Dwarkesh PodcastBrett Harrison — FTX US former president speaks out
EVERY SPOKEN WORD
155 min read · 30,628 words- 0:00 – 8:30
Passive investing & HFT hacks
- BHBrett Harrison
You are probably gonna be fired for this letter that you wrote. Sam is gonna destroy your professional reputation. Like, where do you think you're gonna be able to work after FDX? It, it was, he was threatening me. (screen whooshes) When I knew Sam, when he was 21, 22 years old, he was, like, a happy, healthy-looking kid. When I got to FDX, I saw someone who was very different than that person I remembered. And it felt like he was spending virtually no time helping the company move forward. It was so much about image and brand and PR. Media was primed for the archetype that was Sam. Doesn't matter how little time he spent with the company, doesn't matter how he treated employees internally. Architect, it makes it really easy to access kind of all corners of the digital asset ecosystem. (screen whooshes)
- DPDwarkesh Patel
Okay, today I have the pleasure of speaking with Brett Harrison, who is now the founder of Architect, which provides traders with infrastructure for accessing digital markets. Before that, he was the president of FTX US, and before that he was the head of ETF technology at Citadel, and he has a large amount of experience in leadership positions in finance and tech, so this is gonna be a very interesting conversation. Thanks for coming on to Lunar Society, Brett.
- BHBrett Harrison
Yeah, thanks for, uh, coming out to Chicago.
- DPDwarkesh Patel
Yeah, (laughs) my pleasure, my pleasure. Is the growth of ETFs a good thing for the health of markets? There's one view that as there's more passive investing, you're kind of diluting the power of smart money, and in fact, what these active investors are doing with their fees is subsidizing the price discovery that makes markets efficient, and with passive investing, you're sort of free-riding off of that. Um, you were head of ETF technology at Citadel, so you're the perfect person to ask this. Is it bad that there's so much passive investing?
- BHBrett Harrison
I think on net it's good. I think that most investors in the market shouldn't be trying to pick individual stock names, and the best thing people can do is invest in sort of diversified instruments. And it is far, far, far less expensive to invest in, like, indices now than it ever was in, in history because of the advent of ETFs.
- DPDwarkesh Patel
Yeah. And so maybe it's good for individual investors to put their money in passives, uh, uh, investments, but what about, like, the health of the market as a whole? Is it hampered by how much money goes into passive investments?
- BHBrett Harrison
Uh, it's hard to, it's hard to be able to tell what it would look like if there was less money in passive investment now.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
I, I do think one of the potential downsides is ending up creating extra-correlated activity between instruments purely by virtue of them being included in index products.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
Uh, so, you know, if, like, when Tesla gets added to the S&P 500, like, Tesla doesn't, like, suddenly become a different company-
- DPDwarkesh Patel
Right.
- BHBrett Harrison
... whose, you know, market value is, like, fundamentally changing, but yet it's going to start moving very differently, you know, in terms of its beta correlation between other instruments in the S&P 500 purely as a function of all the passive investing that moves these instruments in the same direction. So, that's the sense in which I think it could be detrimental.
- DPDwarkesh Patel
But naively, you would assume that, like, um, efficient market hypothesis would say that if people know that Tesla stock price would irrationally climb when it's included in the S&P 500, that people would short it and then it, there should be no impact from this irrelevant information. Why isn't that the case? Like, why...
- BHBrett Harrison
Uh, it, it probably mostly is.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
Um, I think that sometimes there can be liquidity differences that cause at least temporary dislocations in stocks, like, I mean, the simplest example is, like, you have an ADR, like, an American depository receipt, that sort of ex- fungible for some underlying foreign stock, and these two things should be, like, almost the same value at all times, like, net of currency conversion and conversion ratios, but if one of the markets is highly illiquid or difficult to access then there's going to be dislocations in price, and that's, like, the job of, like, the Jane Streets of the world to kind of arbitrage away the price over time, and so long run, you wouldn't expect these things to be dislocated for that long. So I'm sure there are people who are understanding, like, the fundamentals of individual names in the S&P 500 and when there's, like, American news and the entire S&P falls, they are, you know, maybe buying S&P and selling individual names and es- expecting that relative value spread to come in over time.
- DPDwarkesh Patel
Speaking of, by the way, these firms, well, how... You don't have to tell me specifics, but how similar are the strategies for market-making or trading at Jane Street versus Citadel and these firms? Is it the same sorts of strategies or are they pretty different?
- BHBrett Harrison
Uh, I think a lot m- more differences than people appreciate from the outside. Um, different companies have, they, they establish different niches in areas, like, Jane Street established its early niche in ETFs, um, at kind of like a mid-frequency level. Um, so not, like, ultrafast but not, like, long-term, year-long discretionary macro, whereas maybe your, you know, Citadel Securities, uh, kind of firm got, you know, built their niche more on, like, lower-latency, you know, options market-making. So, like, it could be, like, all over the place. There are some where they are trying to optimize for really short-term, like, microstructure alphas, like, trying to predict where the order book is gonna move over the course of anywhere from milliseconds to seconds. There are firms that care more about the, like, relative conversions of instruments over the course of hours to days. There's, you know, sophisticated quantitative trading firms that are doing longer term, you know, days to weeks to months-long trades too. So a lot of the infrastructure can be similar, like, either way you need to be able to connect to exchanges-
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
... download market data, establish simulation platforms, build, like, tools for traders to be able to grasp what's going on in the market and especially be able to visualize, like, their own proprietary models and alphas, uh, but beyond that, the actual strategies and the ways they make money can be very different.
- DPDwarkesh Patel
Famously, in other kinds of development, there's these, like, very famous hacks and algorithms, right? So in gaming and graphics, John Carmack has the famous, um, fast inner square root for doing graphs calculations, uh, normalizing vectors faster. Um, you were not only a, a developer in finance, I know what the exact term is for that, um, but you led, you know, teams of hundreds of people who are doing that kind of development. Are there famous examples like this in finance? The equivalent of fast inner square root, but for the kinds of calculations you guys do?
- BHBrett Harrison
Yeah. They're, like, all over the place.
- DPDwarkesh Patel
Uh-huh.
- BHBrett Harrison
Like, there's tons of, you know, hacks and tricks and things like that. I think, like, for example, here's a, a famous one. Not famous, I think I read it in a paper and, like, a bunch of other developers from different other companies told me about this. It's not something I saw at places that I worked but if you're sending a message to, let's say, NASDAQ to buy stock and you want to get there as fast as possible, well, what is a message to NASDAQ? It's a, you know, TCP/IP wrapped, you know, message with a particular proprietary protocol that NASDAQ implements. But let's say your goal is, you know, you know you're gonna trade Apple, but you're not sure, like, what price and at what time, and you're kind of waiting for some signal to buy Apple as fast as possible. So, what you can do is you can pre-construct the entire TCP/IP message. Like, first put the TCP header on there, then the IP header, then, like, the kind of outer protocol that NASDAQ specifies, and the inner protocol except for, like, the byte slot where you put in the price, and then preload that message into the network card's-
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
... sending buffer so that once you're ready to send, you can just, like, pop in the price and send it off and incur as little latency as possible.
- DPDwarkesh Patel
That's awesome. (claps)
- BHBrett Harrison
I think the, the analogy of the video games is a good one because, like, just, like, in, you know, video game graphics, what's the end goal? It's not, like, to produce the most theoretically perfect simulation of environmental graphics. It's to have something that, like, looks good enough and is fast enough for the user. And that's also true in, like, HFT and quantitative finance where, like, the goal is to get to, like, the approximately right trade as f- fast as you can. It's not to have, like, the perfect theoretical model of, you know, underlying price dynamics.
- DPDwarkesh Patel
That is so fascinating. Um,
- 8:30 – 18:38
Is Finance Zero-Sum?
- DPDwarkesh Patel
but th- this actually raises an interesting question. You, if you have some sort of algorithm like this that gets you a few nanoseconds faster to the NASDAQ exchange and that's why you have edge, or you've, like, leased microwave towers to get from New Jersey to Chicago faster, or you've bought an expensive server in the same place that, like, NASDAQ, um, is housed, what fundamentally is the advantage to society as a whole g- from us getting that sort of information faster? Is it just sort of a zero sum game of who can get that, incorporate that signal faster? Like, why is it good for soc- society that peop- so much, so many resources and so much brain power is spent on these kinds of hacks and these kinds of optimizations?
- BHBrett Harrison
Yeah. So I think if you start from the premise that having liquid tight efficient markets is important for the world and you say, like, "How do I design a system that, you know, optimizes for that?" I think you want smart sophisticated technologists competing at the margins. And, of course, the more they compete, the smaller the margins become to the point where you think, like, the little extra activity people are doing to get slightly better don't seem to be, you know, greatly affecting the whole system as much as if, you know, as it was in the earlier days when things were slower and tick sizes were wider. Think it's difficult to imagine designing a market where you say like, "Okay, everyone should innovate up until this point and then stop competing and then just stay in stasis." You know, and maybe you can create certain regulatory or market structures to try to prevent that, but I think on average you want people competing at the margins, even if they seem like they are, you know, minuscule. At the same time, I think it's not zero sum for society for, you know, technologists to be creating like super fast, like, ultra low latency various sophisticated algorithms. Like, maybe, I don't know, we have a lot of geopolitical instability in the world. Who knows if, like, the, our microwave network that we built out in the US could have, like, greater use cases than just for quantitative finance? But, like, quantitative finance subsidized the creation of these towers.
- DPDwarkesh Patel
O- Okay. But... So tha- that's sort of like a contingent potential benefit. Like, I guess people tell a similar story about NASA, right? In this case literally microwaves that they subsidize a lot of the science that ended up becoming, becoming into products. Uh, so that's a, that's a interesting account of the benefits of finance that it has the sa- yeah, i- i- wha- whatever tricks they come up with might be useful elsewhere. But that's not a story about how it's directly useful to have n- you know, nanosecond level latency for filing your, like, Apple stock or something like that. Why is that useful? Directly, I mean.
- BHBrett Harrison
If there is some kind of news that happens in one part of the world and that should affect the current price of stock in a different part of the world, I think that if you care about efficient markets, you want the gap between, like, source of truth events and ultimate price discovery to be as small as possible. I think if you believe... If you wanna question whether getting, you know, a few extra, like, milliseconds or microseconds or nanoseconds is worth it, I think you're then putting some kind of, like, value judgment on, like, the what is the optimal time it takes to get from to price discovery and saying, like, you know, "A second is too slow but a millisecond is too fast." Or, "A millisecond is too slow but a microsecond is too fast." And I d- I just don't think we're, like, in a position to do that. I think we kind of always want as close to instantaneous price discovery as possible.
- DPDwarkesh Patel
Um, I, I, I'm only asking more about this because this is really interesting to me. There is some level of resources where we would say that at this point it's not worth it, right? Like let's say $5 trillion a year was spent on getting it down from like two nanoseconds to like one nanosecond. I know that, that, that's probably not a realistic number, but-
- BHBrett Harrison
Sure.
- DPDwarkesh Patel
... just as soon, like there's some margin at which, uh, for some weird reason, the incentives of society are to spend so many resources on it. Uh, uh, or w- would you say that we haven't reached that margin yet where it's not socially useful, the amount of, uh, brain power and resources that are spent on getting these tight spreads and-
- BHBrett Harrison
Yeah, I don't know how large a percentage of like GDP prop trading is.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
I suspect it's not that large. Uh, so I don't think we're close to that, you know, theoretical limit of where I would start to feel that it's a waste. But I also think like there's a reason why they're willing to spend the money on this kind of technology, because they're obviously profiting from doing so, and it has to come from somewhere. So somehow like the market is subsidizing the creation of this technology, which means that there's still ability for value capture, which means there's still a service that's being provided in exchange for some kind of profit. I think we wouldn't spend five tr- $5 trillion on a microwave network because there isn't $5 trillion of extra value to be created in doing so.
- DPDwarkesh Patel
Got it. Has being a market maker changed your views about civilizational tail risk? Because you're worried and like, you're worried about personally getting run over, right? By some sort of weird event, an adverse selection. Does that change how you think about societies getting run over by a similar thing? Or is, is it... Are the mental models isolated?
- BHBrett Harrison
So I think working in, uh, you know, high speed finance teaches you to understand how to more correctly estimate the probability of rare events.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
And in that sense, you know, working in finance makes me think more about the likelihood of, you know, civilization ending, you know, problems. But it doesn't suggest to me sort of different solutions. You know, there's a very big difference, uh, being in a financial setting where, you know, your positions are numbers that you can put in a spreadsheet and you can model like what happens if like every single position goes against me, you know, three x the wrong way, and what instruments would I have to buy or sell in order to be able to hedge that portfolio. That's like a closed system that you can actually model and do something about. Having like a trader mentality on, you know, future pandemics I don't think helps you much. I think maybe it slightly changes your ability to kind of estimate the probability of such events, but the actual solutions to these problems are a combination of like, you know, collective action problems plus, you know, being able to sort of model the particular type of, you know, unknown unknown about whatever the event is. Uh, and I think those kinds of, you know, solutions should be left to, to the experts in those particular fields and not up to traders. In other words, I don't think like having the trader mentality around rare events in like normal civilization outside of finance really kind of helps you much. And maybe in some ways has led people to think more hubristically that they can do something about it.
- DPDwarkesh Patel
Gee, who could we talk, be talking about? (laughs)
- BHBrett Harrison
(laughs)
- DPDwarkesh Patel
Um, but that, that's really interesting you would say that. Um, I would have... Like, uh, famously, uh, you know, these market making firms really care about having their employees be well calibrated and good at sort of thinking about risk. Uh, I'm surprised you think that there's transfer between thinking about that in financial context and thinking about that in other contexts is, uh, that low.
- BHBrett Harrison
Yeah, again, I think it helps you at estimating probability of rare events-
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
... but it, it does not translate super well to what action then do you take in the face of knowing those rare events.
- DPDwarkesh Patel
Were your circles or people in finance earlier to recognize the dangers of COVID?
- BHBrett Harrison
That's a good question. Um, I think that people in my circles were quicker to take action in the face of knowing about COVID.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
Um, you know, there are a lot of people who like kind of stuck around in, you know, cities and like kind of their existing like, you know, particular situations, like not knowing kind of where this was going to head long term. And I think if you have the, like, the fortune of having the financial flexibility to be able to do something like this, a lot of like the people in kind of the financial circles kind of immediately recognized, okay, there's this like big risk, this unknown, and I don't want to get like adverse selected against in terms of being able to like get out of the like locus of, you know, bad pandemic activity. And so people immediately were fleeing cities, I think faster than other people.
- DPDwarkesh Patel
That seems to point in the opposite direction of them not being able to, you know, estimate and deal with geopolitical risk.
- BHBrett Harrison
Well, I mean, there, there you have like an actual event that has occurred and then in, in the face of the event, what do you do right now?
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
I think that's different than like what do we do about the potential for, you know, AI to destroy civilization in the next hundreds of years.
- DPDwarkesh Patel
(laughs) Yeah. Yeah.
- BHBrett Harrison
Or, or what do we do about the next potential, you know, biological weapon or the next pandemic that could occur?
- 18:38 – 23:10
Interstellar Markets & Periodic Auctions
- BHBrett Harrison
in recent history.
- DPDwarkesh Patel
Matching engines already have to deal with the fact that you can have orders come in from, like, Illinois, you can have orders coming from j- Japan, and given light speed, they're not gonna arrive at the same time. You still kind of have to work around that. Is there any hope of a single market and matching engine for once humanity goes interplanetary or interstellar? Could we ever have, uh, a market between, like, us and Alpha Centauri or even us and Mars, uh, or is the lag too much for that to be possible?
- BHBrett Harrison
Yeah, so I mean, without making any changes to the matching engine, there is nothing that, you know, says that when an order comes in, it can't be older than, you know, X time, right? What it does mean is that, like, the, the actual sender, if, like, they're sending a market order from halfway across the world, by the time that the order reaches the exchange, they might end up with a very different price than the one they were expecting when they sent it. And therefore, like, there's probably a lot of adverse selection sending a market order from halfway across the world than in a co-location facility. So, you can technologically run an interstellar, you know, exchange. It just might not be good for that person living on, uh, on the moon.
- DPDwarkesh Patel
Is- is there any way to make it more fair?
- BHBrett Harrison
Yeah, so I think there's actually a kind of real-world analog of that, which is, like, automated market makers on slow blockchains.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
Because if you're used to working on NASDAQ where, you know, NASDAQ processes, like, a single message in somewhere between, like, tens and hundreds of nanoseconds per order, a blockchain like Ethereum processes, what, like, you know, 15 to 50 messages per second? So, significantly slower by, like, numbers of orders of magnitude, and yet they've been able to establish, like, pretty mature, like, financial marketplaces by saying that rather than you having to, like, send orders with prices on them and then cancel them when the prices aren't good anymore, like, there will be kind of an automated function that moves the prices at the matching engine. And so whenever your order reaches the exchange, it'll always be kind of a predetermined fair price based on the kind of prevailing liquidity at the time. So, one can imagine, like, building a NASDAQ for, like, interstellar market is kind of similar to building, like, Uniswap now on Ethereum in terms of order magnitude and speed. But there's other things you can do too. Like, you could establish, like, periodic auctions instead of, like, continuous matching and things like that, and that could potentially help mitigate some of these issues.
- DPDwarkesh Patel
Yeah, it's actually something else I wanted to ask you about. What do you- what do- what is your opinion of periodic, you know, frequent batch auction systems? Should- should we have more of that instead of...
- BHBrett Harrison
So in theory they help mitigate the advantages of high frequency trading.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Because if you know there's going to be a, an auction every 30 seconds and it's not going to be by time priority, it's gonna be by price, then it doesn't matter if you send that order at the beginning of the 30-second period or the end of the 30-second period. Like, it's- it's really, like, the price that determines whether you get it filled, not something to do with, like, particular latency to the exchange. I think in practice a couple of exchanges around the world that used to have those have switched away from them.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
Like, I think the, like, Taiwan stock exchange used to have periodic auction system and they just thought the, like, the liquidity and price discovery wasn't good and it was, like, complained about a lot, and they eventually moved off of it to a continuous matching system. So I guess in practice it doesn't quite work as well, um, but it's hard to tell. It's really hard to tell.
- DPDwarkesh Patel
What country... And, I mean, in your long experience of developing financial infrastructure, what country do you feel has the best, uh, infrastructure and setup for, uh, good markets?
- BHBrett Harrison
I would say the United States, except what's happened is US companies like NASDAQ have licensed their exchange matching engine technology to other exchanges around the world. So, like, the NASDAQ OMX technology powers a number of the exchanges in Europe and some in Asia, so it's hard to sort of say that, like, the- the... is the technology American? Like, I- I guess so. I'm not sure exactly who wrote a lot of the stuff underneath, you know, NASDAQ technology. But I do think, uh, the US markets are some of the most, you know, efficient and low latency and expansive in products that allowed
- 23:10 – 32:09
Hiring & Programming at Jane Street
- BHBrett Harrison
in the world.
- DPDwarkesh Patel
How do, uh, adverse selection in trading and hiring differ?
- BHBrett Harrison
In hiring, there are, one, many more opportunities for positive selection versus the negative selection you usually encounter in finance, and the other thing is that most financial markets are, like, you know... In- in the US when you think about, like, trading in general, you're thinking about, like, liquid markets. The hiring market is highly inefficient. You know, maybe the, like, pipeline of, you know, orders from, like, Harvard, MIT, Princeton, Yale to Jane Street and Citadel Securities is, like, a very liquid pipeline-
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
... but, like, there are many, many universities and colleges throughout the country and the world that have extremely talented individuals whose resumes will, like, never end up on your doorstep. So you might end up with, like, a resume from, you know, some, you know, graduating senior from college who has no internship experience and your trader mindset might think, "Okay, this is terrible adverse selection," but it actually could be that, like-... that person, if, if he's willing to, you know, put themselves out there and apply to your company from this, like, uh, relatively unknown university, then that might be to signal that is, like, the best person in that entire region, and that might be a positive selection. And so I think that it, it, it's not exactly the same, like, adverse selection dynamics as there is in, uh, traditional trading worlds.
- DPDwarkesh Patel
Yeah. Yeah. Definitely, especially if you have, like, uh, I guess mission-oriented companies have a especially good way of getting rid of adverse selection, right?
- BHBrett Harrison
Yeah. Y- yeah, exactly, like, the companies with really strong brands.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Uh, I mean, that's one of the things we saw at Jane Street was, like, I, I heard stories in the old days of Jane Street that, like, the first resumes from Harvard were, like, the people were terrible.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Like, they couldn't do, like, basic math and they just were this, like, the worst candidates compared to other people that they were able to find. And then they established this brand and this recruiting pri- pipeline and this reputation for having very difficult interviews and for paying people really well and having this, like, amazing work environment that, all of a sudden, all the people getting through the pipeline from Harvard were, like, really, really great. And it wasn't like the quality of students at Harvard changed, there's probably a bell curve there like there is everywhere else. It was just, like, the positive selection resulting from the branding efforts-
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
... and the, like, mission-driven, like, you know, focus of the company that really brought that positive selected pipeline to them.
- DPDwarkesh Patel
Yeah. That's really interesting. Should Jane Street replace OCaml with Rust?
- BHBrett Harrison
No, because there's too much infrastructure already in OCaml.
- DPDwarkesh Patel
Yeah, but starting from scratch?
- BHBrett Harrison
So I guess the world is, um, if they could, like, snap their fingers and suddenly replace all their OCaml infrastructure with Rust at, like, zero cost-
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
... like, would it be worth it?
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Um, in that case, I would say yes, because I think that you get a lot of the sort of static typing and compile time safety in Rust that you get from OCaml, but the base level program that you can write in Rust is much, much faster than one you can write in OCaml. Because of the way OCaml's designed, where there's this kind of automatic garbage collection, the worst thing you can do in high-speed finance is do any memory allocation that results in garbage collection. And so you have to write very, very careful OCaml that almost ends up looking like C in order to end up, you know, staying in functional programming land but not actually creating tons of memory on the heap or the stack that ends up getting collected later.
- DPDwarkesh Patel
I, I guess you've been playing around with Rust a lot recently, right?
- BHBrett Harrison
Yeah.
- DPDwarkesh Patel
What, what are, what is your impression of the language? Have, have you been enjoying it?
- BHBrett Harrison
It's great, and it's come a very long way in the last three to five years. And I think crypto has something to do with that. It seems to be, like, one of the, like, languages of choice for people to write, like, blockchains and smart contracts. And so there's been an enormous amount of, like, open source, uh, contribution to Rust. And so comparison of when I, like, last looked at it a couple years ago, it's a lot easier to write, like, really good sophisticated, uh, programs now in Rust and get all of, like, the type safety and the speed that you get, which is, like, very comparable to C++ on the speed side.
- DPDwarkesh Patel
Well, when I'm writing programs, they're not large code bases with many people contributing, so when I use Rust, it's just, like, a huge pain. And, like, why is... I just want to do something very simple, why do I have to put, like, an arc inside of a box instead of an
- NANarrator
Yeah.
- DPDwarkesh Patel
... option? Like, it just... But, um, I can totally understand if you were, have something where, like, billions of dollars are at stake and-
- BHBrett Harrison
There's definitely a learning curve.
- DPDwarkesh Patel
Right.
- 32:09 – 42:10
Quant Culture
- DPDwarkesh Patel
at these sorts of places, like, uh, Citadel or Jane Street. I mean, you spend some time in Silicon Valley and around, like, traditional sort of, like, startup scene as well. Um, what is the main difference between Silicon Valley tech culture versus, you know, quant, uh, New York culture?
- BHBrett Harrison
Sure.
- DPDwarkesh Patel
Or, I guess, Chicago culture. Yeah.
- BHBrett Harrison
Yeah. So I- I don't- I have a ton of personal experience, like, in- in the Silicon Valley culture or, like, the tech culture since I've only really worked at kind of finances my- finance places my whole life. But the sense I get is that the kind of New York, Chicago quant finance dev culture is one about extreme pragmatism. Like, you know what the outcome is, it's like, you know, be the most profitable at the strategy and you kind of try to draw a straight line between what you're doing and that profitability as fast as you can. Compared to, I think, the Silicon cr- culture is much more about, like, creativity and doing things that are, like, new that no one else has done before. A healthy amount of cross-pollination would be good for both where I think a lot of trading firms are doing, like, the exact same things that all the other trading firms have done. And some, like, healthy, you know, injection of creativity into some of that stuff to maybe think slightly outside the box of, like, you know, as you said earlier, like, get slightly faster to go to NASDAQ or something. Which is like, okay, it's, probably be fine but it's like not that- not that creative, um, would be good for those quant firms. At the same time, the, like, sheer approach to pragmatically getting something done and out there and sold and making money would help a lot of Silicon Valley firms that kind of, like, hang out in this sort of creative land for too long and don't end up getting a product to market.
- DPDwarkesh Patel
Yeah, yeah. No, I- definitely, uh, it seems like there should be one founder from both those cultures at every single startup.
- BHBrett Harrison
(laughs)
- DPDwarkesh Patel
It's similar to what you were saying earlier with, like, SBF and, like, visionaries versus pragmatists in that context. Um, how conspicuous... I mean, uh, you were just mentioning earlier that these traders are making mid six-figure salaries to begin with, let alone where they rise over their careers. How conspicuous is their spending and lifestyle? Is it close to Wolf of Wall Street? Is it just, uh, Walmart T-shirts? Like, where- where are we talking?
- BHBrett Harrison
Uh, it- it's probably a lot closer to Walmart T-shirts than it is Wolf of Wall Street. Um, I mean, certainly it is now. Even when I started it was pretty inconspicuous. I don't think it was that way in, like, the previous decade or two before I joined finance, I guess I'm not really sure. But I get the sense that the current culture around inconspicuous consumption is sort of a function of millennial consumption habits where, you know, people are focusing a lot more on, like, experiences than, you know, having shiny material objects. I think that's had a lot- large effect on kind of, like, the high earning tech and finance culture that exists today.
- DPDwarkesh Patel
Well, I guess, are they spending that much money on experiences either? 'Cause how expensive is a flight to Hawaii, right? And, um, y- uh, even after you subtract that, like, where is this money going? Are they just saving it?
- BHBrett Harrison
Maybe it's not, like, just a flight to Hawaii but it's, like, bring your 10 friends to Hawaii with you with you or something. Or it's, uh, you know, get involved in, like, a charitable organization in a way that, like, someone who is 24, like, normally wouldn't be able to do, purely by able- being able to, like, donate a lot.
- DPDwarkesh Patel
Yeah. What is the social consequence, for lack of a better word, of having a bunch of young nerdy people, often male, often single, having this extraordinary level of wealth? Like, wh- what impact does it have? I don't know if society's the right word but, like, what is the- what is the broader impact of that- that class of people?
- BHBrett Harrison
I think we'll have to play this out over the next decade or two to really see where this goes. If I'm gonna be an optimist about this, I'd like to think that when it was, like, older, you know, single or married males-
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
... you know, getting, you know, hoarding a large amount of wealth that, for the most part, they kept it to themselves and kind of waited until later in life to do anything with it and were the kind of people who really, like, stayed in their same career their whole lives. As opposed to...... if younger and younger generations are amassing wealth through, you know, the, what, what they can actually perform with their skills, then I think that hopefully injects more dynamism into the distribution of that wealth later on-
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
... because those millennials will then, like, or Gen Z or whoever, will go on to, like, found new companies and maybe they'll u- they'll be able to seed the com- the company themselves with their own money and have a lot easier time, like, bringing, like, interesting new things to market. Or they'll be able to donate to, like, really interesting causes, or they'll be able to, you know, help out their friends and family more easily from a young age, or they'll be more selective in the kinds of things that they, you know, give to you or contribute to that don't just involve, like, getting their name on, like, a building of a school or something.
- DPDwarkesh Patel
Yeah, yeah. That's a very optimistic story. I- I hope that's the way it plays out. Um, so tell me about the psychology of being a quant or a trader or a developer in that space because you're responsible, like one lo- wrong keystroke and you're, you've lost millions of dollars, you know, one bug in your code and... There are, you know, historical cases of this, like where entire firms go down because of a bug.
- BHBrett Harrison
Yeah.
- DPDwarkesh Patel
Like, wh- what is the sort of like day-to-day psychological impact of that kind of, uh, responsibility?
- BHBrett Harrison
It... Maybe the, the job selects for the people who don't kind of crumble under the like, you know, theoretical stress of that job. But personally, like I don't lose sleep overnight, at- at- at night over that because within any, like, mature financial institution, like a trading firm, there are typically, like many layers of safeguards in place. Like, you know, limits on like how many dollars you can trade in a minute and how much you can trade overall or for your desk or like how many messages you can send to the exchange. And then there's like limits on like the individual trader and desk level and firm level and there's like layers of different checks. Often, there are actual rules, like regulatory rules to comply with in market access checks, like you know, FINRA is like 15c3-5. And so when you're writing new code, it's not like a completely, like blank slate thing where you're connecting directly to an exchange and like hoping for the best. Usually, you're embedding some piece of code within some very large established framework where the goal is to make something trader-proof. No matter what some trader clicks on or does or configures with their system, there's like a limit to how, uh-
- DPDwarkesh Patel
How badly they can mess it up. Yeah.
- BHBrett Harrison
... how badly it can actually go. Um, and so especially in my particular role as a developer, like actually being able to understand the techno- technological stack and say like, "Oh, I can tell and can sort of, you know, verify that, like these particular safeguards are in place and it is like actually as trader-proof as I think it is," like I sort of, I can sleep at night knowing like nothing too bad's gonna happen. I mean, the times when I actually lose sleep are like, you know, uh, uh, a trader in like London or Hong Kong calls me like in the middle of the night to say like, "Hey, uh, can you explain how this thing works? (laughs) I need your help." Like those are the times where like I actually lose sleep.
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
But it's not over like being c- concerned about risk.
- DPDwarkesh Patel
Yeah. Uh, uh, that's interesting. If you asked the people who work in these firms, "What is the social value you're creating?" separate from the question of what the correct answer to that question is-
- BHBrett Harrison
Yeah.
- DPDwarkesh Patel
... what would the majority of them say? That like, "I'm doing something really valuable." Would the mo- Would they say like, um, it's like, "I'm indifferent to it but it's earning me a lot of money?" Like how, what is, what is their understanding of the value they're creating?
- BHBrett Harrison
It really depends on the company and it depends how diffuse the culture is. At like older firms that have like, um, you know, sort of fewer people impacting the culture on any significant way, I, I think you might not get a clear answer on this. I think for a place like Jane Street where, you know, the firm is really run by like 30 or so, you know, partners and senior employees who have like been there for a really long time and have carried through the core culture of that company up 'til the present day, and with that, like large number of people at the top in a very flat environment, have actually been able to like propagate that culture and maintain it throughout the company, I think you'll find a much more kind of homogeneous view on their social value, which I think they would say is that they provide like the best, like pricing and access to markets that are critical for facilitating like capital allocation throughout the world and allow people to very efficiently invest in vehicles that are global in nature.
- DPDwarkesh Patel
That seems very, um, abstract. And while it, while like it, it co- it is probably very well correct and is very valuable for society, it might not seem that like tangible to somebody who's working in that space. Is there some technique that these firms have of vis- making visceral the impact these traders have? I don't know. Do they, do they bring out some like child who benefited from efficient markets or something? (laughs)
- BHBrett Harrison
(laughs) Uh, I think... Well, probably not like children. I think it's-
- 42:10 – 58:20
FTX - Meeting Sam, Joining FTX US
- DPDwarkesh Patel
Okay. Let's start by addressing the elephant in the room, which is FTX. Let's begin, uh, at Jane Street-
- BHBrett Harrison
All right.
- DPDwarkesh Patel
... which is where you met SBF. Can you tell us sort of the origin story of how you first met him and what your first impressions were?
- BHBrett Harrison
Yeah, absolutely. So I was at Jane Street from 2010 to 2018.... a Sam was at Jane Street for a couple years in the middle of that, I think 2013 to 2017. And one of the things I did at Jane Street was I started this program called OCaml Bootcamp. It was a yearly course for the new trader hires to spend four weeks with me learning programming, uh, in- in OCaml, which was, like, the esoteric programming language that we use at Jane Street, along with a lot of our other proprietary systems. And Sam was in our- was in one of the first cohorts of students, uh, and so I got to meet him through that experience.
- DPDwarkesh Patel
Got it, okay. And what was your impression of him?
- BHBrett Harrison
Yeah, he was a smart kid. He was nice. Uh, he kind of got along well with other people in his class. You know, he was definitely, you know, above average, but not, like, you know, completely stand out at the top, although then again, the bar was extremely high at Jane Street, so I think that's already sort of a compliment. But, uh, yeah, but, you know, people liked him a lot and thought he had a lot of promise, but, you know, he was a young guy like everyone else.
- DPDwarkesh Patel
Got it. And did that perception change over time while you were at Jane Street?
- BHBrett Harrison
Uh, it slowly started to. You know, um, Sam was on one of the largest trading desks at Jane Street, uh, and, you know, had 50 or 60 people on it. He had several managers. And one of my roles at Jane Street was to work with all of the different trading desks on the designs of their particular strategies and systems, and so I would frequently go over to his desk and talk with, uh, his managers about stuff, and I started pulling him into conversations more and more, uh, specifically to talk about some of the, like, lower latency, ETF ARB stuff we were doing, uh, some of the original, like, OTC automation things we were working on. And so he started actually contributing more to the actual, like, design and thought behind some of these systems, and thought he was, you know, precocious and had a lot of really good intuitions about the markets.
- DPDwarkesh Patel
Got it, okay. Um, and so what exactly was your role at Jane Street at this time? And what was his?
- BHBrett Harrison
Yeah, so at- at this time, I was sort of leading the group of software developers building the technology that was closest to actual trading. You can think in a HFT or any kind of sort of trading firm, there's lots of different developers, people who work on, you know, stuff really relating specifically to the- to the trading technology, people who work on kind of the core systems, networking, kind of internal automation tools, tools for developers, so we were on the part of the spectrum that was closest to actual trading. And so my job was to, like, go over to the different trading desks within the company, talk to them about their specific strategy for the products they traded, understand how to, like- their priorities about what venues they want to connect to, what different systems they want to create, what different parameter changes they need in their, like, a- automated trading systems, what kind of research tools will help them do their job better, what user interface would make it easier for them to understand what's going on in the market, and kind of all of that.
- DPDwarkesh Patel
Okay. And w- did SVF at this point have any sort of reputation of either being uncooperative or being cooperative or anything ethical or professional th- that's noteworthy at this time?
- BHBrett Harrison
I don't think there was much that stood out, although, you know, he was, again, you know, pretty, you know, precocious, uh, at- at that particular time period. O- one anecdote that sort of drew me closer to him was Jane Street's offices were in 250 Vesey, uh, they still are, in New York City, and there's a big food cart on, like, the second floor. And so I once went down to meet with a development person from a, uh, a nonprofit that works in animal welfare, and something that my wife and I had donated to for a long time, and I met with this guy and he said, "You know, you're the second person I met from Jane Street today." Which was wild because Jane Street was, like, only a couple hundred people there. It was just, like, a pretty niche-
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
... you know, organization. And, you know, I was like, "That's crazy, you know, who did you meet?" And they said, "Oh, you know, Sam Bankman-Fried." And I was like, "You know Sam? I just came down from talking with him upstairs." And so I went back and we sort of realized we had this kind of shared interest in helping kind of animal welfare causes, we were both, you know, vegans, uh, and we sort of bonded over that, and that's how we kind of became friendly.
- DPDwarkesh Patel
Got it. And did- so it seems then his interest in effective altruism was genuine at this point and early on it was- there was a history of this?
- BHBrett Harrison
Yeah, you know, it wasn't like EA was super popular at Jane Street. I feel like that's a bit of, like, recent sampling bias. Among this, like, younger crew of, uh, Jane Streeters, there definitely- I think it was because of a lot of, uh, association prior to joining Jane Street, uh, that they sort of were into effective altruism, but there were a couple people there who really were fairly vocal about the fact that they were, you know, donating the majority of their yearly salary and bonus to charitable causes, and Sam was one of them, and, uh, and yeah, started to become known for that.
- DPDwarkesh Patel
Got it, okay. So I guess fast-forward to, uh, he's no longer at Jane Street, you're no longer at Jane Street and you're at Citadel. He started FTX. A- actually before we back, uh, go there, were you in contact with him up until the point where you had started talking about a potential?
- BHBrett Harrison
Yeah, off and on. Uh, you know, when- when I first left Jane Street and he left Jane- uh, sorry, you know, when we both left Jane Street around- around the same time, him before me, you know, he had told everyone at Jane Street that he was leaving to join the Center for Effective Altruism full time. And I- I guess he did that, I'm not sure, (laughs) you know, if it actually happened because he very soon after started this trading firm and tried to pull lots of Jane Street people to join him to do this trading firm which didn't make people super happy.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Uh, but it was funny, he- we had a phone call and he told me that it wasn't really going super well. Um, he said it was really great in the beginning, like, they made a lot of money, they had this arbitrage trade, and then a few things kind of went by the wayside and, you know, they had taken out these huge loans to be able to get their initial capital for Alameda, and also there was a big, uh, fracture within the company, you know, half the company split, people left, he really didn't tell me much about that at the time, and he said he's probably gonna do something else. And when I asked him, he said, "I think I'm gonna work on, like, political prediction markets." And I was like, "Okay, it doesn't sound (laughs) super exciting to me. I'm gonna continue on with what I was doing," which was moving to Chicago, taking a new role-... but then fast-forward, I guess that idea, maybe he wasn't telling me the whole truth at the time, but I guess that idea became FTX and he had, uh, supposedly, like, resuscitated Alameda in the process.
- DPDwarkesh Patel
Yeah, yeah. Um, that's really interesting. Do you have some sense of what it was that went sideways?
- BHBrett Harrison
So I pieced together some details over the years because he told me a little bit more after I first joined, I heard a little bit more later from other people, then saw some reporting kind of post-FTX collapse. Uh, I think there were two things. Um, one was, the infrastructure they had built, I think was really poor in the beginning, a lot of, like, Python scripts, like, slapped together, and a couple of times they had sent tokens to the wrong wallet and ended up, like, losing millions in the process, and they had some big non-arbitrage directional bet on in some token. It might have been ETH or something, and it went against them, and so they lost a lot of their trading capital. And then the other thing was that be- after some of their technical problems, they, there was internal disagreements, uh, supposedly, this is what Sam told me, about how to move forward with tech. You know, there was half the crew that wanted to kind of re-write everything from scratch in a different programming language. There was another half that said, like, "Okay, we can make some small incremental changes from here and fix things up," and Sam, and Gary, and Nishad were more in that latter crew. That former crew kind of broke off and started their own thing, and that's what originally happened.
- DPDwarkesh Patel
Okay, got it. And were you aware of the extent of this at the time, or-
- BHBrett Harrison
No.
- DPDwarkesh Patel
... is this something you pieced together, or-
- BHBrett Harrison
Yeah, not at all. I mean, uh, Sam told me a little bit about it but, uh, this was over the course of, you know, years now, where I had two different roles, one at Headlands, one at Citadel Securities, Sam was starting Alameda, we spoke maybe once a year briefly on the phone, um, so all this stuff was happening in the background and I had no clue.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
Like, the first time I even heard about FTX was one of my colleagues from Citadel Securities told me, "Hey, do you, did you ever work with this, like, Sam Bankman-Fried guy?" And I was like, "Yeah, a little bit, why?" And they're like, "Do you know he's, like, a billionaire and he has this, like, like, Hong Kong crypto exchange?" I'm like, "What? No." (laughs) "What, since when?" And then started to see him pop up in articles. There was, like, a Vox article about him and a few other things, especially related to his political donations, and that's kind of when I got back in touch with him and we started talking a little bit.
- DPDwarkesh Patel
When was it that he called you to say that there are s- potentially troubles and I'm considering starting a political prediction market?
- BHBrett Harrison
That was in 2018.
- 58:20 – 1:08:11
FTX - Accomplishments, Beginnings of Trouble
- DPDwarkesh Patel
set up, but yeah, f- feel free to talk in more detail about what it, what were the things that came up during your tenure and what are the accomplishments you're, you know, proud of?
- BHBrett Harrison
Yeah, I guess on a, the professional and personal front, so I guess on a, on a professional front, I'm, you know, most proud of, like, establishing out our team and making significant headway to a lot of our goals to establish these regulated businesses. So for example, LedgerX, you know, we acquired LedgerX and we had this application to the CFTC to enable kind of real time direct to customer 24/7 margining and cross collateralization. And it was an extremely innovative proposal and it felt like we were making real progress towards, uh, establishing, like, new and very exciting regimes for, you know, CFTC regulated derivatives in the US. They also established a broker dealer in the US for the purposes of letting people trade stocks, like similar to Robin Hood. I wrote, like, 90% of all the code for (laughs) that stocks platform myself. Um, and the, yeah, I was very proud of that accomplishment. And then it was, on a personal front, I, it was great to get embedded into the crypto industry. I was very excited by everything that I saw. It was great to make all the connections, you know, through FTX with, like, the different people in the crypto ecosystem and, you know, become friends with these people and that certainly has influenced where I am today. Um, so I'm sort of proud of all of that.
- DPDwarkesh Patel
Mm-hmm. How did you manage the management of, uh, well, I don't, I don't know how big the team was at its peak, and it sounds like you were heavily, I mean, involved is an understatement, in the actual engineering. W- how were you able to manage both roles at the same time?
- BHBrett Harrison
Yeah, so we were, like, between 75 and 100 total people in the US and it was challenging.
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
It was one of my big, biggest complaints-
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
... which I'm sure we'll get into-
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
... which was that, you know, yes, I can write code, but I feel like that's... My comparative advantage is helping kind of to leverage teams of people to get them to, you know, work towards the common goal of, you know, building out large distributed systems that are, you know, complex and multivariate in nature. And the best use of my time was not, you know, me programming between the hours of, like, 10:00 PM and 2:00 AM every night-
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
... while trying to keep on board with, like, what all the personnel were doing.
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
So, eh, eh, so I really wanted to grow the US team significantly to at least-... be, you know, more than a handful of, like, developers. Uh, and so yeah, that, that was one of the initial points of contention.
- DPDwarkesh Patel
Okay. S- s- speak more about that. So he was opposed to growing the team?
- BHBrett Harrison
Sam would frequently talk publicly about how proud he was that all of FTX is built by, like, two developers-
- DPDwarkesh Patel
Yeah.
- BHBrett Harrison
... and all of these crazy organizations that hire thousands of developers and can't get anything done, like, they should learn from me-
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
... about how, like, a small lean team can be, you know, much more effective. And there's some truth to that. You know, I do think the conventional wisdom now is a lot of big tech companies over-hired for software engineers, and not only was it sort of an expense on the balance sheet, but it was also expensive in terms of slowing down the kind of operational efficiency of the organization. And having a small lean team can help you get to your, you know, your first or your nth product a lot more quickly. That's great for a startup, but once you're, like, a north of $10 billion valuation company-
- DPDwarkesh Patel
Right.
- BHBrett Harrison
... like, promising the world to, like, customers and investors, two software developers doesn't really cut it anymore. I mean, at some point you have to, like, you know, grow up and, and, you know, face the reality that-
- DPDwarkesh Patel
Right.
- BHBrett Harrison
... it's time to actually grow an organization into a real kind of managed enterprise with teams of software engineers specializing in certain tasks. And so, th- there was always pushback, you know. People would tell me, like, "Look, we're not trying to be like Jane Street or, or Citadel in terms of our number of, you know, software engineers. Like, we wanna stay lean. That's our comparative advantage." And most importantly, they didn't want two separate development teams, like one in the U.S., one in then the Bahamas. Like, they wanted to keep the nexus of software development underneath, you know, Nishad and Gary in the Bahamas, which I just thought wasn't going to be sustainable long term. Like, if you run a broker dealer in the U.S., you need to have staff that is specifically allocated towards broker dealer activities. It can't be that if s- FINRA comes and says, like, "Well, who's working on the broker dealer?" You say, "Well, it's, like, this Gary guy who-"
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
... "like, lives in the Bahamas, who's sometimes awake at, like, 4:00 AM (laughs) and spends 20 minutes a day thinking about stocks." Like, th- that can't fly.
- DPDwarkesh Patel
Right, right. And has no images of him on the internet (laughs) .
- BHBrett Harrison
No, yeah (laughs) .
- DPDwarkesh Patel
Um, okay. So, uh, were, so were Nishad and Gary contributing code to the FTX US code base?
- BHBrett Harrison
Remember, like, the FTX US side of things was a, like, strict subset of FTX. So, like, in that sense, it kind of flowed into FTX US with the exception of the, like, FTX US derivatives. The LedgerX stuff was, like, actually a completely separate team 'cause that was through an acquisition.
- 1:08:11 – 1:26:53
FTX - SBF's Dysfunctional Leadership
- BHBrett Harrison
- DPDwarkesh Patel
Is there a spec- specific example or set of examples that comes to mind?
- BHBrett Harrison
Y- sure. You know, I, the biggest example for me was, it was just a sort of post my kind of effective resignation, but some of these strategic acquisitions that were being done in the US during the summer of 2022, I would find out from, like, the news or, or, like, it would sort of be mentioned on a Signal chat or something that this was happening. And there was, like, no opportunity to actually, like, wade into the discussion about how this is, like, gonna greatly affect, like, the US business. It's gonna greatly affect our priorities and it wasn't clear if this was, like, a good decision or a bad decision. It was just, like, a unilateral decision that was made, like, oh, this co- we're acquiring this company or we're, we have the option to acquire this company.
- DPDwarkesh Patel
Are there decisions that were made from the Bahamas that stick out to you as being unwise, uh, that, like, I don't know, you try to speak out against? Um, I, I, I mean, you mentioned some of them, right? Like not hiring enough people and, uh, not that, getting more, yeah, and not getting more developers. But are there other things like that that stick out to you as bad decisions?
- BHBrett Harrison
A lot of the spending, I mean, on everything from, like, lavish real estate to all of these, like, partnerships to, like, very, very large venture deals. Like, these were the kinds of things in the company where people asked, like, "When does this stop? Like, to what end are we doing a lot of these things?" And, you know, some of those resulted in sort of, like, direct confrontations, like, just, you know, why are we doing yet another deal with like a sports person or a celebrity? This is like, this is ridiculous. (laughs) Like, this is not doing anything for the company and we're completely distracting from the role that we thought we all had, which is to build a really great core product for, you know, people trading crypto and crypto derivatives.
- DPDwarkesh Patel
Yeah. And did you bring this up directly with SBF?
- BHBrett Harrison
Yeah, multiple times.
- DPDwarkesh Patel
And wha- how would he respond?
- BHBrett Harrison
Sometimes he was, you know, nice about it and he would say, you know, like, "Yeah, like, I see where you're coming from. I do think what we've done so far has been really valuable and we probably should do some more of it, but maybe at some point we should stop." You know, a lot of this sort of, like, hedging language that was ultimately non-confrontational, non-committal. I mean, he was a very non-confrontational person, very conflict avoidant person within the company. So it... And then at- at- at worst, he would just, you know, there were other times where I brought up, like, specific things that I thought, like, he was doing wrong. There was one really unfortunate time where I, it was the first time I visited the Bahamas, uh, in I think it was, like, November of '21 and I'm the kind of person who, like, if I see something wrong at a company, it doesn't matter what company I worked at or how junior or senior I've been, like, I like to go to the person most senior in charge and tell them, like, "This thing seems wrong to me." And that's, I feel like, is one of my, like, superpowers, uh, of just, like, not being afraid and just, like, saying when something seems wrong to me. And sometimes, like, I'm just totally wrong and don't understand the full picture, and sometimes it results in something better happening and people will, you know, thank me for having been honest and bringing to attention something that's actually wrong. Uh, and so I said to Sam, "You know, I think you're doing way too much PR and media." Like, first of all, it's really diluting you and the FTX brand to constantly be doing TV interviews and podcasts and, you know, flying to banking and private equity conferences and, you know, it was so much time spent on this stuff. And also, it was completely taking away from the management of the company. You know, people would sometimes send Sam a Slack or Signal messages and not get responses for weeks at a time, and it felt like he was spending virtually no time helping the company move forward. It was so much about image and brand and PR. And he was really angry at hearing this criticism directly.
- DPDwarkesh Patel
How did- how did he react?
- BHBrett Harrison
I mean, he was just, uh, he- he was sort of emotional, he was worked up. He told me, like, "I- I completely disagree with you." I mean, he said, like, "I think you're completely wrong." He said, "I think the stuff that I've done for PR is maybe the greatest thing that's happened to this company. I should do more of it." (laughs) I didn't think it was physically possible to do more of it.
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
Um, and it- it... I realized at that moment that this was not really gonna work super well long term. Like, if we're not in a relationship where I can give sort of my direct superior, like, real honest, you know, constructive criticism, um, that I thought was for the good of the company, that this wasn't really going to work.
- DPDwarkesh Patel
He actually did my podcast about, I don't know, eight months ago or something. And while I was, like, very grateful he did it, even at the time I'm like... (laughs) I don't know if I would have agreed to this, (laughs) if I was in charge of a $30 billion empire. Uh, but- (laughs)
- BHBrett Harrison
Yeah, I mean, if you, s- sometimes, like, some reporters would say to me, like, you know, "Do you, can you, like, get me in touch with Sam?" And I would say, you know, "Why? Like, I'm, I'm not really his keeper. (laughs) You should contact him yourself." Um, they're like, "Oh, because we want to come to Bahamas and do a special on him." And I always say like, "Okay, well, you're gonna be, like, the sixth one this month."
- DPDwarkesh Patel
(laughs)
- BHBrett Harrison
(laughs)
- DPDwarkesh Patel
There's no exclusives here. So I guess to steel man his point, he did get a lot of good PR at the time, right? Potentially, well, not potentially, like, actually, um, uh, uh, too much and in a way that, like, really created a, at the time, sort of like the king of crypto sort of image. So, I mean, was he right about the impact of the PR at the time?
- BHBrett Harrison
(laughs) Yeah.
- DPDwarkesh Patel
Or may- may- maybe, yeah, so may- let me ask the question a different way. How did he create this image? I mean, people were saying that he's the J.P. Morgan of crypto, like, he could do, uh, do no wrong. Even things that in retrospect seem like clear mistakes, like only having a few developers on the team, uh, universally praised, you know, huge empire run by a few developers. How, how was this image created?
- BHBrett Harrison
I think the media was primed for the archetype that was Sam, this sort of young, upstart prodigy in the realm of fintech. You know, we have a lot of these characters in the, the world of, you know, big tech, and I think that he had a particular role to play in the world of finance. And by making himself so accessible all the time, he gave people, uh, a drug that they were addicted to, which was, like, that constant access. And so, like, any time of day or night, someone could text Sam and get him on the phone with them if they were in media. And they loved it. It was like getting access to a direct expert who was also this famous person, who was also this billionaire, who was also this extremely well-connected person, who was also this, um, very insightful person, who knew a lot what was going on in the industry and can give them, like, insight and tips.
- DPDwarkesh Patel
Mm-hmm.
- BHBrett Harrison
And I think there was s- some amount of, what I like to call, like, reputation laundering going on here, where it was like, okay, so you get the famous celebrity to endorse Sam, which makes, you know, this, you know, politician think highly of Sam because they also like that celebrity. And then also, the investors are writing really, like, great, positive things online about him, but also the media is enforcing how cool it is that Sam is doing all these other things. And it all sort of fed into this, like, flywheel of building up Sam's image over time in a way that didn't necessarily need to, like, match the, the underlying reality of who he was at the company.
- DPDwarkesh Patel
And what was the reaction of other employees at FTX of the sort of, not only the media, um, hype train, but also the amount of time Sam was spending with the media?
- BHBrett Harrison
You know, on one hand, I- I think people were growing frustrated within the company because of the lack of direction and some of, like, the power vacuums that resulted from Sam's continual absence. On the other hand, so many people within the company just hero worship Sam. You know, when you hear all, like, the really tragic stories now of all the employees who kept all of their, like, funds and life savings on FTX, they really, really believed in Sam. And doesn't matter how little time he spent with the company, doesn't matter how he treated employees internally, it was like he was this sort of, uh, genius pioneer and that image couldn't be shaken.
- DPDwarkesh Patel
And I- I certainly don't blame anybody for it. I mean, I interviewed him, I tried to do a lot of research before I interviewed him, and I certainly was, like, totally taken with this, right? I thought he was the most competent person who had ever graced crypto. But, so what was he actually like as a manager and leader, um, other than, I guess, obviously the micromanaging aspect of it? Um, or f- feel free to speak more on that as well, but in terms of the decisions he would make, in terms of business development and prioritizing things, wh- wh- can you describe his sort of management style and leadership?
- BHBrett Harrison
In the beginning, when I joined FTX, my initial impressions were that he had pretty clear intuition and insight into the simple things to do that would work. You know, it, in- in many ways, if you think about what FTX did, uh, it wasn't really super complicated. It was like just be operationally good and give your trading customers as predictable of an experience as possible with regards to collateral management and auto liquidation and matching engine behavior and latency. And so they did it. I- I would say, aside from the intuition, Sam wasn't a details man (laughs) . Like, that was usually left up to the people below him, uh, to really take care of, was like to drive a project to completion, to figure out all the details that had to be done. I think besides that, as a leader, I- I thought he was, you know, fairly incompetent. I thought he was, you know, very conflict avoidant. You know, he didn't like to get into direct confrontation with any of his employees, where most of the reasons why people needed to talk to him were because there were s- you know, significant issues, you know, whether those were personnel or otherwise, and he just blew them off. I mean, that was just a- a- a frequent occurrence within the company. I mean, he was v- he wa- if you went to Bahamas, and I went a- only a couple of times to actually visit the office, he was, if he was in the office, he was there all day on calls-... all day, whether those were with investors or with media, podcasts, whatever. It was just consistently just doing that. And I saw very, very little time where he actually got up and talked to anyone else within the company about anything, you know? So I think, to me, that was the primary impression I got of his leadership, was virtually that there was none, which, you know, made me feel a lot like I and others needed to step up and sort of take, you know, tho- that role in the absence.
Episode duration: 2:37:37
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode yXgDlIlB93A
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome