If this is your first time at AI Curious please consider subscribing, rate it five stars, send to a friend, send to your favorite AI agent trader and they can send into their swarms. We'll have a great time together. Thanks again. See you next time.