Agustin Lebron - Trading, Crypto, and Adverse Selection

Key Takeaways

i tell my kids this all the time like life is not short life is long and what that means is you should think of yourself as having many opportunities to learn things and try things and do things software development is fundamentally an exercise in sociology like in organizing teams and in creating processes and culture and conventions around the building of software i think finance is nine percent of gdp is that too high a price to be paying for liquidity and price discovery [Music] okay today i have the pleasure of speaking with augustine lebron who is the author of the laws of trading a trader's guide to better decision making for everyone um this is one of those books uh you know tyler cowan calls these quake books that completely um shift the models you have of the world um i i really really enjoyed reading this book um so yeah i'll let you describe your background augustine but before that let me just um let me ask this question so um peter thiel says that the straw student reading of zero to one is that you shouldn't start a startup and i think that tell me what you think about this i think the strasian reading of the laws of trading is that you shouldn't trade right because um you probably don't have edge uh because you're not better than a marginal trader and if you think you have edge it's probably because you haven't factored in risks and other costs um so don't trade is that is that what i should take away from this book i think you you pretty much hit the nail on the head like a lot of the times that the people um sort of start thinking about trading seriously they start realizing more and more how how how hard a job it really is to do well and uh and the answer is probably look if you're smart enough and and good enough and hard working enough to to make a go at it and make a living at it in financial markets there's probably an easier way to to make money and you know have a satisfying life most of the time okay yeah so do you want to do you want to talk about uh your background and then what you've been working on in the past and what you're working on now yeah so so my background is engineering that's kind of what i did in university uh i did engineering for about six years professionally i was a chip designer at the time i was playing a lot of online poker back when that was a profitable and arguably legal thing to do and so engineering was getting kind of boring and i wanted to do something else and and so i thought well what's what's halfway between engineering and poker and of course that's qua trading um so january 2008 walked into my boss's office and i said i want to quit uh and and he said oh where are you going and i said i'm going to go into finance and he's like are you sure this is a good time to be doing that um he said yep no i'm dead set on it um and a few months later uh managed to get a job at jane street and and wrote out the implosion of western civilization from the seat of a trading desk um so we did that for a few years and then um left jane street a few years ago and started my own consulting company um basically just helping com tech companies with growth things like management and hiring and that sort of thing um and in the last few months started a new company uh in the crypto space how much are you willing to uh give up your edge by telling us what this is or if you're not willing to talk about it that's okay as well yeah no i mean big picture we're building a crypto protocol that is uh kind of new and has some pretty cool cryptographic guarantees uh against things that people don't like uh when they trade in crypto yeah so let's get into some of the topics in the book um so uh yeah first i want to talk about adverse selection because this was you know this was the most interesting part of the book for me um so let me ask this question um if we think of hiring workers as you know placing bids on them if you're like an employer and then multiple employers can place bills on them doesn't winner's curse imply that the average worker is probably overpaid because the true value of the employee employee is not the highest bid but the average bid that they would get paid on the market yeah you're right from the employer side it's definitely adverse election all around like first of all if you're looking for if you're just sort of posting a job at the applicants that apply are you know selected against in the sense that you're selected against that pool because you know people who are really really good probably their employers know they're really good and so they're really incentivized to keep them and so the people who are kind of on the market are probably at the margin not as good um not only that but even just the mechanics of hiring the person who has the final say in terms of whether this happens or not is the employee and so you're going to get adversely adversely selected there because the people who are really really good are going to have lots of job offers and so they're going to pick from one of many offers the people who aren't so good are going to pick from few offers and so employers just systematically get adverse elected that way now whether that means that they're sort of systematically overpaid i think that's a different question because in the end companies have a pretty good idea or at least should have a pretty good idea of what the marginal value of an additional employee is um it's true certainly that people by and large give up things in order for in order to get the security of working at a company so maybe that counteracts that that sort of adverse election in terms of pay um it's not clear which way it washes out i think to me yeah bern hobart i recently wrote a blog post about your um this chapter in your book about adverse election and um so one of the things he said in a footnote almost in passing was that there should be more adverse selection in industries like finance where um the motivation for people to work on them is money because in industry like if a worker wants to work for spacex there's a story you can tell about like why they're working for you and nobody else in finance you know there's a lot of people obviously as you would as you know who are like might be bidding for really talented people so if they're working for you there's something suspicious about that no i think there's something to that um certainly you know doing a lot of the hiring that i used to do one of the biggest uh almost red flags is when somebody comes to you and says oh i've been wanting to be a trader my whole life because they're not like first of all they don't know what trading is right they haven't known what trading is their whole life they don't know what the job really involves it's not tangible in the way that being a doctor is tangible and so what they're really telling you is i've been wanting to make a lot of money my whole life which is generally a pretty well say like in some jobs it's a good motivation but it's not necessarily the motivation you're 100 looking for out of the gate in hiring someone oh interesting um because your chapter on motivation it seemed like you were implying that that is the motivation you should be looking for because if their motivation is emotional then they're going to be losing to people whose motivation is to make money so um yeah i'd love for you to talk more about what is motivation you are looking for yeah so i mean i think so that the motivation of like winning the game of like making money and and that is sort of how we determine who wins the game i think that that part of the the making money motivation makes and makes a lot of sense for a trader but the like all i want to do is make the most money possible is correlated to things that um that aren't maybe so great like because a lot of the job is um is sort of having an inherent curiosity about random things for example um and um like if you're if your whole motivation is like where can i sort of make the most money today it's not necessarily optimal over the long haul and so you kind of need to sort of balance that against these other things like enjoying the game for its own sake uh enjoying the game for for like you know sort of as an exploratory kind of thing um so maybe that's like maybe a little bit inconsistent with something that i wrote in the book but um but i think at the margin people need to hear the other thing more uh yeah okay interesting so and then how do you figure out if somebody enjoys the game for its own sake i think you said in another interview that um uh it's a company like jane street would hold it against you if you have like retail trading experience because um i guess you can talk more about why that is but yeah so if that's not what you're that's not how you judge whether they would intrinsically enjoy the job how is it that you would judge that so i one of the things i've always had maybe you've heard me say before is um i would love to talk to the person who is the third best player in the world at some weird obscure chess variant because that is probably very correlated with things that i care about such as um a willingness a willingness to really like grind and try to get really really good at something and to do so not because there's a huge pot of gold at the end of the rainbow but because you just find inherent enjoyment in getting really really good at something so i think that's that's pretty good um but yeah just general um again aside from sort of the mathematical and and um and sort of risk-taking parts which are sort of maybe independent from this uh certainly a strong desire to be in a competitive environment and to enjoy being in that environment i think that's you know that can take many forms but i think that's a big part of it for sure so then why is um having domain expertise and trading not important um is it because usually in other industries just like if you the more experience you have in the industry the better and it seems like you guys are often hiring people who are just very analytically smart but um maybe you haven't been traders before um so like how do you guys manage to do that right i guess the the thing i'm thinking of is that the concept of a domain is probably a lot narrower than people understand it to be um like if i'm there sitting there on my robinhood account punting stocks back and forth like that is not the same domain as what a trader at a market maker uh or at a top trading firm would do um and in fact to the extent that you think that that's the same domain that is a thing that you have to unlearn when you come work at at you know we'll say a real company and you know that that can happen but it's just it's kind of a problem like it's just a thing you have you have in the back your mind right like you'd rather take a blank slate a really smart motivated blank slate and sort of teach them what they need to know then undo something and then teach them the thing they need to know you see this a lot of the time the other thing is from again at a meta level probably in expectation the person who's doing trading in their personal account isn't doing positive edge trades like they're probably on average losing money and so you would like the person to realize that maybe this is not a winning game for them and so they shouldn't be playing it and so again there's sort of this adverse selection of well if they can't realize they're playing a losing game here then that's probably not great so yeah you said in the book it takes like six six to 18 months before you can train a trader to be net positive um what is happening in that time like what what are the skills you're teaching them yeah so this varies from company to company and even has varied over the course of the history of jane street certainly like when i started it was very much the the socratic method right you sit next to a senior trader and their jobs to teach you everything they know um and so it's just a continuous stream of questions answers conversations etc um j street to their credit has improved on that um there's now sort of a boot camp that you go through where you basically just intensively learn the fundamentals of everything that you that you know the firm needs feels like you need to know as a trader so that again accelerates the process but it is very much sort of putting people in situations to sort of experience the decision-making process and iterating on that decision-making process like what are you thinking about here what do you think about that hey did you think about that what would you do in this situation why why not et cetera and that just that just takes time i wonder so as you mentioned you've done a lot of um you've helped done a lot of hiring for tech companies i wonder if uh how applicable this model is to the tech industry so i mean um could a company like google just have a very effective bootcamp where they get like people who study like physics or math at mit um and maybe not necessarily computer science but you know if you don't know that much programming you can still come in and then we'll make you you know 10x in a very short amount of time or is that something special about finance and trading i don't think so in fact i think that the most common failure mode i see in tech company hiring is hiring for skills instead of hiring for abilities and potential um and it's just because skills are very legible like it is fairly straightforward to spend an hour with somebody and understand whether they can write code in python right and so it's like the drunk looking for the keys near the lamp post like you just evaluate what's easy to evaluate my dream in some sense and this is something that i can't really work on right now but who knows someday i could is the idea of doing mass mass screening for people around the world like what i'd love to find is the smartest um 0.1 of high school high school graduates around the world india nigeria all these countries that are being massively underserved by their educational system and their opportunities um and putting them in these sort of boot camp situations um for you know six months or something where they learn you know useful skills and at the end of it there's like a six figure job with a western company like there's no reason that that companies like infosys should should be taking the lion's share of that arbitrage opportunity like there's this incredible need in the world for people that are you know smart and motivated and there's this incredible supply that we're just systematically under tapping so my answer to your question is yes there's i i strongly believe there is a there's a trillion dollar business potentially uh or maybe it's a nonprofit i don't know in in closing this arbitrage gap your former colleague sam beckmann freed um he um uh you know obviously the ceo of ftx um and he has you know started a big um charity called the future fund and one of their project ideas is exactly what you're talking about where you would uh there would be like large gains if you could enable talent from the developing world so what is it that you would look for when you're like scouting out this talent yeah so i think one of the things that maybe isn't uh isn't terribly polite to talk about but i think is critical is just uh g intelligence like it strongly predicts outcomes across uh jobs across um industries um and so you that is some element of it that is certainly some element of it but also i would say um i think in an ideal world you would build this this process the selection process kind of like a game like maybe like a mobile game or something where you're sort of people are sort of incentivized to kind of keep trying it stuff and maybe it's it's a little bit of a grind and and again you're sort of selecting for that hard-workingness stick-to-itiveness whatever you want to call it to use a principle skinner term um and so yeah like some combination of those two things i think are pretty are almost definitely predictive of of actual value have you have you heard of pioneer uh the thing started by daniel gross yes i've heard of it i don't know much too much about it uh yeah this sounds a lot like it i i don't know too much about it either but yeah this is this sounds very similar i think they're trying to like make um building a startup like a video game so um with you know the associated risk uh rewards and stuff um how do you deal with adverse selection in cases where theoretically adverse selection should work for you um um but you know like the counterparty prices in the possibility of getting a lemon so like an example would be i'm 21 years old and i'm a male so like car insurance premiums for me are huge even if i'm um if even if i'm a good driver because you know there's like um there's the average selection the insurance company faces and like going back to another example we were talking about if there's like a great employee um who's he might be getting underpaid because the company that's hiring him doesn't know how good of an employee he is before he is hired um so how do you how to deal with such scenarios when you're on the other side of the adverse election yeah certainly i think in the car insurance situation i am fairly sure there are now car insurances that essentially put like uh like a accelerometer and a gps on your car and they essentially monitor how safely you drive or whatever how jerkily you drive probably and uh i imagine that that you can sort of decrease your adverse election by by taking advantage of those kinds of things um in the case of the the employment thing um that's a tougher one um at some level the most important thing you can do is select your co-workers as as a potential employee and so getting really really good at evaluating your interviewers i think is i think it's an undervalued skill not so much because you want to tell like are they good or not but it's more like are they a good fit for me is this company good fit for me and the best signal of whether the company is a good fit for you is who the people are that are interviewing you and what do they ask you to do if a company is at all sensible what they ask you to do in the interview is highly correlated to what you do in the job and so that's kind of maybe like a baseline don't don't adverse select yourself by by just kind of being like meh yeah i think this will probably work out or or perhaps more importantly this is a high status company i am told that it is a high status company and that letting that override your personal understanding of what the experience was i think that happens very very frequently um so once you get past that then you're probably in good shape already and at that point i think it just comes down to you know putting yourself in the right positions and i think that's um that's maybe a a skill that's that you learn over time hopefully yeah so i um there's a common thing that my friends complain about to our programmers which is that when they're interviewing they get asked questions that are very unlike their um actual job so you know questions that are almost brain teasers um right but there's a kind of a chest fence argument you can make that it's like if all the tech companies are doing it there must be some important reasons why they are so um have you figured out the reason why such brain teasers are so common is it just that g is so important that this is the best way to measure exactly so this is the thing right the the dirty secret of of all of this stuff is that explicitly testing for iq is illegal in the united states as a as a as an employment practice um however you can kind of drive a truck through it because companies do like for example wonderlic is a company per maybe people have heard of wunderlich because it's the test they give quarterbacks in the nfl um wonderlic is is a company that is dedicated for example to to building employment testing that is essentially iq testing but has the the you know whether it's a fig leaf or actually legitimate justification that as long as you can show that it is uh important for job performance then you can kind of do the testing right and so essentially i feel like a lot of these brain teaser type questions are as you say you know iq tests disguised i think oftentimes they are badly misapplied by the interviewers like i think it takes actually a lot of really really hard training and experience to ask these sorts of questions in a way that gets you the signal you want um but i think that's a that's a big part of it like the the extent to which you view your job as vocational um is is the extent to which you're going to hate those brain teasers right like so if i'm a programmer and i want my job to be i'm just going to write code all day and sit down and just write code then you're not going to like those brain teasers because you don't think of them as part of your job whereas if you think of your job as a programmer as somewhat more expansive in the sense of like well i'm here to really think about hard problems and i happen to implement them in code then maybe you're going to think of the brain teasers as more correlated to the thing you want to be doing so again select for what you like yeah and maybe it makes sense to select for the latter type of person as well right or i don't know which is preferable to hire but um well so so i think this is the thing about about companies again there's a lot of schizophrenia in tech hiring um one of the things that's clear is everybody says they want to hire a players um but only a small fraction kind of by definition can hire those those sort of high percentage or high percentile kinds of people and so what ends up happening is a lot of startups have the failure mode where they try to build these incredibly selective processes um but the people who who they really really want are never going to accept their offers they're going to go somewhere sort of more high status or more high paying in particular and so you try to select for like an 80th percentile person but you end up selecting like a like a set of 50th percentile person people who look like 80th percentile people which is really really bad and so what you should actually do as a startup is be very clear-eyed and say look if i have a team of 10 i probably need one or two like 90th percentile people and i should evaluate for and in particular pay for that uh and then the rest i should try to hire a kind of 40th percentile people and you know put them in situations where they can be effective that's a much much more cost effective way and more stable way to build a company but nobody wants to hear that and nobody wants to build a company like that that's a great example like a variable strategy um so i'm wondering do you have any ideas of what good arbitrage opportunities in tech hiring might be i know um i think spacex some of their early engineers were from the gaming industry because they're very used to doing optimization problems there but um it's not that's traditionally a high status uh career so uh there's like arbitrage there are you do you have an eds now of like what is a good place you would be looking for really talented potential future programmers if you were if you couldn't compete with pay uh at google or something yeah so i think one of the things i always tell companies is um go more junior like if you look at if you look at the salary somebody just comes out of school and i'm not talking about somebody who just came out of stanford i'm talking about somebody who just came out of like a reasonable cs program right and you look at their salary three years later like it could be almost double sometimes right it's just a crazy crazy jump and that is kind of unjustified i mean you can sort of see the argument for it but it's just be like there's definitely a kink at the two to three year point because every startup or ever i mean every tech company seems to want to have two years of experience and a lot of it is because companies just don't want to or can't see themselves investing in the training of those first two years and if they do they tell themselves well they're just going to leave after two years to go for a higher paying job somewhere else but i think those are terrible answers by and large to the problem like you should be investing in training your people you also get the benefit of training them exactly the way you want and if you put in that work uh and you think carefully about what it is that people are coming to work to do for you day to day probably they're not going to leave right like if if you give them a reason to not leave they're probably not going to leave switching jobs is incredibly costly and risky people don't go out of their way to do so so like you're kind of you're kind of getting the the um the inertia working in your favor anyway so like let's work on these things sounds very similar to the sheepskin effect of the last semester of college um so the it uh brian kaplan has a really good argument about this in the case of education which is that the last semester of college like boosts your earnings many times more than the percentage of college you spend in that last semester and it can't be because you're like learning that much more in the last semester um which i guess sets up an arbitrage opportunity for hiring people in like right before they're about to finish our last year or something but you see like give me like i'll give you a perfect example here in san diego where where startups in san diego tech companies in san diego love to hire intuit employees that have two to three years of experience because intuit hires a bunch of people and they train them and they train them pretty well and and then like they get poached but of course like nobody really actually thinks about the idea that like intuit knows who the good and the bad are after two years and like you're not seeing the really really good ones into it's keeping those right so uh so you see in the book that you've traded over your long career in trading you've traded all kinds of different financial instruments i wonder um what is the reason so is this just um i guess you you just have to do the you had to trade whatever um market that you have to at the moment or because i would think you say in the chapter on edge that one of the ways you can actually get edge is to specialize um so is it a mistake of firms to let their traders over their career trade in multiple different categories or is that necessary in order in order to build your general aptitude as a trader yeah so i think it's a balance um certainly i don't think that again it depends on how big the reference class is certainly i have never done any trading that looks like look at a balance sheet in an income statement and listen to an earnings call and make a bet on that like that's sort of fundamental trading i have never done any of that um and i think it would be a pretty big mistake to put me in that situation um but within we'll say that the the well-defined realm of like quantitative trading um i think a lot of the same skill sets apply in different markets like you're kind of build bringing the same skill set to different markets and having that experience of going around and looking at different kinds of markets and how they work informs like it sort of informs how you think about things and and gives you that that wider vision that i i think makes you a better trader um so yeah i think it's a balance so i i think finance is nine percent of gdp so i understand the argument that you know finance helps allocate scarce uh resources um to where they're needed most but um if we're giving up like a tenth of our resources to make the allocation of the rest of the resources more efficient is that too high a price to be paying for liquidity and price discovery um so is finance too high a fraction of gdp um i go back and forth on this question um i really do um because kind of when you see it from the inside a lot of it is zero zero-sum competition um and and it feels like come on there's got to be a more efficient way to do this um but at the same time kind of outside view we haven't come up with a more efficient way to do this and it's hard to argue with gdp growth and so i kind of go back and forth on it certainly i think the other thing about it is um there's two countervailing forces you can you can sort of be inside something and be really really familiar with it and just your act the act of being very very familiar with something just gives it legitimacy kind of automatically um but at the same time like if you look at something from afar you're like oh that's ridiculous right like that's that's not that's not a thing that should exist right and so it's sort of this perverse thing where the people most like the most well-informed people the people who really could or should be making these decisions about like is this a legitimate thing that we should be doing our bias towards thinking like um yeah you know what this is probably a good thing to be doing or there's value to this and so it's it's hard to sort of disentangle the the like the experience and and the biases that that experience sort of gives you and then would that would that fraction shrink without without harming efficiency if um like are there inefficiencies created by guard regulation or by restrictions on capital flow um or is that like basically what you should expect it to be even in a free market or an in an optimally regulated market let's say that's also a tough one um and it's and it's not that i haven't uh thought a lot about these it's just i feel like i don't have um i don't have a great answer like at the margin what would i like if you sort of made me like regular regulator of the world like at the margin what would i do um there are some things that i would regulate more um and this is probably going to be a very unpopular opinion among my my financial friends but like i think leveraged etfs should be banned for from retail trading like i think this they're just kind of a bad instrument uh in particular like all the volatility products um so i feel like that should probably be regulated some more um but at the same time this sort of qualified investor status thing that people are driving a truck through like that seems weird like should should there be should we just eliminate the qualified investor uh status and let people invest in whatever they want or should we make it even more restrictive um i'm not sure about that one and certainly the other thing about it is like a lot of the regulations especially around capital requirements for banks are incredibly baroque and they feel like job ponzis a lot of the time like we need to figure out a way to employ all these people and like okay we're just going to create like basel three and that's gonna be like an extra thousand employees for every large bank in the world um that's probably kind of a deadweight loss but but doing things more simply doesn't seem like it's gonna get you the thing like the sort of the stability outcomes you want and so yeah it's just i feel like it's just kind of poor trade-offs all around what is the long-run future of trading firms look like so if um if economic growth continues to stay low then you would expect like other financial instruments to stop growing at high rates as well but even if economic rates economic growth speeds up um if markets get more efficient over time then again you would expect the profits that any one trading firm can get to decrease so is there a future for highly profitable uh trade firms like jane street like in the far future so i think to the extent that jane street and companies like it provide a service to the world and i really do think they provide a service to the world then they're going to be around and they're going to be profitable now are they going to gain um like we'll call them excess returns um even that's not so obvious because the thing about trading firms is especially market makers and that sort of thing like most of the time the business is pretty good if you're really good at it um but sometimes it's really good like when when there's lots of market volatility and that sort of thing um but that's precisely because you are the person you are the entity that is willing to take the risks that nobody else is willing to take and to the extent that we're going to still continue to have volatility in terms of either like market volatility or you know economic downturns or whatever there's always going to be um a service that these that these companies are going to provide now over the long run i feel like probably there's going to be more consolidation it seems unlikely to to to stop um just because you sort of gain the the benefits of the economies of scale just kind of keep going up um but then again you have sort of new things that come up like crypto and that sort of thing where like it's the wild west right now and there's going to be like a big consolidation over the next 10 years i think that's the natural arc of things oh interesting so um yeah can you describe what these economies of scale look like in finance and um um and then what is a trade-off where if you're like too big then it's not even worth your time to like look at smaller uh smaller investments where you can't take as big a state without moving the market yeah so the thing about finance or like market-making trading in general is um it's very labor-intensive right so you should think of it almost like the value of a seat or the value of a person's time and so are there going to be are there going to be inefficiencies in the market like pockets in the you know pink sheets or something where it's just not worth a large companies or a large successful company's trader time to look at yes like those will always exist and they'll get slowly competed away by by the by the mom and pop trading operations or or even just the like the former jane street traders who are now at home and kind of doing it on their own for fun um so i think those will always kind of be there is there a potential that markets can get like way way more efficient if we have we don't have much stronger ai and um and at what point will um the work that even traders do that's like much more um i don't know much more model generation and like thinking abstractly at what point can that even get automated away and not just like the road calculations yeah i would say it's already getting and gotten comp like more efficient like when when my former boss started the idea of an options market maker having 10 stocks that they were market makers in was like that was kind of the limit right when i was doing it like we could handle like 100 stocks right market making and 100 stocks again technologist ma technology just made everything more efficient or more efficient in human time um that will continue like you can you can sort of set up things where i'm looking at some data and i can like run a bunch of different models and just select the good ones and make sure that i'm not over fitting because i've all i have all these overfitting protections this is all stuff that you can do now that maybe you couldn't do 20 years ago that will definitely happen i think when people talk about ai and trading i think it's um it's very hard to it like we have to define terms i think that's the hard part is defining terms when we talk about ai because if we talk about if like if you ask um a reasonably aware person what ai means not probably today in 2022 90 of people are going to say oh we're talking about large language models of course that's what ai is right and so is the question like is gpt is gptn going to be a significant force in in markets like i'm honestly kind of skeptical about that i don't know that that the let's just keep making larger transformers is the way that we're gonna get to ai but that's my personal parochial opinion but if we think of ai more broadly as um as slowly but surely uh increasing the range of things that things that machines can do that humans can do like the the more we sort of creep into the things that humans can do that machines can do as well then then yeah then like the the human part is going to slowly start to get disappeared away um i think the the the natural analogy is what happened in the 20th century with manufacturing where like it used to be kind of all human power and a little bit of machine power where you had kind of this like big central like why did factories in the 19th century and early 20th century why were they kind of tall and thin well it's because they had one steam plant and they had to like all these belts and stuff to like use the the power from that one steam plant right and then like electric motors happen it's like okay now factories are horizontal right but over time the trend is for it to be sort of less human power and more machine power and i think the the analogy is perfect i think ai over time is going to take more and more of that sort of cognitive load from the human um that seems inevitable to me i'm curious why you're skeptical um uh that uh like a scaled-up uh gpt3 or other language uh large language model um uh i'm curious so why does it not have applicability um in financial markets like i don't know there's like a toy version where you have like gpt 10 and you ask to complete the sentence the best trade i can make today is and then um so why is that unlikely to happen so there's a couple things that i might say one is is the concept of sample efficiency like these things are incredibly simple and efficient in a way that the way the humans learn are not and so there's something fundamental there that that we're not getting right and the thing that i think we're not getting is um is the things that our brains have which are structures for uh semantic understanding like to the extent that these large language models have semantic understanding it's kind of by accident right it's just like it's the clever hans thing right it's just like a super clever haunts and it's super impressive and i'm not criticizing the models like they're incredibly impressive but it's still a clever hans thing um and so there surely must be a better architecture out there much like our brains have these sort of architectures that um that sort of specialize in certain things that that give these these machines like semantic understanding or at least give them the potential to have semantic understanding um that i don't think gpt3 certainly has has evidenced uh so jc street seems like a mysterious place but what's interesting to me is there seems to be a large overlap with the rationality and ea community so obviously you have saint greed he's um you know he went into jane street with the explicit goal of earning to give yep tyler cowan announced that 20 million dollars have been donated to um his emerging adventures grant program from jane street traders and you know even reading your book like you reference so many thinkers that are prominent in um like rationally spears and you um so there seems to be a big overlap with this community and with at least a part of the trading world that i'm familiar with no that could just be selection selection effects but what is going on here yeah it's a great question i think uh maybe at two levels one is the idea of being very rational and not fooling yourself and uh and to use a yudkowski term just shut up and multiply like i think that that is a that is a thing that is very common i think in the two circles or at least probably it should be um like try to really understand the real world and it matters to do so and doing so using kind of rational mathematical logical approaches i think there's a lot of overlap just inherently there but i think you could say that about any number of finance wall street whatever trading firms i think the one thing that jane street has going for it differentially from those other firms maybe is uh the a culture of collegiality i think that's kind of an important thing that that jane street has developed over the years and continues i think to have um and so i think that's there's a lot of overlap there like it's the kind of place that if you are an ea person thinks about things rationally and just enjoys the enjoys the process of kind of this collegiality and and and working with people and thinking interesting thoughts together jane street's going to be a very natural fit for you and i think maybe that's some of it too when i had bern hobart on the podcast um we talked about whether debugging or finance was a better application of like rationality principles because in each case you had like updates or beliefs and so on and one interesting point he brought up was um in finance you have you not only have to model like a static system um as you would in debugging but you also have to model other asians and their incentives and their motivations which makes it a much more like a dynamic system to get a hold of in your brain which i guess it could even mean that like the tools are like the current rationality movement are not good enough to you know be able to think about those things as well as probably you guys have natively developed in the industry yeah and look i the the cross pollination goes both ways um but yeah the idea of of you being an agent in the world you're trying to study is fundamental in trading um and it makes it like so much more interesting i think that's one of the getting back to the ai thing just because it occurs to me is one of the the big failure modes is to think to think that okay well yeah i'm just gonna like throw some ai and or machine learning or something at this data set and i'm gonna get a trading strategy and okay that's great like let's say you you've figured out something that predicts the price movement 55 of the time like that thing can still actually lose a lot of money in production because of the again so there's the adverse selection effect of you're only going to do a small fraction of the good trades you're gonna do all the bad trades you want um but also if you are actually making money at it this is like a big shining signal to the rest of the world like hey there's money over here like why don't you compete it away um and so yeah that's definitely a huge component of it so you have a very interesting chapter on software and technology in the book and one of the things you argue for is that we should take the concept of technical debt seriously in a financial sense um so is one implication of this interpretation that you should be willing to accept uh technical debt more if you're a rapidly growing company because you know if like you're a startup that's growing fast it makes sense to maybe take out a lot of loans because you can pay back the interest plus way more um but maybe maybe if you don't take it financially maybe that's you would think that if you're like scaling rapidly that's the worst time to take on all the technical debt because you're just going to be hampered the entire way along so yeah so more generally the question is what kinds of firms should be more willing to take on technical debt yeah certainly startups is is the classic example and it and it's and it's non-recourse debt right like if it goes belly up like you have to pay it back right you're done um so so yeah like startup should definitely do this and and you see it all the time right this concept of an mvp where you know let's just get something out there let's get some feedback from the users with the understanding that hopefully with the understanding that you're going to have to essentially rewrite it from scratch if it's successful i think it's a very useful and very very um uh productive way to do software startups um because yeah like the the implied interest rate that you're willing to pay is incredibly high um larger companies it's interesting like if you ask yourself this is kind of a conversation i had with with uh with one of my good friends um who i actually did consulting with he worked at qualcomm for a lot of years and and i asked him because he worked very closely with microsoft like microsoft employees tens of thousands of software engineers like what do they do all day and he said to me like look i don't actually know for a fact but i'm pretty sure the vast majority of them are like well this library is deprecated we need to upgrade this thing let's change like all this like code and all these different little places right so like there's just sort of uh uh like a like a like a sort of an archaeology of software that occurs where where you know if you build if you've been building a software piece software for like 20 some odd years like there's just all this cruft in there that you're just continually trying to maintain so that it's functional as you go from you know this os to this otherwise to the cloud to whatever right um so i think that's that's kind of uh like an accumulated debt that the large companies certainly have yeah that's so interesting they're just like servicing the debt they accumulated in like the 80s and 90s when they're growing rapidly and you can even think of like them moving to a new platform or like rewriting their code is like refinancing their debt or something right exactly in fact like i would say um probably the best me probably the best book i have ever read about software development is actually uh science fiction um vern revenge a deepness in the sky i feel like is uh very crucially about like it sort of takes this idea like what if we've been building on the same software stack for six thousand years what does that look like like what does that world look like um i think it teaches us a lot about how to think about large software projects large long-term software projects yeah so i'm super interested in how uh you guys think about software in the financial industry um i know jane street uses oh camel um so because i mean uh there's like safety you can tell me more why this is but from what i understand it's like there's um there's more safety in a functional uh functional programming language um yeah so how do you think about like obviously there's much more reason to want to have like safe code because you're dealing with an adversary there in some sense so yeah i'm curious like how do you guys make engineering decisions and what are like the trade-offs involved when you're doing when you're working in finance yeah so as you said like james street uses o'camo i think one of the one of the biggest advantages of using that language is it it is strongly and statically typed and so you can put a lot of things um in the like you can use the type system to make uh impossible states unrepresentable this is like a really good software engineering thing you should do and it makes it sort of very easy and um in a rich environment to do that in and so this like oh i didn't know i had to handle this explode problem is kind of minimized um but yeah like you know jane street and companies like it obviously optimize for avoiding hot loops and code that incinerate money really really fast and that is not what your average whatever startup optimizes for uh or it shouldn't be anyway um but i but the thing i keep coming back to in talking to you know technology leaders and that sort of thing is software development is fundamentally an exercise in sociology like in organizing teams and in creating uh processes and culture and conventions uh around the building of software like you know software development is fundamentally the management of complexity like the science of managing complexity because it is incredibly complex right um and so all that sociological stuff ends up being some of the most important stuff to think about now that you're working in finance but you have a startup so you have to think very carefully about this trade-off um how like how are you managing this given that you have to like i guess move fast but you also need to be safe uh hire really really good people honestly like don't skimp on those first few employees is is uh i think a really important thing like where where the bar is kind of uh like the bar is kind of weird like it's not it's not like there's sort of one total ordering over a quality of engineer right there's like they're incredibly multivariate um but certainly um one really really good thoughtful engineer who can build correct code is worth for not so thoughtful people in a spot like that and so that's kind of the thing we're optimizing for right now and um such engineers do you expect or give them a lot of um knowledge about uh finance or can they just function knowing about engineering just just about engineering and then you can just like tell them we need a program that does this or do they need to have an understanding of how trading and finance works so need is probably a hair strong but certainly the culture that i want to build is one where it's almost need like it's almost like want right like i would want to hire somebody i want to hire somebody for whom understanding the problem domain deeply is a critical part of the job they feel they're doing and so is it possible to build something like this um another way probably but that's not the company i want to build um and so in your career you've done so many different things engineering trading consulting yeah so how much carryover and lessons do you feel like you've had between these different domains or do you feel like they're um did they have like self-contained pools of knowledge so i think if there's one uh constant for me it's i'm surprised by how much my previous careers for my next careers like when i when i wanted to to