WEBVTT - DIY Hedge Funds Unleash AI to Crack Wall Street’s Secret Code 0:00:02.520 --> 0:00:07.040 Bloomberg Audio Studios, Podcasts, radio News. 0:00:07.960 --> 0:00:12.000 You're listening to Bloomberg BusinessWeek with Carol Masser and Tim 0:00:12.080 --> 0:00:14.200 Stenoveek on Bloomberg Radio. 0:00:14.640 --> 0:00:16.800 It's one of the most read stories on the Bloomberg terminal. 0:00:16.840 --> 0:00:20.239 It's about how Jenai is transforming retail investing. It's enabling 0:00:20.320 --> 0:00:24.520 individual traders to build and execute institutional grade quant strategies 0:00:24.560 --> 0:00:28.480 that have once been reserved for major Wall Street hedge funds. 0:00:28.680 --> 0:00:31.760 Bernard Guider is one half of the duo behind this story. 0:00:31.760 --> 0:00:34.199 He's US Options reporter for Bloomberg News. He joins us 0:00:34.200 --> 0:00:38.680 here in the Bloomberg Interactive Broker's Studio. Congratulations on the story. 0:00:39.320 --> 0:00:43.120 I'm struck by all of the individuals you highlighted in 0:00:43.200 --> 0:00:45.760 the story because there's this pattern that has emerged. They 0:00:45.800 --> 0:00:49.360 were doing one thing before they quit that job to 0:00:49.520 --> 0:00:52.760 vibe code or create some sort of AI solution to 0:00:52.880 --> 0:00:57.760 trading that they think can get them of salary. How 0:00:57.760 --> 0:00:58.400 are they doing it? 0:00:59.080 --> 0:01:01.480 Yeah, I think it. It's really indicative of that broader 0:01:01.480 --> 0:01:04.560 trend in retail investing we've seen since the pandemic, with 0:01:04.680 --> 0:01:08.560 people putting aside what they were doing before and jumping 0:01:08.600 --> 0:01:11.120 in and trying to make a living out of trading, really, 0:01:11.200 --> 0:01:13.280 and these are some of the more professional side of 0:01:14.080 --> 0:01:17.120 the retail investing world. I think we often like to 0:01:17.200 --> 0:01:20.280 characterize retail investors as having this really high risk appetite, 0:01:20.319 --> 0:01:22.039 But if you're doing this as a job, actually your 0:01:22.120 --> 0:01:24.319 risk appetite is going to be pretty low and reasonable. 0:01:24.360 --> 0:01:25.600 So I think one of the guys had he put 0:01:25.600 --> 0:01:28.360 five percent of his portfolio into this kind of magic 0:01:28.400 --> 0:01:31.960 money machine that's trading for him. Somebody else was doing 0:01:32.000 --> 0:01:34.880 one trade a day on robin Hood, but it was 0:01:34.959 --> 0:01:36.920 an agentic AI, and he goes and does his day 0:01:37.000 --> 0:01:39.440 job in it and lets this thing in the background. 0:01:39.480 --> 0:01:41.720 He'll get a notification on his phone sometime during the 0:01:41.760 --> 0:01:44.360 day being like your trade went through. And that's just 0:01:44.680 --> 0:01:46.760 the way these people are doing. I mean, there's another 0:01:46.760 --> 0:01:49.160 guy who's got to stop loss at ten percent once 0:01:49.200 --> 0:01:51.880 his portfolio. Once his position drops by more than ten percent, 0:01:51.880 --> 0:01:53.840 that stop lost kicks in automatically. And that's the same 0:01:53.840 --> 0:01:55.360 as if you were in a seat at a hedge fund. 0:01:55.560 --> 0:01:57.640 It's a similar kind of system. And I think what 0:01:57.680 --> 0:01:59.520 the AI has done is it's grabbed all of that 0:01:59.560 --> 0:02:01.760 information and from all of the best practices across Wall 0:02:01.800 --> 0:02:04.120 Street and is now giving that to ordinary people. 0:02:04.280 --> 0:02:06.800 So right, I think folks have always tried to kind 0:02:06.800 --> 0:02:09.200 of crack the code right of what goes on at 0:02:09.280 --> 0:02:11.960 some of these elite firms and these hedge funds. But 0:02:12.080 --> 0:02:14.720 AI has given them a tool that's getting them closer 0:02:14.760 --> 0:02:15.839 to it. 0:02:15.840 --> 0:02:16.280 It has. 0:02:16.360 --> 0:02:19.360 But bear in mind that this is giving people the 0:02:19.400 --> 0:02:21.280 same kind of tools that maybe Wall Street had eight 0:02:21.360 --> 0:02:24.760 years ago, which probably means that Wall Street is way, way, 0:02:24.840 --> 0:02:27.920 way further ahead of some of these people. But it 0:02:28.000 --> 0:02:31.959 definitely makes it definitely makes it more viable to try 0:02:32.000 --> 0:02:34.840 and execute trades automatically using AI, and more and more 0:02:34.840 --> 0:02:37.560 people seem to be taking that on. I'm Vlad the 0:02:37.680 --> 0:02:41.840 CEO of Robin Hood posted that one hundred thousand people 0:02:41.880 --> 0:02:44.519 had plugged into their Argentic AI system. They're one of 0:02:44.560 --> 0:02:45.760 the brokers that offers this is. 0:02:45.960 --> 0:02:49.240 Other brokers operate apart from Robin who you read about, 0:02:49.240 --> 0:02:54.040 Futoo Holdings, Moomu Public Holdings all rolling this out. Are 0:02:54.080 --> 0:02:58.760 there notable holdouts of retail trading firms or firms that 0:02:58.800 --> 0:03:00.880 are popular with retail traders are that are saying no 0:03:00.960 --> 0:03:01.959 to this tech as of now? 0:03:02.440 --> 0:03:05.480 I wouldn't I think some firms are more cautious, So 0:03:05.840 --> 0:03:08.760 traditionally Charles Schwab is generally on the more cautious end 0:03:08.800 --> 0:03:10.800 in terms of in terms of this stuff, they do 0:03:10.840 --> 0:03:13.639 have API technology you can plug into, plug a system 0:03:13.639 --> 0:03:16.320 into Schwab, but in terms of like enabling people to 0:03:16.400 --> 0:03:19.280 kind of trade automatically through AI, I definitely point to them. 0:03:19.280 --> 0:03:20.960 And the big one is really the banks. 0:03:22.080 --> 0:03:22.480 A lot of. 0:03:22.440 --> 0:03:25.919 People will use their use that the brokerage that they're 0:03:25.960 --> 0:03:28.920 bank owns to trade through, and those really aren't offering 0:03:29.000 --> 0:03:31.000 that at this point. They're still kind of pushing people 0:03:31.000 --> 0:03:33.440 towards more traditional wealth management services. 0:03:34.200 --> 0:03:36.600 I mean, is this the you know, we always talk 0:03:36.600 --> 0:03:39.200 about financial disruption, right, We talked about fintech and all 0:03:39.200 --> 0:03:41.280 these different things. But I feel like, is this the 0:03:41.360 --> 0:03:45.040 thing that really disrupts kind of Wall Street and trading 0:03:45.080 --> 0:03:47.360 as we know it. 0:03:47.480 --> 0:03:49.640 I think that's what some people in the industry would 0:03:49.680 --> 0:03:53.640 like to think. So Anthony Dennya, the we Will president, 0:03:53.680 --> 0:03:55.360 was quoted in the story as saying, this is zero 0:03:55.360 --> 0:03:57.320 commission two point zero. 0:03:57.120 --> 0:03:58.800 And I thought that was a remarkable thing for him 0:03:58.800 --> 0:04:00.920 to saying, because that changed for trading. 0:04:00.640 --> 0:04:02.000 It completely revolutionized it. 0:04:02.080 --> 0:04:04.520 So there is a I think it's I think you've 0:04:04.520 --> 0:04:06.320 got to kind of separate the people that are very 0:04:06.400 --> 0:04:08.640 much keen for just trading to happen. And Matt Levine 0:04:08.640 --> 0:04:11.600 makes this point in today's newsletter as well in Money Stuff, 0:04:11.680 --> 0:04:13.360 that there's a lot of people who are incentivized just 0:04:13.360 --> 0:04:15.680 to have loads and loads of trading happening, and they 0:04:15.680 --> 0:04:18.400 don't really care what happens to it. I think it 0:04:18.440 --> 0:04:20.120 can have a big impact. I think it's easy to 0:04:20.160 --> 0:04:23.200 overstate what AI does, but I think ultimately one of 0:04:23.200 --> 0:04:25.719 the big changes could actually be the business model, because 0:04:25.800 --> 0:04:28.159 ultimately the payment for order flow, which is this idea 0:04:28.200 --> 0:04:31.560 that you can get free commission free trading, and marketmakers 0:04:31.560 --> 0:04:35.400 are happy to get that. The more sophisticated that these 0:04:35.440 --> 0:04:38.400 retail traders become, the worse a business proposition it becomes 0:04:38.440 --> 0:04:40.719 to trade against them, and that could actually change the 0:04:40.760 --> 0:04:43.160 economics of some of the payment for order flow models. 0:04:43.240 --> 0:04:45.200 I'm laughing when you mentioned payment for order flow, because 0:04:45.240 --> 0:04:47.240 it was just what five years ago at this point, 0:04:47.279 --> 0:04:49.120 that we were talking about payment for order flow every 0:04:49.120 --> 0:04:51.279 single day, and it was seen as a really pretty 0:04:52.160 --> 0:04:55.360 controversial way for us to be able to do free 0:04:55.360 --> 0:04:59.560 trades and essays as retail participants. Okay, so I want 0:04:59.600 --> 0:05:00.320 to talk Oregon. 0:05:00.400 --> 0:05:02.600 Say just one thing though, like if Wall Street's always 0:05:02.720 --> 0:05:05.680 eight you know, eight steps ahead or whatever it is, 0:05:06.160 --> 0:05:06.640 So yeah, I. 0:05:06.600 --> 0:05:07.800 Don't think Jane Street is doing this. 0:05:09.200 --> 0:05:12.320 So is it mean that the retail investor, Yeah, may 0:05:12.360 --> 0:05:15.400 have some fun and do well, but there's always there 0:05:15.440 --> 0:05:19.159 are still ahead wels Like how does that impact this? 0:05:19.800 --> 0:05:21.040 Yeah, I think it's a great question. 0:05:21.560 --> 0:05:23.800 I would say that there are definitely there are definitely 0:05:23.839 --> 0:05:27.040 situations where Main Street is catching up with Wall Street. 0:05:27.880 --> 0:05:30.000 But ultimately, when it comes to market making, when it 0:05:30.000 --> 0:05:31.839 comes to the kind of stuff that Jane Street does 0:05:31.880 --> 0:05:36.200 and sit at our securities and yeah, Hudson River Trading 0:05:36.240 --> 0:05:38.719 in the equity market, Yes, those firms aren't many multiple 0:05:38.720 --> 0:05:41.800 steps ahead, and they're the main counterparty to retail in 0:05:41.880 --> 0:05:44.400 prediction markets as well. Now that in terms of Susquehanna 0:05:44.440 --> 0:05:47.080 is one of the big counter parties Jump Trading as well. 0:05:47.160 --> 0:05:49.640 So yeah, I mean retail is facing off against very 0:05:49.720 --> 0:05:51.039 very powerful professional firms. 0:05:51.120 --> 0:05:51.760 What I'm saying is that. 0:05:51.800 --> 0:05:53.880 Retail has a lot more tools in their arsenal than 0:05:53.920 --> 0:05:56.600 they used to, Okay, to play this game, and maybe 0:05:56.600 --> 0:05:57.680 it's less of an unfair fight. 0:05:57.720 --> 0:05:58.320 Than it used to be. 0:05:58.480 --> 0:06:01.599 We're speaking with Bernard Goiter, options reporter for Bloomberg News. 0:06:01.600 --> 0:06:05.080 He joins us here on set in New York. Bernard, 0:06:05.480 --> 0:06:08.960 I'm wondering about and my question will show that I 0:06:09.000 --> 0:06:11.720 am not a vibe coder, but one question that I've 0:06:11.720 --> 0:06:15.680 had wants trying to be I'll get there one day. 0:06:16.960 --> 0:06:21.320 If if retailers, or if retail participants rather are using 0:06:21.360 --> 0:06:29.960 the same tools to create agents to do this work 0:06:30.000 --> 0:06:33.600 on behalf of them, does that mean that they're sort 0:06:33.600 --> 0:06:37.040 of making the same trades as each other and there's 0:06:37.080 --> 0:06:39.800 going to be like less of an opportunity for them 0:06:39.839 --> 0:06:42.720 to sort of find something that's unique to what they 0:06:42.800 --> 0:06:44.599 actually believe. Does that make sense? 0:06:44.680 --> 0:06:45.279 Yes, it does. 0:06:45.600 --> 0:06:47.919 So there is this question of crowding of people going 0:06:48.000 --> 0:06:50.680 to make a load of the same same decisions. And 0:06:50.720 --> 0:06:53.400 funnily enough, that's also that something that happens on traditional 0:06:53.480 --> 0:06:55.160 markets with Wall Street, you see that with the pun 0:06:55.240 --> 0:06:58.640 shops all entering the same trades, whether that's equity dispersion 0:06:58.640 --> 0:07:01.320 trade is one example that very very popular trade that 0:07:01.360 --> 0:07:03.119 lots of the hedge funds all did at the same time. 0:07:03.680 --> 0:07:05.640 You have it with the in the treasury market. You 0:07:05.680 --> 0:07:08.520 have it with the with the basis trades between treasuries 0:07:08.520 --> 0:07:11.920 and treasury futures. So it's a common problem that exists, 0:07:11.920 --> 0:07:14.280 and maybe it's going to come more so on. But 0:07:14.560 --> 0:07:15.920 I think as long as as long as there are 0:07:15.960 --> 0:07:17.720 some as long as some people are bullish and some 0:07:17.760 --> 0:07:20.360 people are parished, then there should be enough enough kind 0:07:20.360 --> 0:07:21.000 of mix in there. 0:07:21.120 --> 0:07:23.440 It's sort of the same argument that people use about 0:07:23.840 --> 0:07:25.720 investing in index funds, right, or they're going to be 0:07:25.800 --> 0:07:27.920 too many index funds investors, then you won't actually see 0:07:27.960 --> 0:07:30.240 the daily movements. We have not seen that play out, 0:07:30.280 --> 0:07:33.160 and we still have pretty big daily movements in individual stocks. Yeah. 0:07:33.200 --> 0:07:35.880 Absolutely, Like I think, you know, all right, so this 0:07:35.960 --> 0:07:38.440 is the way everything's going. I just think about, like, 0:07:38.600 --> 0:07:42.240 so what happens in a crisis or do we. 0:07:42.800 --> 0:07:45.280 One issue is that some of the model sets that 0:07:45.320 --> 0:07:48.800 people are using, they don't go back that far. So 0:07:48.960 --> 0:07:51.480 if you're trading zero day options, that's only really been 0:07:51.840 --> 0:07:53.760 an asset class that people have been trading in and 0:07:53.760 --> 0:07:56.840 out of since twenty twenty two, that hasn't really seen 0:07:56.880 --> 0:08:00.240 a proper bet market yet, So you could have a 0:08:00.280 --> 0:08:02.960 situation a bit like the run up to two thousand 0:08:03.000 --> 0:08:05.320 and eight housing crisis, where none of the models were 0:08:05.320 --> 0:08:08.640 trained on a scenario where house prices get go down 0:08:08.720 --> 0:08:10.480 in the same way. None of these models are necessarily 0:08:10.520 --> 0:08:14.440 trained in a scenario where share prices fall significantly and 0:08:14.520 --> 0:08:18.040 then and then stay there. So that's definitely a concern 0:08:18.080 --> 0:08:21.160 for people. But people often kind of are interrogating these 0:08:21.160 --> 0:08:22.840 models and asking the right question. But you've got to 0:08:22.840 --> 0:08:25.520 sometimes you've got to push it. You've got to push 0:08:25.560 --> 0:08:27.000 it to be like, hey, did you think of this 0:08:27.120 --> 0:08:28.000 kind of thing? Right? 0:08:28.200 --> 0:08:30.680 That's interesting I would imagine I would have thought that 0:08:30.720 --> 0:08:34.360 the models would be tested or they're getting there. 0:08:34.800 --> 0:08:38.160 Well, it depends there only is there only? It depends 0:08:38.240 --> 0:08:40.160 what data set it's been trained. Yeah, and it depends 0:08:40.200 --> 0:08:42.800 on what questions you ask it. You as a human 0:08:42.880 --> 0:08:44.760 ultimately have to be the one that interrogates it. You 0:08:44.800 --> 0:08:46.800 have to you have a human As a human, you 0:08:46.800 --> 0:08:48.360 have to say, Okay, what happens if there's a two 0:08:48.360 --> 0:08:50.559 thousand and eight style crisis on this on this kind 0:08:50.559 --> 0:08:53.040 of model portfolio, or on this on this trading system. 0:08:53.600 --> 0:08:55.880 So ultimately a lot of it still goes back to 0:08:55.920 --> 0:08:58.280 the decisions being made by the individual, but instead of 0:08:58.320 --> 0:09:01.559 it being a human kind of picking specific trade. It's 0:09:01.600 --> 0:09:04.160 more about designing a model, so your job becomes a 0:09:04.160 --> 0:09:06.240 lot more like a quant at a hedge fund and 0:09:06.320 --> 0:09:09.000 much less like an individual punter kind of throwing darts 0:09:09.000 --> 0:09:09.680 at a dartboard. 0:09:09.920 --> 0:09:13.960 You just very briefly you got to see how these 0:09:13.960 --> 0:09:16.160 models work in the real world and how these agents 0:09:16.160 --> 0:09:18.440 work in the real world. Is there how much further 0:09:18.520 --> 0:09:20.000 does it go in the next couple of years. 0:09:20.360 --> 0:09:21.000 I think. 0:09:22.559 --> 0:09:26.520 I was very surprised by how sophisticated these these these 0:09:26.520 --> 0:09:28.960 trading tools have become, and I think it will go 0:09:29.040 --> 0:09:30.400 a lot further. I think it's going to become a 0:09:30.440 --> 0:09:33.080 lot more mainstream, partly because one of the one of 0:09:33.080 --> 0:09:36.080 the downsides of trading is that it's very labor intensive 0:09:36.080 --> 0:09:39.400 and actually quite boring just sitting there being like, oh, 0:09:39.400 --> 0:09:42.320 it's hit the Bollinger band, Like that's not that fun 0:09:42.400 --> 0:09:45.000 for most people, like unless you're kind of a strange person. 0:09:45.280 --> 0:09:47.600 But I like the idea if I can if it's 0:09:47.640 --> 0:09:49.920 strange here, yeah, but you know, apart. 0:09:49.640 --> 0:09:52.480 From you know, because I'm sure a couple of our listeners, 0:09:52.520 --> 0:09:53.680 that's not actually that exciting. 0:09:53.720 --> 0:09:54.480 Whereas if you can. 0:09:54.400 --> 0:09:56.319 Get a lot of that boring stuff being done for 0:09:56.400 --> 0:09:59.160 you by an AI and it makes money in a 0:09:59.200 --> 0:09:59.760 safe way. 0:10:00.200 --> 0:10:01.960 Yeah. I think it's going to become superbul. 0:10:01.640 --> 0:10:02.920 You can go out and go for a walk. 0:10:03.000 --> 0:10:05.320 Well, what do you do right for a walk? Yeah, 0:10:05.320 --> 0:10:07.080 you're going you go and neicate you the real world. 0:10:07.200 --> 0:10:08.080 YEA touch good. 0:10:08.440 --> 0:10:09.040 They don't want 0:10:10.760 --> 0:10:13.840 Bernard guiter Us options reported for Bloomberg News.