1
00:00:03,180 --> 00:00:06,700
Ejaaz:
It's official. AI models can make you rich.

2
00:00:06,960 --> 00:00:13,840
Ejaaz:
Over the weekend, two AI models doubled their money going from $10,000 to $20,000.

3
00:00:14,260 --> 00:00:18,320
Ejaaz:
But the best part about this is that all their trades were public and available

4
00:00:18,320 --> 00:00:22,480
Ejaaz:
for you to review, analyze, and maybe even trade yourself.

5
00:00:22,780 --> 00:00:27,140
Ejaaz:
In this episode, we're going to unpack which model makes you the most money,

6
00:00:27,340 --> 00:00:29,900
Ejaaz:
how an AI can make you money?

7
00:00:30,080 --> 00:00:34,640
Ejaaz:
Is it just luck or is it skill? And most importantly, how can you do this yourself?

8
00:00:34,940 --> 00:00:38,980
Josh:
So we have six models, $60,000. And in the last two weeks, two of these models

9
00:00:38,980 --> 00:00:40,480
Josh:
have 2X'd their returns.

10
00:00:40,660 --> 00:00:43,540
Josh:
It has been an unbelievable amount of success from this experiment.

11
00:00:43,840 --> 00:00:46,720
Josh:
Some have not done so well, but the ones that did are exceptionally interesting

12
00:00:46,720 --> 00:00:50,140
Josh:
because we can actually emulate the trades. All of the trades are public.

13
00:00:50,280 --> 00:00:51,560
Josh:
The thought processes are public.

14
00:00:51,900 --> 00:00:55,100
Josh:
You can look at the wallets, analyze the trades, and actually recreate this

15
00:00:55,100 --> 00:00:59,940
Josh:
for yourself, not only by copy trading, but also trying to create your own replica

16
00:00:59,940 --> 00:01:01,920
Josh:
model to try to emulate those returns.

17
00:01:02,060 --> 00:01:05,600
Josh:
Now, there are risks. There are two big winners, but there are also two big

18
00:01:05,600 --> 00:01:07,520
Josh:
losers being Gemini and ChatGPT.

19
00:01:07,760 --> 00:01:11,360
Josh:
So there's this really interesting dichotomy split between how agents approach

20
00:01:11,360 --> 00:01:14,260
Josh:
trades and the success that they actually see from these trades,

21
00:01:14,360 --> 00:01:15,620
Josh:
which we're going to get into in this episode.

22
00:01:15,860 --> 00:01:19,500
Josh:
But Ijaz, I want to talk about the top chart, the DeepSeek chart,

23
00:01:19,760 --> 00:01:22,960
Josh:
who is at, what is that number? $22,000?

24
00:01:23,860 --> 00:01:28,020
Josh:
Oh, yeah. That's a lot of money. So walk me through exactly how they made it

25
00:01:28,020 --> 00:01:31,360
Josh:
to this point, please, and how I can make 100% returns on my investment.

26
00:01:31,720 --> 00:01:36,640
Ejaaz:
So the model you just pointed out, DeepSeek, is currently sitting on $22,300,

27
00:01:36,640 --> 00:01:42,300
Ejaaz:
which represents more than 100% return on the initial 10K that it was trading.

28
00:01:42,580 --> 00:01:44,340
Ejaaz:
You want to know the craziest part about this, Josh?

29
00:01:45,080 --> 00:01:48,180
Ejaaz:
When I woke up this morning or when I rather when I went to bed last night,

30
00:01:48,440 --> 00:01:52,260
Ejaaz:
it was number two and Quen was the winner.

31
00:01:52,420 --> 00:01:57,820
Ejaaz:
So it just goes to show how quickly these things move and how quickly these models perform.

32
00:01:58,260 --> 00:02:03,200
Ejaaz:
If we look at the overall standings before we dig into the winners and the losers,

33
00:02:03,460 --> 00:02:07,380
Ejaaz:
I just want to give like a review as to like how these models are performing in general.

34
00:02:07,520 --> 00:02:13,520
Ejaaz:
DeepSeek is right at the top with 122% return. That is in just over a week,

35
00:02:13,540 --> 00:02:18,880
Ejaaz:
which is just kind of insane for any kind of hedge fund that is out there to look at and see perform.

36
00:02:19,260 --> 00:02:22,100
Ejaaz:
And you've got a range of different models that are also performing pretty high up there.

37
00:02:22,260 --> 00:02:25,760
Ejaaz:
Quen is at 90%. And then right at the bottom, as you mentioned,

38
00:02:25,920 --> 00:02:31,380
Ejaaz:
you've got Gemini and GPT, which are down 60%, which is like a horrendous return.

39
00:02:31,680 --> 00:02:36,620
Ejaaz:
But bringing it back to DeepSeek in particular, I found it really interesting,

40
00:02:36,780 --> 00:02:41,800
Ejaaz:
Josh, to kind of unpack how this model trades and why it's been so successful.

41
00:02:42,120 --> 00:02:45,960
Ejaaz:
And to start off, I want to show you something called the model chat,

42
00:02:45,960 --> 00:02:51,820
Ejaaz:
which basically is like this model having a chat GPT conversation with itself.

43
00:02:51,980 --> 00:02:56,800
Ejaaz:
In this conversation, you'll see on the chat log, it's evaluating its trades.

44
00:02:56,800 --> 00:03:00,060
Ejaaz:
It's reviewing its current profit and loss.

45
00:03:00,180 --> 00:03:04,560
Ejaaz:
It's checking the market data that it gets Fed, like, you know,

46
00:03:04,660 --> 00:03:08,520
Ejaaz:
Bitcoin is at this price, this asset is that price, Trump made an announcement

47
00:03:08,520 --> 00:03:14,660
Ejaaz:
on so-and-so, and evaluating whether it should affect the positions and trades that it holds right now.

48
00:03:15,180 --> 00:03:19,640
Ejaaz:
I think this is like really important to kind of like walk through a few of these examples here.

49
00:03:19,800 --> 00:03:27,020
Ejaaz:
So one which it posted just today is, despite all my positions currently being in on the red,

50
00:03:27,500 --> 00:03:30,620
Ejaaz:
technical indicators like RSI, which is like a trading indicator,

51
00:03:30,880 --> 00:03:34,800
Ejaaz:
shows me that my existing trades aren't invalidated just yet.

52
00:03:34,940 --> 00:03:38,020
Ejaaz:
So I'm still holding out for my initial profit targets.

53
00:03:38,040 --> 00:03:42,640
Ejaaz:
So it's a really strategic sense of like thinking, should I hold my positions

54
00:03:42,640 --> 00:03:45,120
Ejaaz:
for long? Does it make sense to cut at this point?

55
00:03:45,320 --> 00:03:48,800
Ejaaz:
Just a really fascinating insight. Josh, do you have any takes on this?

56
00:03:49,430 --> 00:03:52,690
Josh:
The chain of thought thing is fascinating to me because it's a peek inside the brain.

57
00:03:52,870 --> 00:03:57,970
Josh:
It's a way to evaluate how these models think. It's a way to allocate EQ points

58
00:03:57,970 --> 00:04:01,350
Josh:
to each type of model because they all think about these things very differently.

59
00:04:01,890 --> 00:04:05,410
Josh:
One of the things that I'm actually not sure is true is that I don't think these

60
00:04:05,410 --> 00:04:09,190
Josh:
models are given access to news feeds and public sentiment.

61
00:04:09,410 --> 00:04:12,330
Josh:
I think this is mostly just fed price and market data.

62
00:04:12,670 --> 00:04:16,350
Josh:
Learning that, it creates much more of a simple problem in terms of the data

63
00:04:16,350 --> 00:04:18,390
Josh:
ingestion that needs to happen in order for them to make decisions.

64
00:04:18,390 --> 00:04:22,190
Josh:
And it allows it to be a little more precise about how we evaluate these, which is a good thing.

65
00:04:22,570 --> 00:04:25,030
Josh:
One of the things that I really loved, particularly on the other side,

66
00:04:25,110 --> 00:04:28,730
Josh:
which we'll get into, is how they self-reflect on the decisions that they make.

67
00:04:28,990 --> 00:04:32,330
Josh:
Because one of the things, it's not just this pragmatic decision-making tree,

68
00:04:32,490 --> 00:04:33,970
Josh:
there is reflection involved.

69
00:04:34,090 --> 00:04:37,810
Josh:
And I remember, Ijez, you showed me a funny one about ChatGPT and how it's like,

70
00:04:37,930 --> 00:04:41,950
Josh:
all of my positions are down now, I'm doing bad. I should probably try to figure out how to do better.

71
00:04:42,310 --> 00:04:47,250
Josh:
And it's fascinating to see into the brain, the chain of thought of how these

72
00:04:47,250 --> 00:04:50,230
Josh:
things work. and see the differences.

73
00:04:50,430 --> 00:04:53,230
Josh:
So I haven't had a chance to look through a lot of these logs,

74
00:04:53,370 --> 00:04:54,390
Josh:
but I just, I know you have.

75
00:04:54,570 --> 00:05:00,030
Josh:
Is there any specific differences that you notice between the top and the bottom specifically?

76
00:05:00,250 --> 00:05:02,550
Josh:
Because in the first episode, and for people who haven't watched it last week,

77
00:05:02,850 --> 00:05:04,870
Josh:
our biggest episode ever. Thank you for the support. Thank you for watching.

78
00:05:04,990 --> 00:05:05,950
Josh:
Go check it out if you haven't.

79
00:05:06,110 --> 00:05:09,630
Josh:
But in that episode, we mentioned the fact that ChatGPT was the early loser

80
00:05:09,630 --> 00:05:12,550
Josh:
and we kind of projected it to continue to be the biggest loser.

81
00:05:12,730 --> 00:05:17,830
Josh:
Because ChatGPT is this very thoughtful, very sycophantic, very wanting to please.

82
00:05:18,150 --> 00:05:20,570
Josh:
And the reality is that markets are a lot more hardcore than that.

83
00:05:20,710 --> 00:05:25,150
Josh:
So I think we were probably right in our guess about this, but I love that we

84
00:05:25,150 --> 00:05:26,790
Josh:
have the concrete evidence now.

85
00:05:26,910 --> 00:05:29,490
Josh:
So have you noticed any differences in how they handle each other differently?

86
00:05:30,370 --> 00:05:35,450
Ejaaz:
I have. So DeepSeek, probably unsurprisingly, as it was created by,

87
00:05:35,610 --> 00:05:40,770
Ejaaz:
this model was created by a hedge fund, trades like a hedge fund trader or an analyst.

88
00:05:41,050 --> 00:05:43,750
Ejaaz:
So let's look at a few different things to kind of prove that.

89
00:05:43,950 --> 00:05:46,570
Ejaaz:
Looking at the chat log that it's having with itself.

90
00:05:46,990 --> 00:05:51,590
Ejaaz:
One thing that is strikingly obvious in this entire discussion with itself is

91
00:05:51,590 --> 00:05:55,030
Ejaaz:
that it's constantly evaluating its stop loss,

92
00:05:55,170 --> 00:05:59,650
Ejaaz:
which is like when its trade thesis gets invalidated and when it shut off the

93
00:05:59,650 --> 00:06:02,990
Ejaaz:
trade, with the current price that that asset is at.

94
00:06:03,130 --> 00:06:06,370
Ejaaz:
If you compare it to the bottom model, which I'm going to show you in a second,

95
00:06:06,510 --> 00:06:10,050
Ejaaz:
which is ChatGPT, GPT-5, it almost never does that.

96
00:06:10,270 --> 00:06:16,890
Ejaaz:
It just reflects on the current P&L that its trade has versus like looking at it more analytically.

97
00:06:17,150 --> 00:06:20,950
Ejaaz:
The second component for the top model, which is DeepSeek, which has made the

98
00:06:20,950 --> 00:06:26,210
Ejaaz:
most money, is if you look at its completed trades, Josh, you'll notice one

99
00:06:26,210 --> 00:06:30,070
Ejaaz:
thing in common, which is DeepSeek is constantly making trades.

100
00:06:30,250 --> 00:06:35,230
Ejaaz:
It's actually the model that has made its second highest number of trades in

101
00:06:35,230 --> 00:06:38,590
Ejaaz:
this entire experiment so far. It's constantly opening positions.

102
00:06:38,650 --> 00:06:42,930
Ejaaz:
It's constantly closing positions. It's constantly reevaluating where it is

103
00:06:42,930 --> 00:06:44,570
Ejaaz:
in the market and what it needs to do.

104
00:06:44,690 --> 00:06:48,590
Ejaaz:
And you'll notice right at the top here in the most recent trade that it's closed,

105
00:06:48,750 --> 00:06:53,030
Ejaaz:
it booked just over $7,000, which has put it up in its first place.

106
00:06:53,150 --> 00:06:57,170
Ejaaz:
So again, it's trading more like a quantitative analyst, which is taking wins

107
00:06:57,170 --> 00:07:01,310
Ejaaz:
when it can and taking losses that are incredibly small.

108
00:07:01,490 --> 00:07:04,830
Ejaaz:
Like notice this, right? Like normally we don't highlight the losses of a model.

109
00:07:05,750 --> 00:07:11,150
Ejaaz:
If you notice, all its red numbers are tiny compared to the profit numbers that

110
00:07:11,150 --> 00:07:12,450
Ejaaz:
it makes when it is right.

111
00:07:12,590 --> 00:07:15,350
Ejaaz:
So really, really strategic in its positioning.

112
00:07:15,530 --> 00:07:21,170
Ejaaz:
Now, if you compare that to the worst model, which is GPT-5,

113
00:07:21,710 --> 00:07:23,870
Ejaaz:
you'll notice a few things.

114
00:07:24,570 --> 00:07:29,630
Ejaaz:
Mainly, there's a bunch of green and red that you can see, mainly red.

115
00:07:29,990 --> 00:07:34,430
Ejaaz:
In its green positions where it's completed a trade, Josh, you'll notice something

116
00:07:34,430 --> 00:07:37,450
Ejaaz:
pretty different, which is the numbers are pretty small. Look at this.

117
00:07:37,710 --> 00:07:41,710
Ejaaz:
It's only booking tiny profits with each of its different trades,

118
00:07:41,850 --> 00:07:46,650
Ejaaz:
which tells me that it's not taking enough risk and it's closing the trades

119
00:07:46,650 --> 00:07:48,930
Ejaaz:
way too early for its thesis.

120
00:07:49,150 --> 00:07:53,790
Ejaaz:
So it's trading more like a cautious trader, like a lot of people that I know, actually.

121
00:07:54,210 --> 00:07:57,530
Ejaaz:
And then if you look at the model chat where it's talking to itself,

122
00:07:57,750 --> 00:07:58,930
Ejaaz:
you mentioned earlier...

123
00:07:59,850 --> 00:08:05,970
Ejaaz:
Here's an example. It goes, I'm still in the red with a minus 61% total return,

124
00:08:06,130 --> 00:08:10,350
Ejaaz:
but my ETH and XRP positions are showing gains, suggesting a slight upward momentum

125
00:08:10,350 --> 00:08:13,210
Ejaaz:
in those altcoins, despite the overall market downturn.

126
00:08:13,350 --> 00:08:16,650
Ejaaz:
So I'm holding strong and waiting for those profit targets to hit.

127
00:08:16,870 --> 00:08:21,330
Ejaaz:
And so you might think, huh, that's not too crazy. That sounds like a sensible strategy.

128
00:08:21,470 --> 00:08:27,290
Ejaaz:
If you look at its profit targets, Josh, it's like super small from where the

129
00:08:27,290 --> 00:08:31,150
Ejaaz:
price currently is, which means that even if it does hit those profit targets,

130
00:08:31,350 --> 00:08:33,210
Ejaaz:
it only ends up booking like 50 bucks.

131
00:08:33,670 --> 00:08:37,910
Ejaaz:
So overall, the reason why this model is underperformed is it hasn't taken enough

132
00:08:37,910 --> 00:08:42,950
Ejaaz:
risk whilst the winning models have taken either too much risk or just enough

133
00:08:42,950 --> 00:08:44,650
Ejaaz:
risk to put them ahead of the game.

134
00:08:44,750 --> 00:08:47,510
Josh:
There's a lot of notes in there that I think humans can take on just the stay

135
00:08:47,510 --> 00:08:49,070
Josh:
of psychology around trading markets.

136
00:08:49,570 --> 00:08:52,430
Josh:
And I'm sure if you kind of follow these models long enough,

137
00:08:52,530 --> 00:08:55,370
Josh:
you'll start to understand the patterns that perhaps you as a human should follow

138
00:08:55,370 --> 00:08:58,570
Josh:
and learn something from deep seek versus open ai being very conservative

139
00:08:58,570 --> 00:09:01,690
Josh:
but now that we've kind of laid out the foundation the framework of how this works there

140
00:09:01,690 --> 00:09:04,470
Josh:
there are two big questions that i'm really interested in answering one of

141
00:09:04,470 --> 00:09:08,550
Josh:
these is should i use this model to trade for me the other one is how can i

142
00:09:08,550 --> 00:09:11,970
Josh:
use this model to trade for me because listen i like a little bit of risk i

143
00:09:11,970 --> 00:09:15,350
Josh:
can deal with the downside in exchange for like a nice upside and it looks like

144
00:09:15,350 --> 00:09:18,790
Josh:
the odds are about split between all of these so the first question i think

145
00:09:18,790 --> 00:09:23,490
Josh:
i want to ask you just maybe i'll get your take first is like, is this a benchmark?

146
00:09:23,850 --> 00:09:28,670
Josh:
Is this real signal? Or is this kind of just a reality TV show?

147
00:09:28,790 --> 00:09:31,470
Josh:
Is this esports for AI models?

148
00:09:31,650 --> 00:09:35,890
Josh:
Is this just a fun way to kind of throw our intelligence at this lottery machine

149
00:09:35,890 --> 00:09:39,810
Josh:
that everyone loves to watch and see if it could beat us in the hope that one

150
00:09:39,810 --> 00:09:43,730
Josh:
day an AI will beat the system enough to give us an edge and actually make us

151
00:09:43,730 --> 00:09:48,230
Josh:
money personally as portfolio owners so what what do you think about

152
00:09:48,230 --> 00:09:52,170
Ejaaz:
That okay I'm gonna give you the same response Josh,

153
00:09:52,870 --> 00:09:55,970
Ejaaz:
And then I'm going to give you the optimist's approach. Oh, yeah.

154
00:09:56,130 --> 00:09:56,850
Josh:
Bring it on. Let's hear it.

155
00:09:57,370 --> 00:10:04,770
Ejaaz:
The sane response to this is this experiment is way too tiny to make any kind

156
00:10:04,770 --> 00:10:06,510
Ejaaz:
of major financial decision on.

157
00:10:06,630 --> 00:10:12,630
Ejaaz:
And you would be stupid to risk putting your money with an AI model to trade for you.

158
00:10:12,930 --> 00:10:17,890
Ejaaz:
Incredibly stupid. Why? Well, this is one experiment. It's six models.

159
00:10:18,430 --> 00:10:23,150
Ejaaz:
Have you replicated those models? Like, what if you had 10 of the same models

160
00:10:23,150 --> 00:10:25,750
Ejaaz:
trading the same amount of money? Would they make the same trades? Probably not.

161
00:10:26,030 --> 00:10:30,030
Ejaaz:
And actually, the founder of this experiment highlights this problem that you

162
00:10:30,030 --> 00:10:33,770
Ejaaz:
speak about, which is, is this just skill versus noise?

163
00:10:33,930 --> 00:10:36,650
Ejaaz:
And the point he makes in this tweet is like, of course it is, right?

164
00:10:36,730 --> 00:10:40,590
Ejaaz:
Because this is such a limited data set. And he goes on to explain that they're

165
00:10:40,590 --> 00:10:43,730
Ejaaz:
going to be doing experiments which involve like more of these models doing

166
00:10:43,730 --> 00:10:46,050
Ejaaz:
the same kind of thing. So you can get statistical significance.

167
00:10:46,230 --> 00:10:50,470
Ejaaz:
So the logic answer is, yes, it's insane. But the optimist take,

168
00:10:50,750 --> 00:10:53,810
Ejaaz:
Josh, and I have to give the optimist take, is...

169
00:10:55,030 --> 00:11:00,890
Ejaaz:
This is giving us, or rather giving the public unparalleled access to data to

170
00:11:00,890 --> 00:11:03,530
Ejaaz:
which they never would have gotten access to in the first place,

171
00:11:03,690 --> 00:11:07,870
Ejaaz:
which is they can take this training data and not take it too seriously,

172
00:11:08,050 --> 00:11:12,730
Ejaaz:
but use it to teach themselves what maybe not to do or what maybe not to trade

173
00:11:12,730 --> 00:11:14,450
Ejaaz:
with. How about you? Do you have a different take?

174
00:11:15,010 --> 00:11:18,110
Josh:
There's a couple of different perspectives I have on this because there's the

175
00:11:18,110 --> 00:11:21,170
Josh:
fun speculative side of things, the gambling, the investing,

176
00:11:21,410 --> 00:11:22,210
Josh:
whatever you want to call it.

177
00:11:22,210 --> 00:11:25,910
Josh:
And then there's the actual technical benchmarking part of this that we spoke

178
00:11:25,910 --> 00:11:29,070
Josh:
about briefly in the last episode, which one of the things I was really excited

179
00:11:29,070 --> 00:11:32,750
Josh:
about when this came out was the idea of having a real-world benchmark that

180
00:11:32,750 --> 00:11:35,710
Josh:
operated in dynamic conditions that cannot be gamified.

181
00:11:36,030 --> 00:11:38,690
Josh:
So a lot of these benchmarks, this is the way you evaluate AI models,

182
00:11:38,950 --> 00:11:41,090
Josh:
they are done based on a fixed problem set.

183
00:11:41,210 --> 00:11:44,310
Josh:
And a lot of times when you're training an AI model, these big labs can do tricks

184
00:11:44,310 --> 00:11:49,010
Josh:
to gamify these benchmarks. with this case and using real world data and real

185
00:11:49,010 --> 00:11:51,910
Josh:
world markets, you can actually put them into the real world.

186
00:11:52,050 --> 00:11:54,670
Josh:
And there's no way to gamify these benchmarks because if there was,

187
00:11:54,810 --> 00:11:56,710
Josh:
everyone would be rich and you'd be able to predict markets.

188
00:11:57,210 --> 00:12:01,630
Josh:
To that point, though, there is a lot of problems with using this as a benchmark

189
00:12:01,630 --> 00:12:04,290
Josh:
because, I mean, one is the fixed data set, like he mentioned,

190
00:12:04,470 --> 00:12:07,690
Josh:
is that this has only been around for one to two weeks. We need a lot more data to confirm this.

191
00:12:08,130 --> 00:12:12,810
Josh:
The second is that this isn't really a very holistic approach to investing and

192
00:12:12,810 --> 00:12:17,550
Josh:
to gambling because it really doesn't have all of the data required to make good decisions.

193
00:12:17,550 --> 00:12:21,970
Josh:
It's only analyzing the price action and the volumes and whatever technical

194
00:12:21,970 --> 00:12:26,130
Josh:
specs you can see on a single page without understanding the context around

195
00:12:26,130 --> 00:12:29,630
Josh:
the moves. So let's say that Bitcoin's encryption got hacked.

196
00:12:29,810 --> 00:12:33,570
Josh:
It would have, and Bitcoin falls 50%, it has no idea why Bitcoin is going down

197
00:12:33,570 --> 00:12:36,950
Josh:
50%. And because of that, it's a huge disadvantage that it doesn't know how to trade.

198
00:12:37,270 --> 00:12:40,270
Josh:
Now, granted, these are unlocks. These are things that will change.

199
00:12:40,430 --> 00:12:44,790
Josh:
And I assume the natural progression of this will lead towards more of a steady state benchmark.

200
00:12:44,990 --> 00:12:47,830
Josh:
But it is a very tricky thing because markets are so unpredictable.

201
00:12:48,050 --> 00:12:50,790
Josh:
So is this a viable benchmark? I don't know.

202
00:12:51,010 --> 00:12:54,510
Josh:
Probably I'm leaning towards no because market conditions change a lot.

203
00:12:54,650 --> 00:12:56,110
Josh:
It's not quite there with the capabilities.

204
00:12:56,610 --> 00:13:01,510
Josh:
The other part of me is so stoked about this because the same way we love watching

205
00:13:01,510 --> 00:13:05,090
Josh:
esports or we love watching, a big thing on Twitch right now is gamblers.

206
00:13:05,250 --> 00:13:07,230
Josh:
You guys, I don't know if you've seen these in real life. People will sit there

207
00:13:07,230 --> 00:13:11,310
Josh:
and play, like they'll gamble blackjack on a live stream and people will just

208
00:13:11,310 --> 00:13:14,030
Josh:
watch them play virtual. Yeah, those types of services.

209
00:13:14,470 --> 00:13:21,170
Josh:
This in very much feels like an early prototype for a new type of fun form of

210
00:13:21,170 --> 00:13:24,510
Josh:
entertainment, which could be something where it's just, it's high stakes trading.

211
00:13:24,510 --> 00:13:28,830
Josh:
Imagine if this was done with $10 million per wallet and you got to watch these

212
00:13:28,830 --> 00:13:30,390
Josh:
AIs trade and there was real money on the line.

213
00:13:30,530 --> 00:13:32,810
Josh:
This feels sort of like a form of,

214
00:13:33,290 --> 00:13:36,550
Josh:
almost e-sport entertainment where I could see competing labs builds,

215
00:13:36,950 --> 00:13:40,950
Josh:
competing AI models to trade markets, and winners are given access to certain prizes.

216
00:13:41,130 --> 00:13:43,670
Josh:
In terms of trading for myself, which is the last point I'm going to make on

217
00:13:43,670 --> 00:13:47,210
Josh:
this, I am not very excited to take on these risks.

218
00:13:47,330 --> 00:13:49,610
Josh:
For the same reason, I'm not really excited to bet on sports.

219
00:13:49,750 --> 00:13:53,390
Josh:
And I imagine my opinions vary a lot from others, but this is very much a gamble.

220
00:13:53,530 --> 00:13:56,310
Josh:
There's no way you can skew this in which it is not a gamble.

221
00:13:56,650 --> 00:13:59,810
Josh:
The interesting part is there's a near perfect data split between them.

222
00:13:59,890 --> 00:14:02,330
Josh:
There's two big winners, two big losers. The rest are kind of sitting around the median.

223
00:14:02,730 --> 00:14:05,850
Ejaaz:
Okay, but I'm going to push back on you a bit here, Josh.

224
00:14:06,010 --> 00:14:11,110
Ejaaz:
The earlier point you made was it doesn't have access to all the necessary data

225
00:14:11,110 --> 00:14:13,710
Ejaaz:
that it might need to make more informed trades.

226
00:14:13,910 --> 00:14:19,270
Ejaaz:
And I would argue, well, isn't the whole point of the benchmark, can you make money?

227
00:14:19,490 --> 00:14:25,110
Ejaaz:
And the fact that two of these models have made over 100% returns in less than

228
00:14:25,110 --> 00:14:29,110
Ejaaz:
a week or just over a week is proof that it can make money to some extent.

229
00:14:29,350 --> 00:14:31,270
Ejaaz:
The second point I'll make is.

230
00:14:32,060 --> 00:14:36,340
Ejaaz:
We throw around the term like gambling, which is actually what I would say the

231
00:14:36,340 --> 00:14:39,180
Ejaaz:
majority of these models in this experiment are doing.

232
00:14:39,380 --> 00:14:43,940
Ejaaz:
But they are one or two models that are actually way more strategic and trade

233
00:14:43,940 --> 00:14:48,340
Ejaaz:
much, much better than the average trader that you trade against, if that makes sense.

234
00:14:48,480 --> 00:14:52,820
Ejaaz:
So if we take DeepSeek, which is the number one model, if you look at its trades,

235
00:14:52,980 --> 00:14:59,560
Ejaaz:
at an initial glance, you might see that it's using 25x leverage and be like, that is so ridiculous.

236
00:14:59,760 --> 00:15:03,680
Ejaaz:
I'm not even going to pay attention to this, right? But if you dig into the

237
00:15:03,680 --> 00:15:09,160
Ejaaz:
position that it holds under 25X leverage, you'll notice that it's actually not at 25X.

238
00:15:09,300 --> 00:15:13,380
Ejaaz:
It's using only a small amount of its capital to do a very specific trade over

239
00:15:13,380 --> 00:15:18,260
Ejaaz:
like a five to 10 minute period, which automatically makes it a much more strategic

240
00:15:18,260 --> 00:15:23,820
Ejaaz:
technical trader than the average trader than that is just gambling their money away.

241
00:15:24,060 --> 00:15:28,560
Ejaaz:
But the point you made around it being fair distribution, and this is my last

242
00:15:28,560 --> 00:15:33,200
Ejaaz:
counterpoint to you, Josh, you pointed out that it seems to be very even distribution, right?

243
00:15:33,260 --> 00:15:37,540
Ejaaz:
You've got two at the top, two at the bottom, and two right bang in the middle, right?

244
00:15:37,880 --> 00:15:43,840
Ejaaz:
I wonder whether actually GPT and Gemini are actually the best traders,

245
00:15:44,040 --> 00:15:47,820
Ejaaz:
even though they're at the bottom, if you just inversely traded them.

246
00:15:48,630 --> 00:15:52,350
Ejaaz:
It's it's it's it's zero sum. And it's the point that the founder of the experiment

247
00:15:52,350 --> 00:15:54,330
Ejaaz:
makes right here where he goes markets are zero sum.

248
00:15:54,410 --> 00:15:58,770
Ejaaz:
If you find a strategy that consistently loses money, it's just as good as finding

249
00:15:58,770 --> 00:16:00,690
Ejaaz:
one that makes money. Just do the opposite.

250
00:16:01,550 --> 00:16:04,730
Josh:
Yeah, absolutely. And it'll take time to for these to play out because I imagine

251
00:16:04,730 --> 00:16:08,050
Josh:
there is they are kind of tuned for a specific type of trading.

252
00:16:08,450 --> 00:16:10,870
Josh:
So in the case a few weeks ago, there was a huge liquidation event in crypto.

253
00:16:11,350 --> 00:16:15,270
Josh:
Things go down. Well, in a down market, some might trade way better than others.

254
00:16:15,610 --> 00:16:19,310
Josh:
And the point you made about leverage, it got me thinking it was really interesting. like

255
00:16:19,310 --> 00:16:22,050
Josh:
because I don't use 20x leverage and I imagine most people

256
00:16:22,050 --> 00:16:24,710
Josh:
don't but with AIs they they're able to hold a

257
00:16:24,710 --> 00:16:27,550
Josh:
lot more in their memory and it reminded me of the the

258
00:16:27,550 --> 00:16:30,330
Josh:
AlphaGo case Google where an AI

259
00:16:30,330 --> 00:16:34,890
Josh:
model played a professional at AlphaGo and there was one move that was way outside

260
00:16:34,890 --> 00:16:39,610
Josh:
of the expected data set move 37 which was the famous move and it turned out

261
00:16:39,610 --> 00:16:43,470
Josh:
that that was a move that no human could have ever seen but it resulted in the

262
00:16:43,470 --> 00:16:48,410
Josh:
AI winning the game and it kind of broke open the rule set and expectations

263
00:16:48,410 --> 00:16:49,790
Josh:
around the game of AlphaGo.

264
00:16:49,930 --> 00:16:53,990
Josh:
And I wonder if we'll get some sort of breakthrough with that around AI trading,

265
00:16:54,150 --> 00:16:57,950
Josh:
where we have this very fixed set of outcomes that we do and strategies that we do.

266
00:16:58,190 --> 00:17:01,970
Josh:
But AIs might actually just destroy a lot of these barriers that we,

267
00:17:02,110 --> 00:17:05,770
Josh:
or perceived barriers that we have in exchange for these like really weird strategies,

268
00:17:05,930 --> 00:17:06,910
Josh:
like 20x longing everything.

269
00:17:07,090 --> 00:17:10,190
Josh:
So I don't know, there's a lot to talk about when it comes to this.

270
00:17:10,290 --> 00:17:13,230
Josh:
But another of the big questions that I want to answer, because this was something

271
00:17:13,230 --> 00:17:17,790
Josh:
I was interested in, is how can I use these for myself? Let's say I am a degenerate gambler.

272
00:17:17,970 --> 00:17:20,950
Josh:
I want to make two acts in a week or at least give myself a chance to do it.

273
00:17:21,110 --> 00:17:23,790
Josh:
I want to know, how can I use these models to trade for myself?

274
00:17:23,930 --> 00:17:25,430
Josh:
What do I need to do to get involved in this?

275
00:17:26,090 --> 00:17:30,370
Ejaaz:
Yeah, it has been the number one question and feedback that we got on our previous

276
00:17:30,370 --> 00:17:33,370
Ejaaz:
episode from our listeners is, I've got it up on a tweet here.

277
00:17:33,470 --> 00:17:37,090
Ejaaz:
How do I profit from this trading? How do I do this for myself?

278
00:17:37,310 --> 00:17:43,390
Ejaaz:
I have one simple answer for you, which is the platform that these AI models

279
00:17:43,390 --> 00:17:47,430
Ejaaz:
are trading their tens of thousands of dollars on And Josh is public.

280
00:17:48,100 --> 00:17:53,760
Ejaaz:
It's open. It's available for anyone to log onto right now and see what trades

281
00:17:53,760 --> 00:17:58,740
Ejaaz:
each of these models open up when they close it and what their inevitable strategy is.

282
00:17:58,820 --> 00:18:02,480
Ejaaz:
I'm going to give you an example here with the number one model,

283
00:18:02,760 --> 00:18:06,180
Ejaaz:
DeepSeek, which has doubled its money in just over a week.

284
00:18:06,480 --> 00:18:10,700
Ejaaz:
The platform that these models are trading on is called Hyperliquid. It's a blockchain.

285
00:18:11,060 --> 00:18:13,520
Ejaaz:
Blockchains are known for being transparent and open. The fact that you can

286
00:18:13,520 --> 00:18:16,160
Ejaaz:
kind of see all the things that these models are doing.

287
00:18:16,360 --> 00:18:19,660
Ejaaz:
And if I just scroll down over here, you'll notice a few things.

288
00:18:19,840 --> 00:18:24,640
Ejaaz:
Number one, these are all the positions that this model currently has open.

289
00:18:24,820 --> 00:18:28,840
Ejaaz:
This isn't made up, this isn't on someone's word and you have to trust them.

290
00:18:29,080 --> 00:18:31,520
Ejaaz:
This is all verifiable using a blockchain.

291
00:18:31,780 --> 00:18:35,200
Ejaaz:
So the whole point of a blockchain is that you are able to verify what is real

292
00:18:35,200 --> 00:18:38,280
Ejaaz:
and what is not real without having to trust someone on this.

293
00:18:38,400 --> 00:18:42,520
Ejaaz:
You can look into its holdings and you can see how much that it currently holds,

294
00:18:42,520 --> 00:18:44,760
Ejaaz:
like in terms of like money or in terms of like dollars.

295
00:18:44,840 --> 00:18:48,840
Ejaaz:
You can also look at the trades that it's completed as well.

296
00:18:48,940 --> 00:18:54,400
Ejaaz:
So the point I'm making is you can't currently go onto DeepSeek and say.

297
00:18:54,740 --> 00:19:00,100
Ejaaz:
Hey, can I give you $10,000 and you go make me money like I've just heard about on this video?

298
00:19:00,340 --> 00:19:05,040
Ejaaz:
It won't be able to work. But what you can do is you can go onto a site like

299
00:19:05,040 --> 00:19:08,100
Ejaaz:
this and look at the trades that they're making yourself.

300
00:19:08,360 --> 00:19:12,440
Ejaaz:
And again, this is not financial advice, potentially copy those trades or make

301
00:19:12,440 --> 00:19:16,000
Ejaaz:
those trades yourself in order to trade like how these models are.

302
00:19:16,280 --> 00:19:21,800
Ejaaz:
Now, the last point I'll make is the founder of this experiment has all the

303
00:19:21,800 --> 00:19:26,880
Ejaaz:
intention to allow you and me to trade with these models directly.

304
00:19:27,040 --> 00:19:30,300
Ejaaz:
That is, you can speak to the model, give it your money, and it can do that.

305
00:19:30,420 --> 00:19:33,200
Ejaaz:
And to your point, Josh, it's up to you whether you want to do it from an entertainment

306
00:19:33,200 --> 00:19:36,620
Ejaaz:
basis where it's just all gambling or whether you actually want to invest serious money into this.

307
00:19:36,720 --> 00:19:41,200
Ejaaz:
That will come in later iterations, probably around a couple of months from now.

308
00:19:42,260 --> 00:19:45,360
Josh:
So there's kind of two ways to copy trade. There's one you could actually copy trade.

309
00:19:45,600 --> 00:19:48,200
Josh:
Or another way to get into it is if you're feeling a little more ambitious,

310
00:19:48,200 --> 00:19:52,920
Josh:
you can actually generate one of these yourself. You can create like a mini alpha arena bot.

311
00:19:53,260 --> 00:19:56,200
Josh:
The way to do that is pretty simple. I was kind of curious. I was like,

312
00:19:56,260 --> 00:19:57,660
Josh:
what does it take to actually build one of these things?

313
00:19:57,940 --> 00:20:01,440
Josh:
You choose your fighter. So you pick a model that you want. And then you kind

314
00:20:01,440 --> 00:20:03,760
Josh:
of pipe market data in from Hyperliquid that you showed.

315
00:20:03,880 --> 00:20:06,660
Josh:
So Hyperliquid has this endpoint, not to get too technical, but you can kind

316
00:20:06,660 --> 00:20:08,040
Josh:
of feed the model this data.

317
00:20:08,040 --> 00:20:11,460
Josh:
And then the difficult part, the tricky part, and the thing that we haven't

318
00:20:11,460 --> 00:20:15,160
Josh:
been able to talk about because we don't actually know, is the system prompts

319
00:20:15,160 --> 00:20:19,180
Josh:
behind the recursive loop that happens as these models receive this data.

320
00:20:19,460 --> 00:20:22,680
Josh:
So the way it works is you choose a model, you give it feedback,

321
00:20:22,860 --> 00:20:26,820
Josh:
or you give it data, and then you write a prompt for the model to run in between

322
00:20:26,820 --> 00:20:28,360
Josh:
each iteration of receiving new data.

323
00:20:28,660 --> 00:20:34,040
Josh:
What that prompt says is how it makes a decision. The problem is that is all of the value.

324
00:20:34,160 --> 00:20:36,400
Josh:
All of the value sits within that prompt. And the prompt is just written in

325
00:20:36,400 --> 00:20:39,520
Josh:
plain English. Like we always say, the hottest language in the world is English.

326
00:20:39,700 --> 00:20:43,440
Josh:
So there is some string of words that you as a developer or just a novice can

327
00:20:43,440 --> 00:20:46,240
Josh:
write into this to generate you more money than other people.

328
00:20:46,320 --> 00:20:48,760
Josh:
So I encourage people who are feeling a little ambitious to actually try this

329
00:20:48,760 --> 00:20:53,040
Josh:
out, to write a prompt yourself and see if you can get a bot to try and kind of trade like this.

330
00:20:53,100 --> 00:20:55,700
Josh:
And if we ever do get the system prompts from this, we will certainly share

331
00:20:55,700 --> 00:21:00,100
Josh:
because it'll be fascinating to see the behind the scenes and what happens to

332
00:21:00,100 --> 00:21:02,900
Josh:
produce those outputs that we were reading a little bit earlier in the show.

333
00:21:02,900 --> 00:21:05,500
Josh:
So that's kind of how you can get involved if you're interested.

334
00:21:05,840 --> 00:21:08,620
Josh:
Copy trade, maybe inverse copy trade. I think if I were to do this,

335
00:21:08,700 --> 00:21:11,720
Josh:
I'd probably go to ChatGPT's trading history, sit there refreshing,

336
00:21:11,800 --> 00:21:13,780
Josh:
and then just hit the opposite of whatever they decide to do.

337
00:21:13,880 --> 00:21:14,660
Josh:
That seems pretty consistent.

338
00:21:15,300 --> 00:21:18,320
Josh:
But yeah, that is how this whole thing works. It's pretty fascinating.

339
00:21:18,480 --> 00:21:23,080
Josh:
It's been amazing how the internet has kind of gotten behind this and it has spread like wildfire.

340
00:21:23,480 --> 00:21:28,260
Ejaaz:
The thing is, I don't think they'll ever make the system prompt for this or

341
00:21:28,260 --> 00:21:31,840
Ejaaz:
any other successful trading AI publicly available.

342
00:21:32,060 --> 00:21:36,600
Ejaaz:
The reason is that's the secret sauce. And why would you let everyone have access

343
00:21:36,600 --> 00:21:41,360
Ejaaz:
to it when you can use it yourself and make a ton of money? And that's what D.D.

344
00:21:41,480 --> 00:21:45,160
Ejaaz:
Das demonstrates in this tweet. He says, I've heard six people tell me they're

345
00:21:45,160 --> 00:21:50,140
Ejaaz:
doing this using Vibe coding apps to algorithmically trade on the stock or crypto market.

346
00:21:50,360 --> 00:21:53,640
Ejaaz:
But the thing to remember is this is a dangerous game to play.

347
00:21:54,160 --> 00:21:57,440
Ejaaz:
Algo trading is the last thing I expect AI to democratize.

348
00:21:57,480 --> 00:22:01,180
Ejaaz:
The point being, if you have a successful algo, you're probably not going to

349
00:22:01,180 --> 00:22:03,260
Ejaaz:
democratize access to it. Full stop.

350
00:22:03,740 --> 00:22:07,180
Ejaaz:
That being said, I do think you can't stop AI trading.

351
00:22:07,950 --> 00:22:12,070
Ejaaz:
Entering the investment and financial scene. I think it's going to make people

352
00:22:12,070 --> 00:22:14,390
Ejaaz:
way more financially literate than they already are.

353
00:22:14,650 --> 00:22:18,110
Ejaaz:
Look how ChatGPT has made so many people proficient in other things that they

354
00:22:18,110 --> 00:22:19,770
Ejaaz:
had previously no idea about.

355
00:22:20,030 --> 00:22:23,610
Ejaaz:
So I think AI is inevitably going to be integrated. It's going to make markets way more efficient.

356
00:22:23,770 --> 00:22:26,910
Ejaaz:
It's going to give you access to knowledge that can make you do trades that

357
00:22:26,910 --> 00:22:29,890
Ejaaz:
you otherwise wouldn't have known of five minutes prior to that.

358
00:22:30,010 --> 00:22:32,290
Ejaaz:
But will it make you a super trading god?

359
00:22:32,670 --> 00:22:35,950
Ejaaz:
No. I think that it'll evolve the trading scene, though.

360
00:22:36,030 --> 00:22:39,390
Ejaaz:
I think the hedge funds that are successful today will look very different to

361
00:22:39,390 --> 00:22:44,650
Ejaaz:
the hedge funds that are successful in an AGI or AI world where AI is available

362
00:22:44,650 --> 00:22:45,530
Ejaaz:
pretty much everywhere.

363
00:22:46,010 --> 00:22:48,050
Josh:
Yeah, AI needs to be integrated into all these trading strategies.

364
00:22:48,190 --> 00:22:50,390
Josh:
So to me, it's no brainer that it will be.

365
00:22:50,530 --> 00:22:54,130
Josh:
The extent of that integration is kind of what is up for debate and what we'll

366
00:22:54,130 --> 00:22:56,690
Josh:
see in this answering the big question.

367
00:22:56,790 --> 00:22:58,930
Josh:
Is this a benchmark or is this just a reality show?

368
00:22:59,130 --> 00:23:02,030
Josh:
And is this just a toy or is this real technology baked into this?

369
00:23:02,190 --> 00:23:04,690
Josh:
It seems as if AI will slowly creep its way in.

370
00:23:04,930 --> 00:23:08,510
Josh:
I'm looking forward to tracking this. It ends next week, so we'll probably add

371
00:23:08,510 --> 00:23:11,230
Josh:
some follow-ups on this first trading competition, the result to how it turns out.

372
00:23:11,350 --> 00:23:15,110
Josh:
But that is a part two in our little saga of this crazy weird thing that's happening

373
00:23:15,110 --> 00:23:16,390
Josh:
in AI crypto trading world.

374
00:23:16,770 --> 00:23:19,910
Josh:
I hope you enjoyed this episode. You enjoyed the last one a lot.

375
00:23:19,970 --> 00:23:22,750
Josh:
It was amazing. So thank you for watching, sharing with your friends, liking and commenting.

376
00:23:22,930 --> 00:23:26,230
Josh:
It really goes a long way. It's been amazing to see the growth and support from

377
00:23:26,230 --> 00:23:28,050
Josh:
everybody watching. So thank you for that.

378
00:23:28,250 --> 00:23:31,450
Josh:
More of this to come. We have a couple more episodes slated for this week that

379
00:23:31,450 --> 00:23:35,030
Josh:
are pretty exciting about autonomy and robotics and just a whole bunch of interesting

380
00:23:35,030 --> 00:23:36,430
Josh:
things. So stick around for that.

381
00:23:36,690 --> 00:23:39,250
Josh:
We'll be back in the next one. And I just, I think that's it.

382
00:23:39,330 --> 00:23:40,290
Josh:
Any final parting words?

383
00:23:40,630 --> 00:23:43,010
Ejaaz:
That's it. Let us know what you want to hear more of as well.

384
00:23:43,090 --> 00:23:45,330
Ejaaz:
If you're loving this trading stuff and you have some other ideas,

385
00:23:45,430 --> 00:23:46,210
Ejaaz:
let us know in the comments.

386
00:23:47,210 --> 00:23:50,110
Josh:
Absolutely. All right. Well, that's been another episode of Limitless. Thank you for tuning in.