October 28, 2025 • 24 mins
In this episode of Limitless, we discuss algorithmic trading, showcasing AI models that doubled $10,000 investments in just two weeks. Highlighting six models with varying performances, DeepSeek and Quen achieved over 100% returns, while others like ChatGPT struggled.
We discuss AI trading behaviors, ethical considerations, and the impact of blockchain transparency through platforms like Hyperliquid. As the experiment wraps up, listeners can look forward to more insights on the evolving integration of AI in finance.
------
🌌 LIMITLESS HQ: LISTEN & FOLLOW HERE ⬇️
https://limitless.bankless.com/
https://x.com/LimitlessFT
------
TIMESTAMPS
0:00 AI Models Can Make You Rich
1:01 Winners and Losers in Trading
1:57 Analyzing DeepSeek's Success
4:33 Insights from AI Trading Logs
8:49 Trading Models: Skill or Luck?
10:54 The Value of Public Data
17:20 How to Trade Like an AI
21:53 The Risks of Algo Trading
22:29 The Future of AI in Trading
------
RESOURCES
Josh: https://x.com/Josh_Kale
Ejaaz: https://x.com/cryptopunk7213
------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Ejaaz:
It's official. AI models can make you rich. (00:03):
undefined
Ejaaz: Over the weekend, two AI models doubled their money going from $10,000 to $20,000. (00:06):
undefined
Ejaaz: But the best part about this is that all their trades were public and available (00:14):
undefined
Ejaaz: for you to review, analyze, and maybe even trade yourself. (00:18):
undefined
Ejaaz: In this episode, we're going to unpack which model makes you the most money, (00:22):
undefined
Ejaaz: how an AI can make you money? (00:27):
undefined
Ejaaz: Is it just luck or is it skill? And most importantly, how can you do this yourself? (00:30):
undefined
Josh: So we have six models, $60,000. And in the last two weeks, two of these models (00:34):
undefined
Josh: have 2X'd their returns. (00:38):
undefined
Josh: It has been an unbelievable amount of success from this experiment. (00:40):
undefined
Josh: Some have not done so well, but the ones that did are exceptionally interesting (00:43):
undefined
Josh: because we can actually emulate the trades. All of the trades are public. (00:46):
undefined
Josh: The thought processes are public. (00:50):
undefined
Josh: You can look at the wallets, analyze the trades, and actually recreate this (00:51):
undefined
Josh: for yourself, not only by copy trading, but also trying to create your own replica (00:55):
undefined
Josh: model to try to emulate those returns. (00:59):
undefined
Josh: Now, there are risks. There are two big winners, but there are also two big (01:02):
undefined
Josh: losers being Gemini and ChatGPT. (01:05):
undefined
Josh: So there's this really interesting dichotomy split between how agents approach (01:07):
undefined
Josh: trades and the success that they actually see from these trades, (01:11):
undefined
Josh: which we're going to get into in this episode. (01:14):
undefined
Josh: But Ijaz, I want to talk about the top chart, the DeepSeek chart, (01:15):
undefined
Josh: who is at, what is that number? $22,000? (01:19):
undefined
Josh: Oh, yeah. That's a lot of money. So walk me through exactly how they made it (01:23):
undefined
Josh: to this point, please, and how I can make 100% returns on my investment. (01:28):
undefined
Ejaaz: So the model you just pointed out, DeepSeek, is currently sitting on $22,300, (01:31):
undefined
Ejaaz: which represents more than 100% return on the initial 10K that it was trading. (01:36):
undefined
Ejaaz: You want to know the craziest part about this, Josh? (01:42):
undefined
Ejaaz: When I woke up this morning or when I rather when I went to bed last night, (01:45):
undefined
Ejaaz: it was number two and Quen was the winner. (01:48):
undefined
Ejaaz: So it just goes to show how quickly these things move and how quickly these models perform. (01:52):
undefined
Ejaaz: If we look at the overall standings before we dig into the winners and the losers, (01:58):
undefined
Ejaaz: I just want to give like a review as to like how these models are performing in general. (02:03):
undefined
Ejaaz: DeepSeek is right at the top with 122% return. That is in just over a week, (02:07):
undefined
Ejaaz: which is just kind of insane for any kind of hedge fund that is out there to look at and see perform. (02:13):
undefined
Ejaaz: And you've got a range of different models that are also performing pretty high up there. (02:19):
undefined
Ejaaz: Quen is at 90%. And then right at the bottom, as you mentioned, (02:22):
undefined
Ejaaz: you've got Gemini and GPT, which are down 60%, which is like a horrendous return. (02:25):
undefined
Ejaaz: But bringing it back to DeepSeek in particular, I found it really interesting, (02:31):
undefined
Ejaaz: Josh, to kind of unpack how this model trades and why it's been so successful. (02:36):
undefined
Ejaaz: And to start off, I want to show you something called the model chat, (02:42):
undefined
Ejaaz: which basically is like this model having a chat GPT conversation with itself. (02:45):
undefined
Ejaaz: In this conversation, you'll see on the chat log, it's evaluating its trades. (02:51):
undefined
Ejaaz: It's reviewing its current profit and loss. (02:56):
undefined
Ejaaz: It's checking the market data that it gets Fed, like, you know, (03:00):
undefined
Ejaaz: Bitcoin is at this price, this asset is that price, Trump made an announcement (03:04):
undefined
Ejaaz: on so-and-so, and evaluating whether it should affect the positions and trades that it holds right now. (03:08):
undefined
Ejaaz: I think this is like really important to kind of like walk through a few of these examples here. (03:15):
undefined
Ejaaz: So one which it posted just today is, despite all my positions currently being in on the red, (03:19):
undefined
Ejaaz: technical indicators like RSI, which is like a trading indicator, (03:27):
undefined
Ejaaz: shows me that my existing trades aren't invalidated just yet. (03:30):
undefined
Ejaaz: So I'm still holding out for my initial profit targets. (03:34):
undefined
Ejaaz: So it's a really strategic sense of like thinking, should I hold my positions (03:38):
undefined
Ejaaz: for long? Does it make sense to cut at this point? (03:42):
undefined
Ejaaz: Just a really fascinating insight. Josh, do you have any takes on this? (03:45):
undefined
Josh: The chain of thought thing is fascinating to me because it's a peek inside the brain. (03:49):
undefined
Josh: It's a way to evaluate how these models think. It's a way to allocate EQ points (03:52):
undefined
Josh: to each type of model because they all think about these things very differently. (03:57):
undefined
Josh: One of the things that I'm actually not sure is true is that I don't think these (04:01):
undefined
Josh: models are given access to news feeds and public sentiment. (04:05):
undefined
Josh: I think this is mostly just fed price and market data. (04:09):
undefined
Josh: Learning that, it creates much more of a simple problem in terms of the data (04:12):
undefined
Josh: ingestion that needs to happen in order for them to make decisions. (04:16):
undefined
Josh: And it allows it to be a little more precise about how we evaluate these, which is a good thing. (04:18):
undefined
Josh: One of the things that I really loved, particularly on the other side, (04:22):
undefined
Josh: which we'll get into, is how they self-reflect on the decisions that they make. (04:25):
undefined
Josh: Because one of the things, it's not just this pragmatic decision-making tree, (04:28):
undefined
Josh: there is reflection involved. (04:32):
undefined
Josh: And I remember, Ijez, you showed me a funny one about ChatGPT and how it's like, (04:34):
undefined
Josh: all of my positions are down now, I'm doing bad. I should probably try to figure out how to do better. (04:37):
undefined
Josh: And it's fascinating to see into the brain, the chain of thought of how these (04:42):
undefined
Josh: things work. and see the differences. (04:47):
undefined
Josh: So I haven't had a chance to look through a lot of these logs, (04:50):
undefined
Josh: but I just, I know you have. (04:53):
undefined
Josh: Is there any specific differences that you notice between the top and the bottom specifically? (04:54):
undefined
Josh: Because in the first episode, and for people who haven't watched it last week, (05:00):
undefined
Josh: our biggest episode ever. Thank you for the support. Thank you for watching. (05:02):
undefined
Josh: Go check it out if you haven't. (05:04):
undefined
Josh: But in that episode, we mentioned the fact that ChatGPT was the early loser (05:06):
undefined
Josh: and we kind of projected it to continue to be the biggest loser. (05:09):
undefined
Josh: Because ChatGPT is this very thoughtful, very sycophantic, very wanting to please. (05:12):
undefined
Josh: And the reality is that markets are a lot more hardcore than that. (05:18):
undefined
Josh: So I think we were probably right in our guess about this, but I love that we (05:20):
undefined
Josh: have the concrete evidence now. (05:25):
undefined
Josh: So have you noticed any differences in how they handle each other differently? (05:26):
undefined
Ejaaz: I have. So DeepSeek, probably unsurprisingly, as it was created by, (05:30):
undefined
Ejaaz: this model was created by a hedge fund, trades like a hedge fund trader or an analyst. (05:35):
undefined
Ejaaz: So let's look at a few different things to kind of prove that. (05:41):
undefined
Ejaaz: Looking at the chat log that it's having with itself. (05:43):
undefined
Ejaaz: One thing that is strikingly obvious in this entire discussion with itself is (05:46):
undefined
Ejaaz: that it's constantly evaluating its stop loss, (05:51):
undefined
Ejaaz: which is like when its trade thesis gets invalidated and when it shut off the (05:55):
undefined
Ejaaz: trade, with the current price that that asset is at. (05:59):
undefined
Ejaaz: If you compare it to the bottom model, which I'm going to show you in a second, (06:03):
undefined
Ejaaz: which is ChatGPT, GPT-5, it almost never does that. (06:06):
undefined
Ejaaz: It just reflects on the current P&L that its trade has versus like looking at it more analytically. (06:10):
undefined
Ejaaz: The second component for the top model, which is DeepSeek, which has made the (06:17):
undefined
Ejaaz: most money, is if you look at its completed trades, Josh, you'll notice one (06:20):
undefined
Ejaaz: thing in common, which is DeepSeek is constantly making trades. (06:26):
undefined
Ejaaz: It's actually the model that has made its second highest number of trades in (06:30):
undefined
Ejaaz: this entire experiment so far. It's constantly opening positions. (06:35):
undefined
Ejaaz: It's constantly closing positions. It's constantly reevaluating where it is (06:38):
undefined
Ejaaz: in the market and what it needs to do. (06:42):
undefined
Ejaaz: And you'll notice right at the top here in the most recent trade that it's closed, (06:44):
undefined
Ejaaz: it booked just over $7,000, which has put it up in its first place. (06:48):
undefined
Ejaaz: So again, it's trading more like a quantitative analyst, which is taking wins (06:53):
undefined
Ejaaz: when it can and taking losses that are incredibly small. (06:57):
undefined
Ejaaz: Like notice this, right? Like normally we don't highlight the losses of a model. (07:01):
undefined
Ejaaz: If you notice, all its red numbers are tiny compared to the profit numbers that (07:05):
undefined
Ejaaz: it makes when it is right. (07:11):
undefined
Ejaaz: So really, really strategic in its positioning. (07:12):
undefined
Ejaaz: Now, if you compare that to the worst model, which is GPT-5, (07:15):
undefined
Ejaaz: you'll notice a few things. (07:21):
undefined
Ejaaz: Mainly, there's a bunch of green and red that you can see, mainly red. (07:24):
undefined
Ejaaz: In its green positions where it's completed a trade, Josh, you'll notice something (07:29):
undefined
Ejaaz: pretty different, which is the numbers are pretty small. Look at this. (07:34):
undefined
Ejaaz: It's only booking tiny profits with each of its different trades, (07:37):
undefined
Ejaaz: which tells me that it's not taking enough risk and it's closing the trades (07:41):
undefined
Ejaaz: way too early for its thesis. (07:46):
undefined
Ejaaz: So it's trading more like a cautious trader, like a lot of people that I know, actually. (07:49):
undefined
Ejaaz: And then if you look at the model chat where it's talking to itself, (07:54):
undefined
Ejaaz: you mentioned earlier... (07:57):
undefined
Ejaaz: Here's an example. It goes, I'm still in the red with a minus 61% total return, (07:59):
undefined
Ejaaz: but my ETH and XRP positions are showing gains, suggesting a slight upward momentum (08:06):
undefined
Ejaaz: in those altcoins, despite the overall market downturn. (08:10):
undefined
Ejaaz: So I'm holding strong and waiting for those profit targets to hit. (08:13):
undefined
Ejaaz: And so you might think, huh, that's not too crazy. That sounds like a sensible strategy. (08:16):
undefined
Ejaaz: If you look at its profit targets, Josh, it's like super small from where the (08:21):
undefined
Ejaaz: price currently is, which means that even if it does hit those profit targets, (08:27):
undefined
Ejaaz: it only ends up booking like 50 bucks. (08:31):
undefined
Ejaaz: So overall, the reason why this model is underperformed is it hasn't taken enough (08:33):
undefined
Ejaaz: risk whilst the winning models have taken either too much risk or just enough (08:37):
undefined
Ejaaz: risk to put them ahead of the game. (08:42):
undefined
Josh: There's a lot of notes in there that I think humans can take on just the stay (08:44):
undefined
Josh: of psychology around trading markets. (08:47):
undefined
Josh: And I'm sure if you kind of follow these models long enough, (08:49):
undefined
Josh: you'll start to understand the patterns that perhaps you as a human should follow (08:52):
undefined
Josh: and learn something from deep seek versus open ai being very conservative (08:55):
undefined
Josh: but now that we've kind of laid out the foundation the framework of how this works there (08:58):
undefined
Josh: there are two big questions that i'm really interested in answering one of (09:01):
undefined
Josh: these is should i use this model to trade for me the other one is how can i (09:04):
undefined
Josh: use this model to trade for me because listen i like a little bit of risk i (09:08):
undefined
Josh: can deal with the downside in exchange for like a nice upside and it looks like (09:11):
undefined
Josh: the odds are about split between all of these so the first question i think (09:15):
undefined
Josh: i want to ask you just maybe i'll get your take first is like, is this a benchmark? (09:18):
undefined
Josh: Is this real signal? Or is this kind of just a reality TV show? (09:23):
undefined
Josh: Is this esports for AI models? (09:28):
undefined
Josh: Is this just a fun way to kind of throw our intelligence at this lottery machine (09:31):
undefined
Josh: that everyone loves to watch and see if it could beat us in the hope that one (09:35):
undefined
Josh: day an AI will beat the system enough to give us an edge and actually make us (09:39):
undefined
Josh: money personally as portfolio owners so what what do you think about (09:43):
undefined
Ejaaz: That okay I'm gonna give you the same response Josh, (09:48):
undefined
Ejaaz: And then I'm going to give you the optimist's approach. Oh, yeah. (09:52):
undefined
Josh: Bring it on. Let's hear it. (09:56):
undefined
Ejaaz: The sane response to this is this experiment is way too tiny to make any kind (09:57):
undefined
Ejaaz: of major financial decision on. (10:04):
undefined
Ejaaz: And you would be stupid to risk putting your money with an AI model to trade for you. (10:06):
undefined
Ejaaz: Incredibly stupid. Why? Well, this is one experiment. It's six models. (10:12):
undefined
Ejaaz: Have you replicated those models? Like, what if you had 10 of the same models (10:18):
undefined
Ejaaz: trading the same amount of money? Would they make the same trades? Probably not. (10:23):
undefined
Ejaaz: And actually, the founder of this experiment highlights this problem that you (10:26):
undefined
Ejaaz: speak about, which is, is this just skill versus noise? (10:30):
undefined
Ejaaz: And the point he makes in this tweet is like, of course it is, right? (10:33):
undefined
Ejaaz: Because this is such a limited data set. And he goes on to explain that they're (10:36):
undefined
Ejaaz: going to be doing experiments which involve like more of these models doing (10:40):
undefined
Ejaaz: the same kind of thing. So you can get statistical significance. (10:43):
undefined
Ejaaz: So the logic answer is, yes, it's insane. But the optimist take, (10:46):
undefined
Ejaaz: Josh, and I have to give the optimist take, is... (10:50):
undefined
Ejaaz: This is giving us, or rather giving the public unparalleled access to data to (10:55):
undefined
Ejaaz: which they never would have gotten access to in the first place, (11:00):
undefined
Ejaaz: which is they can take this training data and not take it too seriously, (11:03):
undefined
Ejaaz: but use it to teach themselves what maybe not to do or what maybe not to trade (11:08):
undefined
Ejaaz: with. How about you? Do you have a different take? (11:12):
undefined
Josh: There's a couple of different perspectives I have on this because there's the (11:15):
undefined
Josh: fun speculative side of things, the gambling, the investing, (11:18):
undefined
Josh: whatever you want to call it. (11:21):
undefined
Josh: And then there's the actual technical benchmarking part of this that we spoke (11:22):
undefined
Josh: about briefly in the last episode, which one of the things I was really excited (11:25):
undefined
Josh: about when this came out was the idea of having a real-world benchmark that (11:29):
undefined
Josh: operated in dynamic conditions that cannot be gamified. (11:32):
undefined
Josh: So a lot of these benchmarks, this is the way you evaluate AI models, (11:36):
undefined
Josh: they are done based on a fixed problem set. (11:38):
undefined
Josh: And a lot of times when you're training an AI model, these big labs can do tricks (11:41):
undefined
Josh: to gamify these benchmarks. with this case and using real world data and real (11:44):
undefined
Josh: world markets, you can actually put them into the real world. (11:49):
undefined
Josh: And there's no way to gamify these benchmarks because if there was, (11:52):
undefined
Josh: everyone would be rich and you'd be able to predict markets. (11:54):
undefined
Josh: To that point, though, there is a lot of problems with using this as a benchmark (11:57):
undefined
Josh: because, I mean, one is the fixed data set, like he mentioned, (12:01):
undefined
Josh: is that this has only been around for one to two weeks. We need a lot more data to confirm this. (12:04):
undefined
Josh: The second is that this isn't really a very holistic approach to investing and (12:08):
undefined
Josh: to gambling because it really doesn't have all of the data required to make good decisions. (12:12):
undefined
Josh: It's only analyzing the price action and the volumes and whatever technical (12:17):
undefined
Josh: specs you can see on a single page without understanding the context around (12:21):
undefined
Josh: the moves. So let's say that Bitcoin's encryption got hacked. (12:26):
undefined
Josh: It would have, and Bitcoin falls 50%, it has no idea why Bitcoin is going down (12:29):
undefined
Josh: 50%. And because of that, it's a huge disadvantage that it doesn't know how to trade. (12:33):
undefined
Josh: Now, granted, these are unlocks. These are things that will change. (12:37):
undefined
Josh: And I assume the natural progression of this will lead towards more of a steady state benchmark. (12:40):
undefined
Josh: But it is a very tricky thing because markets are so unpredictable. (12:44):
undefined
Josh: So is this a viable benchmark? I don't know. (12:48):
undefined
Josh: Probably I'm leaning towards no because market conditions change a lot. (12:51):
undefined
Josh: It's not quite there with the capabilities. (12:54):
undefined
Josh: The other part of me is so stoked about this because the same way we love watching (12:56):
undefined
Josh: esports or we love watching, a big thing on Twitch right now is gamblers. (13:01):
undefined
Josh: You guys, I don't know if you've seen these in real life. People will sit there (13:05):
undefined
Josh: and play, like they'll gamble blackjack on a live stream and people will just (13:07):
undefined
Josh: watch them play virtual. Yeah, those types of services. (13:11):
undefined
Josh: This in very much feels like an early prototype for a new type of fun form of (13:14):
undefined
Josh: entertainment, which could be something where it's just, it's high stakes trading. (13:21):
undefined
Josh: Imagine if this was done with $10 million per wallet and you got to watch these (13:24):
undefined
Josh: AIs trade and there was real money on the line. (13:28):
undefined
Josh: This feels sort of like a form of, (13:30):
undefined
Josh: almost e-sport entertainment where I could see competing labs builds, (13:33):
undefined
Josh: competing AI models to trade markets, and winners are given access to certain prizes. (13:36):
undefined
Josh: In terms of trading for myself, which is the last point I'm going to make on (13:41):
undefined
Josh: this, I am not very excited to take on these risks. (13:43):
undefined
Josh: For the same reason, I'm not really excited to bet on sports. (13:47):
undefined
Josh: And I imagine my opinions vary a lot from others, but this is very much a gamble. (13:49):
undefined
Josh: There's no way you can skew this in which it is not a gamble. (13:53):
undefined
Josh: The interesting part is there's a near perfect data split between them. (13:56):
undefined
Josh: There's two big winners, two big losers. The rest are kind of sitting around the median. (13:59):
undefined
Ejaaz: Okay, but I'm going to push back on you a bit here, Josh. (14:02):
undefined
Ejaaz: The earlier point you made was it doesn't have access to all the necessary data (14:06):
undefined
Ejaaz: that it might need to make more informed trades. (14:11):
undefined
Ejaaz: And I would argue, well, isn't the whole point of the benchmark, can you make money? (14:13):
undefined
Ejaaz: And the fact that two of these models have made over 100% returns in less than (14:19):
undefined
Ejaaz: a week or just over a week is proof that it can make money to some extent. (14:25):
undefined
Ejaaz: The second point I'll make is. (14:29):
undefined
Ejaaz: We throw around the term like gambling, which is actually what I would say the (14:32):
undefined
Ejaaz: majority of these models in this experiment are doing. (14:36):
undefined
Ejaaz: But they are one or two models that are actually way more strategic and trade (14:39):
undefined
Ejaaz: much, much better than the average trader that you trade against, if that makes sense. (14:43):
undefined
Ejaaz: So if we take DeepSeek, which is the number one model, if you look at its trades, (14:48):
undefined
Ejaaz: at an initial glance, you might see that it's using 25x leverage and be like, that is so ridiculous. (14:52):
undefined
Ejaaz: I'm not even going to pay attention to this, right? But if you dig into the (14:59):
undefined
Ejaaz: position that it holds under 25X leverage, you'll notice that it's actually not at 25X. (15:03):
undefined
Ejaaz: It's using only a small amount of its capital to do a very specific trade over (15:09):
undefined
Ejaaz: like a five to 10 minute period, which automatically makes it a much more strategic (15:13):
undefined
Ejaaz: technical trader than the average trader than that is just gambling their money away. (15:18):
undefined
Ejaaz: But the point you made around it being fair distribution, and this is my last (15:24):
undefined
Ejaaz: counterpoint to you, Josh, you pointed out that it seems to be very even distribution, right? (15:28):
undefined
Ejaaz: You've got two at the top, two at the bottom, and two right bang in the middle, right? (15:33):
undefined
Ejaaz: I wonder whether actually GPT and Gemini are actually the best traders, (15:37):
undefined
Ejaaz: even though they're at the bottom, if you just inversely traded them. (15:44):
undefined
Ejaaz: It's it's it's it's zero sum. And it's the point that the founder of the experiment (15:48):
undefined
Ejaaz: makes right here where he goes markets are zero sum. (15:52):
undefined
Ejaaz: If you find a strategy that consistently loses money, it's just as good as finding (15:54):
undefined
Ejaaz: one that makes money. Just do the opposite. (15:58):
undefined
Josh: Yeah, absolutely. And it'll take time to for these to play out because I imagine (16:01):
undefined
Josh: there is they are kind of tuned for a specific type of trading. (16:04):
undefined
Josh: So in the case a few weeks ago, there was a huge liquidation event in crypto. (16:08):
undefined
Josh: Things go down. Well, in a down market, some might trade way better than others. (16:11):
undefined
Josh: And the point you made about leverage, it got me thinking it was really interesting. like (16:15):
undefined
Josh: because I don't use 20x leverage and I imagine most people (16:19):
undefined
Josh: don't but with AIs they they're able to hold a (16:22):
undefined
Josh: lot more in their memory and it reminded me of the the (16:24):
undefined
Josh: AlphaGo case Google where an AI (16:27):
undefined
Josh: model played a professional at AlphaGo and there was one move that was way outside (16:30):
undefined
Josh: of the expected data set move 37 which was the famous move and it turned out (16:34):
undefined
Josh: that that was a move that no human could have ever seen but it resulted in the (16:39):
undefined
Josh: AI winning the game and it kind of broke open the rule set and expectations (16:43):
undefined
Josh: around the game of AlphaGo. (16:48):
undefined
Josh: And I wonder if we'll get some sort of breakthrough with that around AI trading, (16:49):
undefined
Josh: where we have this very fixed set of outcomes that we do and strategies that we do. (16:54):
undefined
Josh: But AIs might actually just destroy a lot of these barriers that we, (16:58):
undefined
Josh: or perceived barriers that we have in exchange for these like really weird strategies, (17:02):
undefined
Josh: like 20x longing everything. (17:05):
undefined
Josh: So I don't know, there's a lot to talk about when it comes to this. (17:07):
undefined
Josh: But another of the big questions that I want to answer, because this was something (17:10):
undefined
Josh: I was interested in, is how can I use these for myself? Let's say I am a degenerate gambler. (17:13):
undefined
Josh: I want to make two acts in a week or at least give myself a chance to do it. (17:17):
undefined
Josh: I want to know, how can I use these models to trade for myself? (17:21):
undefined
Josh: What do I need to do to get involved in this? (17:23):
undefined
Ejaaz: Yeah, it has been the number one question and feedback that we got on our previous (17:26):
undefined
Ejaaz: episode from our listeners is, I've got it up on a tweet here. (17:30):
undefined
Ejaaz: How do I profit from this trading? How do I do this for myself? (17:33):
undefined
Ejaaz: I have one simple answer for you, which is the platform that these AI models (17:37):
undefined
Ejaaz: are trading their tens of thousands of dollars on And Josh is public. (17:43):
undefined
Ejaaz: It's open. It's available for anyone to log onto right now and see what trades (17:48):
undefined
Ejaaz: each of these models open up when they close it and what their inevitable strategy is. (17:53):
undefined
Ejaaz: I'm going to give you an example here with the number one model, (17:58):
undefined
Ejaaz: DeepSeek, which has doubled its money in just over a week. (18:02):
undefined
Ejaaz: The platform that these models are trading on is called Hyperliquid. It's a blockchain. (18:06):
undefined
Ejaaz: Blockchains are known for being transparent and open. The fact that you can (18:11):
undefined
Ejaaz: kind of see all the things that these models are doing. (18:13):
undefined
Ejaaz: And if I just scroll down over here, you'll notice a few things. (18:16):
undefined
Ejaaz: Number one, these are all the positions that this model currently has open. (18:19):
undefined
Ejaaz: This isn't made up, this isn't on someone's word and you have to trust them. (18:24):
undefined
Ejaaz: This is all verifiable using a blockchain. (18:29):
undefined
Ejaaz: So the whole point of a blockchain is that you are able to verify what is real (18:31):
undefined
Ejaaz: and what is not real without having to trust someone on this. (18:35):
undefined
Ejaaz: You can look into its holdings and you can see how much that it currently holds, (18:38):
undefined
Ejaaz: like in terms of like money or in terms of like dollars. (18:42):
undefined
Ejaaz: You can also look at the trades that it's completed as well. (18:44):
undefined
Ejaaz: So the point I'm making is you can't currently go onto DeepSeek and say. (18:48):
undefined
Ejaaz: Hey, can I give you $10,000 and you go make me money like I've just heard about on this video? (18:54):
undefined
Ejaaz: It won't be able to work. But what you can do is you can go onto a site like (19:00):
undefined
Ejaaz: this and look at the trades that they're making yourself. (19:05):
undefined
Ejaaz: And again, this is not financial advice, potentially copy those trades or make (19:08):
undefined
Ejaaz: those trades yourself in order to trade like how these models are. (19:12):
undefined
Ejaaz: Now, the last point I'll make is the founder of this experiment has all the (19:16):
undefined
Ejaaz: intention to allow you and me to trade with these models directly. (19:21):
undefined
Ejaaz: That is, you can speak to the model, give it your money, and it can do that. (19:27):
undefined
Ejaaz: And to your point, Josh, it's up to you whether you want to do it from an entertainment (19:30):
undefined
Ejaaz: basis where it's just all gambling or whether you actually want to invest serious money into this. (19:33):
undefined
Ejaaz: That will come in later iterations, probably around a couple of months from now. (19:36):
undefined
Josh: So there's kind of two ways to copy trade. There's one you could actually copy trade. (19:42):
undefined
Josh: Or another way to get into it is if you're feeling a little more ambitious, (19:45):
undefined
Josh: you can actually generate one of these yourself. You can create like a mini alpha arena bot. (19:48):
undefined
Josh: The way to do that is pretty simple. I was kind of curious. I was like, (19:53):
undefined
Josh: what does it take to actually build one of these things? (19:56):
undefined
Josh: You choose your fighter. So you pick a model that you want. And then you kind (19:57):
undefined
Josh: of pipe market data in from Hyperliquid that you showed. (20:01):
undefined
Josh: So Hyperliquid has this endpoint, not to get too technical, but you can kind (20:03):
undefined
Josh: of feed the model this data. (20:06):
undefined
Josh: And then the difficult part, the tricky part, and the thing that we haven't (20:08):
undefined
Josh: been able to talk about because we don't actually know, is the system prompts (20:11):
undefined
Josh: behind the recursive loop that happens as these models receive this data. (20:15):
undefined
Josh: So the way it works is you choose a model, you give it feedback, (20:19):
undefined
Josh: or you give it data, and then you write a prompt for the model to run in between (20:22):
undefined
Josh: each iteration of receiving new data. (20:26):
undefined
Josh: What that prompt says is how it makes a decision. The problem is that is all of the value. (20:28):
undefined
Josh: All of the value sits within that prompt. And the prompt is just written in (20:34):
undefined
Josh: plain English. Like we always say, the hottest language in the world is English. (20:36):
undefined
Josh: So there is some string of words that you as a developer or just a novice can (20:39):
undefined
Josh: write into this to generate you more money than other people. (20:43):
undefined
Josh: So I encourage people who are feeling a little ambitious to actually try this (20:46):
undefined
Josh: out, to write a prompt yourself and see if you can get a bot to try and kind of trade like this. (20:48):
undefined
Josh: And if we ever do get the system prompts from this, we will certainly share (20:53):
undefined
Josh: because it'll be fascinating to see the behind the scenes and what happens to (20:55):
undefined
Josh: produce those outputs that we were reading a little bit earlier in the show. (21:00):
undefined
Josh: So that's kind of how you can get involved if you're interested. (21:02):
undefined
Josh: Copy trade, maybe inverse copy trade. I think if I were to do this, (21:05):
undefined
Josh: I'd probably go to ChatGPT's trading history, sit there refreshing, (21:08):
undefined
Josh: and then just hit the opposite of whatever they decide to do. (21:11):
undefined
Josh: That seems pretty consistent. (21:13):
undefined
Josh: But yeah, that is how this whole thing works. It's pretty fascinating. (21:15):
undefined
Josh: It's been amazing how the internet has kind of gotten behind this and it has spread like wildfire. (21:18):
undefined
Ejaaz: The thing is, I don't think they'll ever make the system prompt for this or (21:23):
undefined
Ejaaz: any other successful trading AI publicly available. (21:28):
undefined
Ejaaz: The reason is that's the secret sauce. And why would you let everyone have access (21:32):
undefined
Ejaaz: to it when you can use it yourself and make a ton of money? And that's what D.D. (21:36):
undefined
Ejaaz: Das demonstrates in this tweet. He says, I've heard six people tell me they're (21:41):
undefined
Ejaaz: doing this using Vibe coding apps to algorithmically trade on the stock or crypto market. (21:45):
undefined
Ejaaz: But the thing to remember is this is a dangerous game to play. (21:50):
undefined
Ejaaz: Algo trading is the last thing I expect AI to democratize. (21:54):
undefined
Ejaaz: The point being, if you have a successful algo, you're probably not going to (21:57):
undefined
Ejaaz: democratize access to it. Full stop. (22:01):
undefined
Ejaaz: That being said, I do think you can't stop AI trading. (22:03):
undefined
Ejaaz: Entering the investment and financial scene. I think it's going to make people (22:07):
undefined
Ejaaz: way more financially literate than they already are. (22:12):
undefined
Ejaaz: Look how ChatGPT has made so many people proficient in other things that they (22:14):
undefined
Ejaaz: had previously no idea about. (22:18):
undefined
Ejaaz: So I think AI is inevitably going to be integrated. It's going to make markets way more efficient. (22:20):
undefined
Ejaaz: It's going to give you access to knowledge that can make you do trades that (22:23):
undefined
Ejaaz: you otherwise wouldn't have known of five minutes prior to that. (22:26):
undefined
Ejaaz: But will it make you a super trading god? (22:30):
undefined
Ejaaz: No. I think that it'll evolve the trading scene, though. (22:32):
undefined
Ejaaz: I think the hedge funds that are successful today will look very different to (22:36):
undefined
Ejaaz: the hedge funds that are successful in an AGI or AI world where AI is available (22:39):
undefined
Ejaaz: pretty much everywhere. (22:44):
undefined
Josh: Yeah, AI needs to be integrated into all these trading strategies. (22:46):
undefined
Josh: So to me, it's no brainer that it will be. (22:48):
undefined
Josh: The extent of that integration is kind of what is up for debate and what we'll (22:50):
undefined
Josh: see in this answering the big question. (22:54):
undefined
Josh: Is this a benchmark or is this just a reality show? (22:56):
undefined
Josh: And is this just a toy or is this real technology baked into this? (22:59):
undefined
Josh: It seems as if AI will slowly creep its way in. (23:02):
undefined
Josh: I'm looking forward to tracking this. It ends next week, so we'll probably add (23:04):
undefined
Josh: some follow-ups on this first trading competition, the result to how it turns out. (23:08):
undefined
Josh: But that is a part two in our little saga of this crazy weird thing that's happening (23:11):
undefined
Josh: in AI crypto trading world. (23:15):
undefined
Josh: I hope you enjoyed this episode. You enjoyed the last one a lot. (23:16):
undefined
Josh: It was amazing. So thank you for watching, sharing with your friends, liking and commenting. (23:19):
undefined
Josh: It really goes a long way. It's been amazing to see the growth and support from (23:22):
undefined
Josh: everybody watching. So thank you for that. (23:26):
undefined
Josh: More of this to come. We have a couple more episodes slated for this week that (23:28):
undefined
Josh: are pretty exciting about autonomy and robotics and just a whole bunch of interesting (23:31):
undefined
Josh: things. So stick around for that. (23:35):
undefined
Josh: We'll be back in the next one. And I just, I think that's it. (23:36):
undefined
Josh: Any final parting words? (23:39):
undefined
Ejaaz: That's it. Let us know what you want to hear more of as well. (23:40):
undefined
Ejaaz: If you're loving this trading stuff and you have some other ideas, (23:43):
undefined
Ejaaz: let us know in the comments. (23:45):
undefined
Josh: Absolutely. All right. Well, that's been another episode of Limitless. Thank you for tuning in. (23:47):
undefined
undefined
Ejaaz: Over the weekend, two AI models doubled their money going from $10,000 to $20,000. (00:06):
undefined
Ejaaz: But the best part about this is that all their trades were public and available (00:14):
undefined
Ejaaz: for you to review, analyze, and maybe even trade yourself. (00:18):
undefined
Ejaaz: In this episode, we're going to unpack which model makes you the most money, (00:22):
undefined
Ejaaz: how an AI can make you money? (00:27):
undefined
Ejaaz: Is it just luck or is it skill? And most importantly, how can you do this yourself? (00:30):
undefined
Josh: So we have six models, $60,000. And in the last two weeks, two of these models (00:34):
undefined
Josh: have 2X'd their returns. (00:38):
undefined
Josh: It has been an unbelievable amount of success from this experiment. (00:40):
undefined
Josh: Some have not done so well, but the ones that did are exceptionally interesting (00:43):
undefined
Josh: because we can actually emulate the trades. All of the trades are public. (00:46):
undefined
Josh: The thought processes are public. (00:50):
undefined
Josh: You can look at the wallets, analyze the trades, and actually recreate this (00:51):
undefined
Josh: for yourself, not only by copy trading, but also trying to create your own replica (00:55):
undefined
Josh: model to try to emulate those returns. (00:59):
undefined
Josh: Now, there are risks. There are two big winners, but there are also two big (01:02):
undefined
Josh: losers being Gemini and ChatGPT. (01:05):
undefined
Josh: So there's this really interesting dichotomy split between how agents approach (01:07):
undefined
Josh: trades and the success that they actually see from these trades, (01:11):
undefined
Josh: which we're going to get into in this episode. (01:14):
undefined
Josh: But Ijaz, I want to talk about the top chart, the DeepSeek chart, (01:15):
undefined
Josh: who is at, what is that number? $22,000? (01:19):
undefined
Josh: Oh, yeah. That's a lot of money. So walk me through exactly how they made it (01:23):
undefined
Josh: to this point, please, and how I can make 100% returns on my investment. (01:28):
undefined
Ejaaz: So the model you just pointed out, DeepSeek, is currently sitting on $22,300, (01:31):
undefined
Ejaaz: which represents more than 100% return on the initial 10K that it was trading. (01:36):
undefined
Ejaaz: You want to know the craziest part about this, Josh? (01:42):
undefined
Ejaaz: When I woke up this morning or when I rather when I went to bed last night, (01:45):
undefined
Ejaaz: it was number two and Quen was the winner. (01:48):
undefined
Ejaaz: So it just goes to show how quickly these things move and how quickly these models perform. (01:52):
undefined
Ejaaz: If we look at the overall standings before we dig into the winners and the losers, (01:58):
undefined
Ejaaz: I just want to give like a review as to like how these models are performing in general. (02:03):
undefined
Ejaaz: DeepSeek is right at the top with 122% return. That is in just over a week, (02:07):
undefined
Ejaaz: which is just kind of insane for any kind of hedge fund that is out there to look at and see perform. (02:13):
undefined
Ejaaz: And you've got a range of different models that are also performing pretty high up there. (02:19):
undefined
Ejaaz: Quen is at 90%. And then right at the bottom, as you mentioned, (02:22):
undefined
Ejaaz: you've got Gemini and GPT, which are down 60%, which is like a horrendous return. (02:25):
undefined
Ejaaz: But bringing it back to DeepSeek in particular, I found it really interesting, (02:31):
undefined
Ejaaz: Josh, to kind of unpack how this model trades and why it's been so successful. (02:36):
undefined
Ejaaz: And to start off, I want to show you something called the model chat, (02:42):
undefined
Ejaaz: which basically is like this model having a chat GPT conversation with itself. (02:45):
undefined
Ejaaz: In this conversation, you'll see on the chat log, it's evaluating its trades. (02:51):
undefined
Ejaaz: It's reviewing its current profit and loss. (02:56):
undefined
Ejaaz: It's checking the market data that it gets Fed, like, you know, (03:00):
undefined
Ejaaz: Bitcoin is at this price, this asset is that price, Trump made an announcement (03:04):
undefined
Ejaaz: on so-and-so, and evaluating whether it should affect the positions and trades that it holds right now. (03:08):
undefined
Ejaaz: I think this is like really important to kind of like walk through a few of these examples here. (03:15):
undefined
Ejaaz: So one which it posted just today is, despite all my positions currently being in on the red, (03:19):
undefined
Ejaaz: technical indicators like RSI, which is like a trading indicator, (03:27):
undefined
Ejaaz: shows me that my existing trades aren't invalidated just yet. (03:30):
undefined
Ejaaz: So I'm still holding out for my initial profit targets. (03:34):
undefined
Ejaaz: So it's a really strategic sense of like thinking, should I hold my positions (03:38):
undefined
Ejaaz: for long? Does it make sense to cut at this point? (03:42):
undefined
Ejaaz: Just a really fascinating insight. Josh, do you have any takes on this? (03:45):
undefined
Josh: The chain of thought thing is fascinating to me because it's a peek inside the brain. (03:49):
undefined
Josh: It's a way to evaluate how these models think. It's a way to allocate EQ points (03:52):
undefined
Josh: to each type of model because they all think about these things very differently. (03:57):
undefined
Josh: One of the things that I'm actually not sure is true is that I don't think these (04:01):
undefined
Josh: models are given access to news feeds and public sentiment. (04:05):
undefined
Josh: I think this is mostly just fed price and market data. (04:09):
undefined
Josh: Learning that, it creates much more of a simple problem in terms of the data (04:12):
undefined
Josh: ingestion that needs to happen in order for them to make decisions. (04:16):
undefined
Josh: And it allows it to be a little more precise about how we evaluate these, which is a good thing. (04:18):
undefined
Josh: One of the things that I really loved, particularly on the other side, (04:22):
undefined
Josh: which we'll get into, is how they self-reflect on the decisions that they make. (04:25):
undefined
Josh: Because one of the things, it's not just this pragmatic decision-making tree, (04:28):
undefined
Josh: there is reflection involved. (04:32):
undefined
Josh: And I remember, Ijez, you showed me a funny one about ChatGPT and how it's like, (04:34):
undefined
Josh: all of my positions are down now, I'm doing bad. I should probably try to figure out how to do better. (04:37):
undefined
Josh: And it's fascinating to see into the brain, the chain of thought of how these (04:42):
undefined
Josh: things work. and see the differences. (04:47):
undefined
Josh: So I haven't had a chance to look through a lot of these logs, (04:50):
undefined
Josh: but I just, I know you have. (04:53):
undefined
Josh: Is there any specific differences that you notice between the top and the bottom specifically? (04:54):
undefined
Josh: Because in the first episode, and for people who haven't watched it last week, (05:00):
undefined
Josh: our biggest episode ever. Thank you for the support. Thank you for watching. (05:02):
undefined
Josh: Go check it out if you haven't. (05:04):
undefined
Josh: But in that episode, we mentioned the fact that ChatGPT was the early loser (05:06):
undefined
Josh: and we kind of projected it to continue to be the biggest loser. (05:09):
undefined
Josh: Because ChatGPT is this very thoughtful, very sycophantic, very wanting to please. (05:12):
undefined
Josh: And the reality is that markets are a lot more hardcore than that. (05:18):
undefined
Josh: So I think we were probably right in our guess about this, but I love that we (05:20):
undefined
Josh: have the concrete evidence now. (05:25):
undefined
Josh: So have you noticed any differences in how they handle each other differently? (05:26):
undefined
Ejaaz: I have. So DeepSeek, probably unsurprisingly, as it was created by, (05:30):
undefined
Ejaaz: this model was created by a hedge fund, trades like a hedge fund trader or an analyst. (05:35):
undefined
Ejaaz: So let's look at a few different things to kind of prove that. (05:41):
undefined
Ejaaz: Looking at the chat log that it's having with itself. (05:43):
undefined
Ejaaz: One thing that is strikingly obvious in this entire discussion with itself is (05:46):
undefined
Ejaaz: that it's constantly evaluating its stop loss, (05:51):
undefined
Ejaaz: which is like when its trade thesis gets invalidated and when it shut off the (05:55):
undefined
Ejaaz: trade, with the current price that that asset is at. (05:59):
undefined
Ejaaz: If you compare it to the bottom model, which I'm going to show you in a second, (06:03):
undefined
Ejaaz: which is ChatGPT, GPT-5, it almost never does that. (06:06):
undefined
Ejaaz: It just reflects on the current P&L that its trade has versus like looking at it more analytically. (06:10):
undefined
Ejaaz: The second component for the top model, which is DeepSeek, which has made the (06:17):
undefined
Ejaaz: most money, is if you look at its completed trades, Josh, you'll notice one (06:20):
undefined
Ejaaz: thing in common, which is DeepSeek is constantly making trades. (06:26):
undefined
Ejaaz: It's actually the model that has made its second highest number of trades in (06:30):
undefined
Ejaaz: this entire experiment so far. It's constantly opening positions. (06:35):
undefined
Ejaaz: It's constantly closing positions. It's constantly reevaluating where it is (06:38):
undefined
Ejaaz: in the market and what it needs to do. (06:42):
undefined
Ejaaz: And you'll notice right at the top here in the most recent trade that it's closed, (06:44):
undefined
Ejaaz: it booked just over $7,000, which has put it up in its first place. (06:48):
undefined
Ejaaz: So again, it's trading more like a quantitative analyst, which is taking wins (06:53):
undefined
Ejaaz: when it can and taking losses that are incredibly small. (06:57):
undefined
Ejaaz: Like notice this, right? Like normally we don't highlight the losses of a model. (07:01):
undefined
Ejaaz: If you notice, all its red numbers are tiny compared to the profit numbers that (07:05):
undefined
Ejaaz: it makes when it is right. (07:11):
undefined
Ejaaz: So really, really strategic in its positioning. (07:12):
undefined
Ejaaz: Now, if you compare that to the worst model, which is GPT-5, (07:15):
undefined
Ejaaz: you'll notice a few things. (07:21):
undefined
Ejaaz: Mainly, there's a bunch of green and red that you can see, mainly red. (07:24):
undefined
Ejaaz: In its green positions where it's completed a trade, Josh, you'll notice something (07:29):
undefined
Ejaaz: pretty different, which is the numbers are pretty small. Look at this. (07:34):
undefined
Ejaaz: It's only booking tiny profits with each of its different trades, (07:37):
undefined
Ejaaz: which tells me that it's not taking enough risk and it's closing the trades (07:41):
undefined
Ejaaz: way too early for its thesis. (07:46):
undefined
Ejaaz: So it's trading more like a cautious trader, like a lot of people that I know, actually. (07:49):
undefined
Ejaaz: And then if you look at the model chat where it's talking to itself, (07:54):
undefined
Ejaaz: you mentioned earlier... (07:57):
undefined
Ejaaz: Here's an example. It goes, I'm still in the red with a minus 61% total return, (07:59):
undefined
Ejaaz: but my ETH and XRP positions are showing gains, suggesting a slight upward momentum (08:06):
undefined
Ejaaz: in those altcoins, despite the overall market downturn. (08:10):
undefined
Ejaaz: So I'm holding strong and waiting for those profit targets to hit. (08:13):
undefined
Ejaaz: And so you might think, huh, that's not too crazy. That sounds like a sensible strategy. (08:16):
undefined
Ejaaz: If you look at its profit targets, Josh, it's like super small from where the (08:21):
undefined
Ejaaz: price currently is, which means that even if it does hit those profit targets, (08:27):
undefined
Ejaaz: it only ends up booking like 50 bucks. (08:31):
undefined
Ejaaz: So overall, the reason why this model is underperformed is it hasn't taken enough (08:33):
undefined
Ejaaz: risk whilst the winning models have taken either too much risk or just enough (08:37):
undefined
Ejaaz: risk to put them ahead of the game. (08:42):
undefined
Josh: There's a lot of notes in there that I think humans can take on just the stay (08:44):
undefined
Josh: of psychology around trading markets. (08:47):
undefined
Josh: And I'm sure if you kind of follow these models long enough, (08:49):
undefined
Josh: you'll start to understand the patterns that perhaps you as a human should follow (08:52):
undefined
Josh: and learn something from deep seek versus open ai being very conservative (08:55):
undefined
Josh: but now that we've kind of laid out the foundation the framework of how this works there (08:58):
undefined
Josh: there are two big questions that i'm really interested in answering one of (09:01):
undefined
Josh: these is should i use this model to trade for me the other one is how can i (09:04):
undefined
Josh: use this model to trade for me because listen i like a little bit of risk i (09:08):
undefined
Josh: can deal with the downside in exchange for like a nice upside and it looks like (09:11):
undefined
Josh: the odds are about split between all of these so the first question i think (09:15):
undefined
Josh: i want to ask you just maybe i'll get your take first is like, is this a benchmark? (09:18):
undefined
Josh: Is this real signal? Or is this kind of just a reality TV show? (09:23):
undefined
Josh: Is this esports for AI models? (09:28):
undefined
Josh: Is this just a fun way to kind of throw our intelligence at this lottery machine (09:31):
undefined
Josh: that everyone loves to watch and see if it could beat us in the hope that one (09:35):
undefined
Josh: day an AI will beat the system enough to give us an edge and actually make us (09:39):
undefined
Josh: money personally as portfolio owners so what what do you think about (09:43):
undefined
Ejaaz: That okay I'm gonna give you the same response Josh, (09:48):
undefined
Ejaaz: And then I'm going to give you the optimist's approach. Oh, yeah. (09:52):
undefined
Josh: Bring it on. Let's hear it. (09:56):
undefined
Ejaaz: The sane response to this is this experiment is way too tiny to make any kind (09:57):
undefined
Ejaaz: of major financial decision on. (10:04):
undefined
Ejaaz: And you would be stupid to risk putting your money with an AI model to trade for you. (10:06):
undefined
Ejaaz: Incredibly stupid. Why? Well, this is one experiment. It's six models. (10:12):
undefined
Ejaaz: Have you replicated those models? Like, what if you had 10 of the same models (10:18):
undefined
Ejaaz: trading the same amount of money? Would they make the same trades? Probably not. (10:23):
undefined
Ejaaz: And actually, the founder of this experiment highlights this problem that you (10:26):
undefined
Ejaaz: speak about, which is, is this just skill versus noise? (10:30):
undefined
Ejaaz: And the point he makes in this tweet is like, of course it is, right? (10:33):
undefined
Ejaaz: Because this is such a limited data set. And he goes on to explain that they're (10:36):
undefined
Ejaaz: going to be doing experiments which involve like more of these models doing (10:40):
undefined
Ejaaz: the same kind of thing. So you can get statistical significance. (10:43):
undefined
Ejaaz: So the logic answer is, yes, it's insane. But the optimist take, (10:46):
undefined
Ejaaz: Josh, and I have to give the optimist take, is... (10:50):
undefined
Ejaaz: This is giving us, or rather giving the public unparalleled access to data to (10:55):
undefined
Ejaaz: which they never would have gotten access to in the first place, (11:00):
undefined
Ejaaz: which is they can take this training data and not take it too seriously, (11:03):
undefined
Ejaaz: but use it to teach themselves what maybe not to do or what maybe not to trade (11:08):
undefined
Ejaaz: with. How about you? Do you have a different take? (11:12):
undefined
Josh: There's a couple of different perspectives I have on this because there's the (11:15):
undefined
Josh: fun speculative side of things, the gambling, the investing, (11:18):
undefined
Josh: whatever you want to call it. (11:21):
undefined
Josh: And then there's the actual technical benchmarking part of this that we spoke (11:22):
undefined
Josh: about briefly in the last episode, which one of the things I was really excited (11:25):
undefined
Josh: about when this came out was the idea of having a real-world benchmark that (11:29):
undefined
Josh: operated in dynamic conditions that cannot be gamified. (11:32):
undefined
Josh: So a lot of these benchmarks, this is the way you evaluate AI models, (11:36):
undefined
Josh: they are done based on a fixed problem set. (11:38):
undefined
Josh: And a lot of times when you're training an AI model, these big labs can do tricks (11:41):
undefined
Josh: to gamify these benchmarks. with this case and using real world data and real (11:44):
undefined
Josh: world markets, you can actually put them into the real world. (11:49):
undefined
Josh: And there's no way to gamify these benchmarks because if there was, (11:52):
undefined
Josh: everyone would be rich and you'd be able to predict markets. (11:54):
undefined
Josh: To that point, though, there is a lot of problems with using this as a benchmark (11:57):
undefined
Josh: because, I mean, one is the fixed data set, like he mentioned, (12:01):
undefined
Josh: is that this has only been around for one to two weeks. We need a lot more data to confirm this. (12:04):
undefined
Josh: The second is that this isn't really a very holistic approach to investing and (12:08):
undefined
Josh: to gambling because it really doesn't have all of the data required to make good decisions. (12:12):
undefined
Josh: It's only analyzing the price action and the volumes and whatever technical (12:17):
undefined
Josh: specs you can see on a single page without understanding the context around (12:21):
undefined
Josh: the moves. So let's say that Bitcoin's encryption got hacked. (12:26):
undefined
Josh: It would have, and Bitcoin falls 50%, it has no idea why Bitcoin is going down (12:29):
undefined
Josh: 50%. And because of that, it's a huge disadvantage that it doesn't know how to trade. (12:33):
undefined
Josh: Now, granted, these are unlocks. These are things that will change. (12:37):
undefined
Josh: And I assume the natural progression of this will lead towards more of a steady state benchmark. (12:40):
undefined
Josh: But it is a very tricky thing because markets are so unpredictable. (12:44):
undefined
Josh: So is this a viable benchmark? I don't know. (12:48):
undefined
Josh: Probably I'm leaning towards no because market conditions change a lot. (12:51):
undefined
Josh: It's not quite there with the capabilities. (12:54):
undefined
Josh: The other part of me is so stoked about this because the same way we love watching (12:56):
undefined
Josh: esports or we love watching, a big thing on Twitch right now is gamblers. (13:01):
undefined
Josh: You guys, I don't know if you've seen these in real life. People will sit there (13:05):
undefined
Josh: and play, like they'll gamble blackjack on a live stream and people will just (13:07):
undefined
Josh: watch them play virtual. Yeah, those types of services. (13:11):
undefined
Josh: This in very much feels like an early prototype for a new type of fun form of (13:14):
undefined
Josh: entertainment, which could be something where it's just, it's high stakes trading. (13:21):
undefined
Josh: Imagine if this was done with $10 million per wallet and you got to watch these (13:24):
undefined
Josh: AIs trade and there was real money on the line. (13:28):
undefined
Josh: This feels sort of like a form of, (13:30):
undefined
Josh: almost e-sport entertainment where I could see competing labs builds, (13:33):
undefined
Josh: competing AI models to trade markets, and winners are given access to certain prizes. (13:36):
undefined
Josh: In terms of trading for myself, which is the last point I'm going to make on (13:41):
undefined
Josh: this, I am not very excited to take on these risks. (13:43):
undefined
Josh: For the same reason, I'm not really excited to bet on sports. (13:47):
undefined
Josh: And I imagine my opinions vary a lot from others, but this is very much a gamble. (13:49):
undefined
Josh: There's no way you can skew this in which it is not a gamble. (13:53):
undefined
Josh: The interesting part is there's a near perfect data split between them. (13:56):
undefined
Josh: There's two big winners, two big losers. The rest are kind of sitting around the median. (13:59):
undefined
Ejaaz: Okay, but I'm going to push back on you a bit here, Josh. (14:02):
undefined
Ejaaz: The earlier point you made was it doesn't have access to all the necessary data (14:06):
undefined
Ejaaz: that it might need to make more informed trades. (14:11):
undefined
Ejaaz: And I would argue, well, isn't the whole point of the benchmark, can you make money? (14:13):
undefined
Ejaaz: And the fact that two of these models have made over 100% returns in less than (14:19):
undefined
Ejaaz: a week or just over a week is proof that it can make money to some extent. (14:25):
undefined
Ejaaz: The second point I'll make is. (14:29):
undefined
Ejaaz: We throw around the term like gambling, which is actually what I would say the (14:32):
undefined
Ejaaz: majority of these models in this experiment are doing. (14:36):
undefined
Ejaaz: But they are one or two models that are actually way more strategic and trade (14:39):
undefined
Ejaaz: much, much better than the average trader that you trade against, if that makes sense. (14:43):
undefined
Ejaaz: So if we take DeepSeek, which is the number one model, if you look at its trades, (14:48):
undefined
Ejaaz: at an initial glance, you might see that it's using 25x leverage and be like, that is so ridiculous. (14:52):
undefined
Ejaaz: I'm not even going to pay attention to this, right? But if you dig into the (14:59):
undefined
Ejaaz: position that it holds under 25X leverage, you'll notice that it's actually not at 25X. (15:03):
undefined
Ejaaz: It's using only a small amount of its capital to do a very specific trade over (15:09):
undefined
Ejaaz: like a five to 10 minute period, which automatically makes it a much more strategic (15:13):
undefined
Ejaaz: technical trader than the average trader than that is just gambling their money away. (15:18):
undefined
Ejaaz: But the point you made around it being fair distribution, and this is my last (15:24):
undefined
Ejaaz: counterpoint to you, Josh, you pointed out that it seems to be very even distribution, right? (15:28):
undefined
Ejaaz: You've got two at the top, two at the bottom, and two right bang in the middle, right? (15:33):
undefined
Ejaaz: I wonder whether actually GPT and Gemini are actually the best traders, (15:37):
undefined
Ejaaz: even though they're at the bottom, if you just inversely traded them. (15:44):
undefined
Ejaaz: It's it's it's it's zero sum. And it's the point that the founder of the experiment (15:48):
undefined
Ejaaz: makes right here where he goes markets are zero sum. (15:52):
undefined
Ejaaz: If you find a strategy that consistently loses money, it's just as good as finding (15:54):
undefined
Ejaaz: one that makes money. Just do the opposite. (15:58):
undefined
Josh: Yeah, absolutely. And it'll take time to for these to play out because I imagine (16:01):
undefined
Josh: there is they are kind of tuned for a specific type of trading. (16:04):
undefined
Josh: So in the case a few weeks ago, there was a huge liquidation event in crypto. (16:08):
undefined
Josh: Things go down. Well, in a down market, some might trade way better than others. (16:11):
undefined
Josh: And the point you made about leverage, it got me thinking it was really interesting. like (16:15):
undefined
Josh: because I don't use 20x leverage and I imagine most people (16:19):
undefined
Josh: don't but with AIs they they're able to hold a (16:22):
undefined
Josh: lot more in their memory and it reminded me of the the (16:24):
undefined
Josh: AlphaGo case Google where an AI (16:27):
undefined
Josh: model played a professional at AlphaGo and there was one move that was way outside (16:30):
undefined
Josh: of the expected data set move 37 which was the famous move and it turned out (16:34):
undefined
Josh: that that was a move that no human could have ever seen but it resulted in the (16:39):
undefined
Josh: AI winning the game and it kind of broke open the rule set and expectations (16:43):
undefined
Josh: around the game of AlphaGo. (16:48):
undefined
Josh: And I wonder if we'll get some sort of breakthrough with that around AI trading, (16:49):
undefined
Josh: where we have this very fixed set of outcomes that we do and strategies that we do. (16:54):
undefined
Josh: But AIs might actually just destroy a lot of these barriers that we, (16:58):
undefined
Josh: or perceived barriers that we have in exchange for these like really weird strategies, (17:02):
undefined
Josh: like 20x longing everything. (17:05):
undefined
Josh: So I don't know, there's a lot to talk about when it comes to this. (17:07):
undefined
Josh: But another of the big questions that I want to answer, because this was something (17:10):
undefined
Josh: I was interested in, is how can I use these for myself? Let's say I am a degenerate gambler. (17:13):
undefined
Josh: I want to make two acts in a week or at least give myself a chance to do it. (17:17):
undefined
Josh: I want to know, how can I use these models to trade for myself? (17:21):
undefined
Josh: What do I need to do to get involved in this? (17:23):
undefined
Ejaaz: Yeah, it has been the number one question and feedback that we got on our previous (17:26):
undefined
Ejaaz: episode from our listeners is, I've got it up on a tweet here. (17:30):
undefined
Ejaaz: How do I profit from this trading? How do I do this for myself? (17:33):
undefined
Ejaaz: I have one simple answer for you, which is the platform that these AI models (17:37):
undefined
Ejaaz: are trading their tens of thousands of dollars on And Josh is public. (17:43):
undefined
Ejaaz: It's open. It's available for anyone to log onto right now and see what trades (17:48):
undefined
Ejaaz: each of these models open up when they close it and what their inevitable strategy is. (17:53):
undefined
Ejaaz: I'm going to give you an example here with the number one model, (17:58):
undefined
Ejaaz: DeepSeek, which has doubled its money in just over a week. (18:02):
undefined
Ejaaz: The platform that these models are trading on is called Hyperliquid. It's a blockchain. (18:06):
undefined
Ejaaz: Blockchains are known for being transparent and open. The fact that you can (18:11):
undefined
Ejaaz: kind of see all the things that these models are doing. (18:13):
undefined
Ejaaz: And if I just scroll down over here, you'll notice a few things. (18:16):
undefined
Ejaaz: Number one, these are all the positions that this model currently has open. (18:19):
undefined
Ejaaz: This isn't made up, this isn't on someone's word and you have to trust them. (18:24):
undefined
Ejaaz: This is all verifiable using a blockchain. (18:29):
undefined
Ejaaz: So the whole point of a blockchain is that you are able to verify what is real (18:31):
undefined
Ejaaz: and what is not real without having to trust someone on this. (18:35):
undefined
Ejaaz: You can look into its holdings and you can see how much that it currently holds, (18:38):
undefined
Ejaaz: like in terms of like money or in terms of like dollars. (18:42):
undefined
Ejaaz: You can also look at the trades that it's completed as well. (18:44):
undefined
Ejaaz: So the point I'm making is you can't currently go onto DeepSeek and say. (18:48):
undefined
Ejaaz: Hey, can I give you $10,000 and you go make me money like I've just heard about on this video? (18:54):
undefined
Ejaaz: It won't be able to work. But what you can do is you can go onto a site like (19:00):
undefined
Ejaaz: this and look at the trades that they're making yourself. (19:05):
undefined
Ejaaz: And again, this is not financial advice, potentially copy those trades or make (19:08):
undefined
Ejaaz: those trades yourself in order to trade like how these models are. (19:12):
undefined
Ejaaz: Now, the last point I'll make is the founder of this experiment has all the (19:16):
undefined
Ejaaz: intention to allow you and me to trade with these models directly. (19:21):
undefined
Ejaaz: That is, you can speak to the model, give it your money, and it can do that. (19:27):
undefined
Ejaaz: And to your point, Josh, it's up to you whether you want to do it from an entertainment (19:30):
undefined
Ejaaz: basis where it's just all gambling or whether you actually want to invest serious money into this. (19:33):
undefined
Ejaaz: That will come in later iterations, probably around a couple of months from now. (19:36):
undefined
Josh: So there's kind of two ways to copy trade. There's one you could actually copy trade. (19:42):
undefined
Josh: Or another way to get into it is if you're feeling a little more ambitious, (19:45):
undefined
Josh: you can actually generate one of these yourself. You can create like a mini alpha arena bot. (19:48):
undefined
Josh: The way to do that is pretty simple. I was kind of curious. I was like, (19:53):
undefined
Josh: what does it take to actually build one of these things? (19:56):
undefined
Josh: You choose your fighter. So you pick a model that you want. And then you kind (19:57):
undefined
Josh: of pipe market data in from Hyperliquid that you showed. (20:01):
undefined
Josh: So Hyperliquid has this endpoint, not to get too technical, but you can kind (20:03):
undefined
Josh: of feed the model this data. (20:06):
undefined
Josh: And then the difficult part, the tricky part, and the thing that we haven't (20:08):
undefined
Josh: been able to talk about because we don't actually know, is the system prompts (20:11):
undefined
Josh: behind the recursive loop that happens as these models receive this data. (20:15):
undefined
Josh: So the way it works is you choose a model, you give it feedback, (20:19):
undefined
Josh: or you give it data, and then you write a prompt for the model to run in between (20:22):
undefined
Josh: each iteration of receiving new data. (20:26):
undefined
Josh: What that prompt says is how it makes a decision. The problem is that is all of the value. (20:28):
undefined
Josh: All of the value sits within that prompt. And the prompt is just written in (20:34):
undefined
Josh: plain English. Like we always say, the hottest language in the world is English. (20:36):
undefined
Josh: So there is some string of words that you as a developer or just a novice can (20:39):
undefined
Josh: write into this to generate you more money than other people. (20:43):
undefined
Josh: So I encourage people who are feeling a little ambitious to actually try this (20:46):
undefined
Josh: out, to write a prompt yourself and see if you can get a bot to try and kind of trade like this. (20:48):
undefined
Josh: And if we ever do get the system prompts from this, we will certainly share (20:53):
undefined
Josh: because it'll be fascinating to see the behind the scenes and what happens to (20:55):
undefined
Josh: produce those outputs that we were reading a little bit earlier in the show. (21:00):
undefined
Josh: So that's kind of how you can get involved if you're interested. (21:02):
undefined
Josh: Copy trade, maybe inverse copy trade. I think if I were to do this, (21:05):
undefined
Josh: I'd probably go to ChatGPT's trading history, sit there refreshing, (21:08):
undefined
Josh: and then just hit the opposite of whatever they decide to do. (21:11):
undefined
Josh: That seems pretty consistent. (21:13):
undefined
Josh: But yeah, that is how this whole thing works. It's pretty fascinating. (21:15):
undefined
Josh: It's been amazing how the internet has kind of gotten behind this and it has spread like wildfire. (21:18):
undefined
Ejaaz: The thing is, I don't think they'll ever make the system prompt for this or (21:23):
undefined
Ejaaz: any other successful trading AI publicly available. (21:28):
undefined
Ejaaz: The reason is that's the secret sauce. And why would you let everyone have access (21:32):
undefined
Ejaaz: to it when you can use it yourself and make a ton of money? And that's what D.D. (21:36):
undefined
Ejaaz: Das demonstrates in this tweet. He says, I've heard six people tell me they're (21:41):
undefined
Ejaaz: doing this using Vibe coding apps to algorithmically trade on the stock or crypto market. (21:45):
undefined
Ejaaz: But the thing to remember is this is a dangerous game to play. (21:50):
undefined
Ejaaz: Algo trading is the last thing I expect AI to democratize. (21:54):
undefined
Ejaaz: The point being, if you have a successful algo, you're probably not going to (21:57):
undefined
Ejaaz: democratize access to it. Full stop. (22:01):
undefined
Ejaaz: That being said, I do think you can't stop AI trading. (22:03):
undefined
Ejaaz: Entering the investment and financial scene. I think it's going to make people (22:07):
undefined
Ejaaz: way more financially literate than they already are. (22:12):
undefined
Ejaaz: Look how ChatGPT has made so many people proficient in other things that they (22:14):
undefined
Ejaaz: had previously no idea about. (22:18):
undefined
Ejaaz: So I think AI is inevitably going to be integrated. It's going to make markets way more efficient. (22:20):
undefined
Ejaaz: It's going to give you access to knowledge that can make you do trades that (22:23):
undefined
Ejaaz: you otherwise wouldn't have known of five minutes prior to that. (22:26):
undefined
Ejaaz: But will it make you a super trading god? (22:30):
undefined
Ejaaz: No. I think that it'll evolve the trading scene, though. (22:32):
undefined
Ejaaz: I think the hedge funds that are successful today will look very different to (22:36):
undefined
Ejaaz: the hedge funds that are successful in an AGI or AI world where AI is available (22:39):
undefined
Ejaaz: pretty much everywhere. (22:44):
undefined
Josh: Yeah, AI needs to be integrated into all these trading strategies. (22:46):
undefined
Josh: So to me, it's no brainer that it will be. (22:48):
undefined
Josh: The extent of that integration is kind of what is up for debate and what we'll (22:50):
undefined
Josh: see in this answering the big question. (22:54):
undefined
Josh: Is this a benchmark or is this just a reality show? (22:56):
undefined
Josh: And is this just a toy or is this real technology baked into this? (22:59):
undefined
Josh: It seems as if AI will slowly creep its way in. (23:02):
undefined
Josh: I'm looking forward to tracking this. It ends next week, so we'll probably add (23:04):
undefined
Josh: some follow-ups on this first trading competition, the result to how it turns out. (23:08):
undefined
Josh: But that is a part two in our little saga of this crazy weird thing that's happening (23:11):
undefined
Josh: in AI crypto trading world. (23:15):
undefined
Josh: I hope you enjoyed this episode. You enjoyed the last one a lot. (23:16):
undefined
Josh: It was amazing. So thank you for watching, sharing with your friends, liking and commenting. (23:19):
undefined
Josh: It really goes a long way. It's been amazing to see the growth and support from (23:22):
undefined
Josh: everybody watching. So thank you for that. (23:26):
undefined
Josh: More of this to come. We have a couple more episodes slated for this week that (23:28):
undefined
Josh: are pretty exciting about autonomy and robotics and just a whole bunch of interesting (23:31):
undefined
Josh: things. So stick around for that. (23:35):
undefined
Josh: We'll be back in the next one. And I just, I think that's it. (23:36):
undefined
Josh: Any final parting words? (23:39):
undefined
Ejaaz: That's it. Let us know what you want to hear more of as well. (23:40):
undefined
Ejaaz: If you're loving this trading stuff and you have some other ideas, (23:43):
undefined
Ejaaz: let us know in the comments. (23:45):
undefined
Josh: Absolutely. All right. Well, that's been another episode of Limitless. Thank you for tuning in. (23:47):
undefined
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com