Note: this is a long post! You may want to read it on the site instead of in email.
Money, Credit, Trust, and FTX
“Credit” comes from the Latin credere, “to believe, to trust.” When this point gets trotted out, it's usually because someone was a little too credulous and has lost a lot of money. Trust is typically bilateral: person A trusts person B, but that doesn't tell you much about what person C thinks of them. Money makes trust multilateral. Instead of “I trust that you owe me a favor,” it’s “I trust that a favor is owed to me."
The collapse of FTX is a story about credit and credibility: about how leverage can produce fabulous wealth, but how it can also lead to vulnerable systems. It's about how a good story and a high-trust community can inspire people to work hard in order to improve the world (and get rich), but how it can also be abused.
Creating credit is one valuable thing financial systems do: there's a practical mismatch between savers and projects worth investing in, which the financial industry intermediates. And a growing economy typically operates with a maturity mismatch in real assets ($), which makes individual banks vulnerable but which can be patched up by central banks ($). Constraining credit creation is a service that's essential in any system where credit can be created. This is usually done through some combination of regulation and market action: if you're going to be a money-creating entity like a bank, you'll need to follow lots of rules; if you're going to create credit in other ways, you're always running the risk of a liquidity crisis.
Credit is a bit like the Greek myth of Tantalus. Financial assets are moneylike to the extent that you don’t need money, and very much not moneylike when you’re desperate for cash.1
The collapse of FTX raises questions about market structure—about how differences in regulation create the worst of both worlds, and whether or not investors can be fooled into pattern-matching to poorly-coiffed socially-awkward geniuses wearing ill-fitting casual clothes2, etc. But truly understanding FTX’s collapse requires asking more basic questions too, like: what is the good life? And: what is the nature of reality? Sam Bankman-Fried had unusual answers to these questions, which matters because he took those answers seriously and is a talented person who was operating in a space where he could have a significant impact quickly.
A plausible but boring theory is that SBF was a scam artist the whole time. If that were true, he probably would have blown up earlier, and at a smaller scale, or would have retired to the Bahamas instead of moving there for tax/regulatory purposes and sleeping in a beanbag chair in the office. Instead, let’s explore a plausible and exciting theory, where SBF and other FTX/Alameda employees were brilliant, idealistic, and just made some fundamental mistakes—including fraud eventually, but not right away.
Crypto Versus the Crypto Industry
We designed a system that we think will withstand huge market moves and huge volume without leading to any clawbacks. And if there’s ever a clawback on FTX, we fucked up.
It's useful to distinguish between cryptocurrencies as a technology and the companies that use them. On the other hand, it's important not to draw that distinction too finely, since it amounts to distinguishing between crypto in theory and crypto in practice. It's like putting more weight on what Marx said a communist system would do than what Stalin or Mao actually did. In other words, technologies have to be evaluated in terms of what they do, not just what they promise. Don't get me wrong—the promise is still important: the technology adoption curve is different from the ideology adoption curve, because new technologies can be deployed into small institutions that grow into large ones, while changes in the function of a government or economic system tend to be one-off.3
The original idea of crypto is that instead of trusting centralized intermediaries run by humans, like your local bank or a payments provider such as Western Union, you can trust decentralized protocols. This idea turns out to have some surprising implications: it means that crypto projects are composable, that tokens can be airdropped in order to bootstrap adoption, and that anyone can spin up a new currency whenever they feel like it.
It also means there's an $800bn+ asset class with poor discoverability, lots of volatility, 24/7 activity, with asset prices mostly quoted in US dollars. And this has created an incentive for centralized entities like FTX, Binance, and many dearly-departed exchanges, to perform three functions:
- Provide a nice user interface, so interacting with crypto can mean tapping on an app rather than running something from the command line.
- Provide a nice regulatory interface, connecting the legacy financial system to the crypto system.
- Offering leverage: those assets are volatile, but speculators love more volatility. (Given someone a stock that rockets up and down for no discernable reason, like Gamestop or AMC, and they'll want to trade derivatives on it, too.)
That leverage is exactly what users are asking for, but it's not part of the original spirit of crypto. Crypto is designed around a hard money system, where "money" represents finite assets owned by specific people or entities. A system with credit is usually one where the vast majority of money consists of balance sheet entries that are not backed one-to-one with finite, physical currency. The US has $2.3tr of currency in circulation backing $21.4tr of bank deposits and deposit-like assets. Clearly we'd need a lot more physical currency to maintain our current level of economic activity in a system where credit creation was harder!
Crypto intermediaries create credit in a variety of ways: there are direct loans, where someone can buy $x worth of a cryptocurrency with less than $x in cash upfront. There are derivatives enabling levered speculation, which FTX focused on. There are stablecoins, which are supposed to be backed 1:1 with cash but which sometimes aren't. Crypto purists long for Satoshi to come down from Mount Sinai, see his erstwhile followers worshiping a calf made of credit and fiat, and wrathfully address the situation. It hasn't happened yet.
Leverage in the real economy enables growth that would otherwise happen, but not at the same speed. So the parts of the crypto ecosystem that allow leverage grow much faster than the ones that don't. That's why it took Coinbase almost a decade to reach the point where it could go public, but FTX only a few years to become one of the biggest crypto exchanges.
It gets tiresome to hear this, but it's technically true that the levered collapse of an obscure centralized institution is an argument for decentralized and transparent financial systems. But it also indicates some combination of:
- A decentralized system can create opportunities for centralized layers that are run much less responsibly than the rest of the financial system,
- The persistent blowups in crypto may indicate that it’s a sound technology that just isn’t compatible with human behavior, and
- There may be a load-bearing level of opacity and deception in the financial system that actually creates wealth.
Steelmanning Stochastic Fraud, Or: The Kelly Criterion and Quantum Financial Suicide
I think I was a pretty good manager when everything was going really well, conditional on there being no real problems... As soon as sh-t hit the fan, I had no idea what the f-ck to do.
—SBF, Conversations with Tyler
The sudden collapse of FTX/Alameda raises the question of how long it was misbehaving. There's a plausible story that early Alameda, at least, was indeed highly profitable. The big trade SBF talked about is taking advantage of the price discrepancy between Korean and Japanese crypto prices and US prices. Korea had a bigger arbitrage, and stricter controls; Japan required some fancy footwork but generated 10% returns per weekday. SBF worked at Jane Street from 2014 to 2017. They pay well, and he seems to have had a low personal burn rate. It's plausible to imagine him exiting with a six- or low-seven-figure sum stashed away. Call it $500k. A 10%-per-weekday rate of compounding means doubling the money every 10.2 calendar days. Doing that for two months straight turns $500k into $30m. Doing it for four months is enough to mint a billion-dollar fortune. It certainly demands an explanation that SBF got to a $10bn fortune at the age at which he did, but it's not impossible for this to have been legitimate.4
At some point the trade that compounds at 10% per weekday stops scaling, whether it's because of competitors or because one liquidity provider is trading in enough size to offset the premium, but there are many strategies available when supply and demand are fragmented across exchanges and many traders are levered. Given traders who have sufficient energy, merely taking advantage of (or precipitating) cascading liquidations in markets where the main traders are asleep—i.e. doing short-term momentum trading on Binance during the wee hours in China and doing the same thing on Gemini in the dead of night in the US—is a potential source of alpha.
And we have evidence for this: Alameda was a frequent presence on the BitMEX leaderboard for most profitable traders. Blockchain intelligence company Arkham has identified a wallet associated with Alameda that generated peak profits of almost $2bn by May of 2021, with declines since then including a one-day $600m loss during the Terra/Luna collapse. This is not a comprehensive look at their performance (they may have other wallets, and this wouldn't show derivatives transactions or tradfi transactions like venture investments). They were trying to borrow money at 15% in 2018, claiming 110% annualized returns. Which sounds suspicious, especially since they claim this is a no-risk opportunity. But:
- There were not many sources for crypto-related leverage in 2018. To find a willing lender, you'd need to find someone who understood crypto and systematic trading, and had enough liquidity to make a meaningful loan.
- They could have raised equity capital instead—people tried!—but if those returns were real, then getting paid 2 & 20 would mean an annual cost of capital of ~86%. Borrowing, even at high rates, is a much better way to fund a capacity-constrained by high-sharpe strategy.
So a likely story is that, circa 2019, Sam Bankman-Fried had a wonderful money machine that could easily generate tens of millions of dollars in annual profits. Why put any of that at risk by starting an exchange, much less by eventually moving client funds from the exchange to the trading firm and creating a "backdoor" that hid the evidence?
Answering this question goes back to Sam Bankman-Fried's fundamental beliefs. He's a self-described utilitarian, and takes this belief very seriously. And as it turns out, utilitarianism has some interesting implications for risk management.
Specifically, the north star for intelligent risk management is the Kelly Criterion, which indicates the optimal size of a bet given particular odds of winning and payoffs. Kelly is designed to maximize log(wealth); in other words, identifying the bet size that will produce the highest compounding returns without any risk of either a) losing everything, or b) equivalently, losing so much that there's no way to recover. Most people should not aim for maximum wealth, since it also means accepting the highest recoverable drawdown, which can be very high indeed.
But for a utilitarian who plans to give away all of their money, the utility of wealth is roughly linear for all plausible sums one person could earn. There are 700 million people who earn less than $2/day. Suppose you decide to top them all up by $1. You need $25bn each year to accomplish that. And if providing for the poor is a) good, and b) good in proportion to how much you provide if you're spending at a smaller scale than major governments, then you should make all the positive-EV bets you can, even if most of them will not pay off!
Was the utilitarianism stuff just bluster to defend a sociopathic moneymaking scheme? Probably not. SBF was writing about utilitarianism for years before he started Alameda, and before he was even working in finance. So it's reasonable to treat this belief system as action-guiding.
And these beliefs come with a meta belief system, around biting the bullet even if it's weird. This kind of linear thinking can sometimes compound, especially if it leads to selecting a peer group that thinks the same way. SBF has acknowledged that he did this. In this Reddit thread discussing which college he should attend, he says:
So, it's definitely true for a lot of people that a more well rounded environment is better. But I'm not one of those people. I've realized this about myself--I'm much more comfortable and happy when I'm around people like me.
So SBF had a principled reason to bet big, even if there were high odds of failure. He also had a reason to think that the odds of failure were not that high—Alameda was printing money, and it was also clearly aware that some crypto exchanges operated suboptimally.
Continuously betting big, with nonzero odds of failure, will eventually lead to some kind of drawdown. Doing a full Kelly bet can lead to extreme drawdowns, even with perfect play; maximizing log(wealth) also means having a level of volatility that few people can reasonably tolerate. That's one reason that practitioners who use Kelly betting typically don't do full Kelly bets; half- or a third-Kelly is a common compromise. Another reason for this is that bets in financial markets, and bets on new startups, do not actually fit the framework for Kelly, since Kelly tells you how to bet given total certainty about the odds and the payoff. Since you can't be more than 100% right, but can be up to 100% wrong, so you ought to bet less than Kelly in conditions of uncertainty, i.e. all real-world situations.5
SBF argued that you shouldn't bet based on expected(log(wealth)), but based on expected(wealth), because the utility of money diminishes so slowly when it's being given away. (But maximizing log(wealth) is actually the way to maximize expected wealth growth over the long term). Going past Kelly doesn't mean trading more risk for more return, it means taking on so much risk that there's no possibility of a good long-term return!
There are three plausible ways this overbetting strategy can be rescued from the curse of blowup risk. The first is that there's an overlap between the Effective Altruist community and belief in the Many-Worlds interpretation of, which can be interpreted to mean that when probabilities don't work out, there's a world in which they do, and you can't privilege this world over that one just because you happen to be stuck in the unlucky version of reality. It's quantum suicide but for finance. Quantum suicide is mostly a thought experiment, but it can be enhanced by adding even more implausible details. Suppose an evil wizard offered to donate $1 billion to the charity of your choice, but only if you played a round of Russian Roulette. $1 billion will save lives even if it's spent fairly inefficiently, much less if it goes to high-ROI interventions. So long as we're maximizing utility, you're morally obligated to take the deal!
More realistically, a second option is that this kind of utilitarian risk-on behavior is happening in parallel. From the perspective of a smart utilitarian:
- They believe they are utilitarian because it's correct and they're smart enough to see that.
- Being smart creates opportunities for making money.
- So the utilitarian can be confident that there are other equally-smart people who are also taking moonshot bets. It's not one person betting on a 1% chance to 1,000x their wealth: it's thousands of people with the opportunity to take those bets! So the effective bankroll is much bigger; each of them should be betting more on the assumption that, of the people who don't blow up immediately, most of them systematically underbet.
So when SBF looks at a bet with a 1% chance of paying off, he shouldn’t think of this as a one-time proposition, because other EAs will find similar bets, and take them. If other people’s utility matches your own and enough other people are running the same moral source code you are, then the Law of Large Numbers applies and the expected value is exactly what the world will get from this decision framework, even if any single decisionmaker may find himself bankrupt, pilloried in the media, credibly worried about death threats, etc. In this model, full Kelly underbets because the bankroll is much bigger than it looks.
And a third possibility is that SBF was maximizing arithmetic returns rather than geometric returns because he saw the crypto boom as a one-time, and temporary, opportunity to do wholesale wealth transfer from speculators to worthy causes. If you don't care about compounding, because your bet won't compound, you should indeed maximize arithmetic returns rather than geometric returns.
Regardless of the exact justification, FTX/Alameda persistently bet big. And, at some point, they tapped the invisible line of credit ($), moved assets to where they shouldn't be, and eventually got caught.6
Unfortunately, this is not a parallel process in which lots of Effective Altruists are running uncorrelated bets at the same time: Sam’s blowup hurts other people running the same strategies or raising for the same causes. Expected(Log(wealth)) is hard to escape, because linear extrapolation to arbitrary values will always produce absurdities. Nature abhors an oversimplified model, though the way nature deals with it is only obvious in retrospect.
Continuously doubling down, and always taking the 51% chance to double your money over the 49% chance of losing everything, will eventually blow up. It isn't strictly fraudulent to do this, but it's a form of stochastic fraud: given enough rounds, it will lead to unexpectedly bad outcomes for everyone involved even if they're right about the odds and are never tempted to cheat.
Is This a Game to You?
I should clarify, too, that I don't exactly find league fun, or learn much from it, or really get anything out of it, exactly.
A review of SBF's interviews and blog posts shows that analogies from games are a frequent theme. Magic: The Gathering was an influence on some of his thinking about probability, this post on thinking about the multiplicative benefits of charitable donations to outcomes that depend on one another uses League of Legends to explain some of its points, and FTX celebrated a venture round by playing Bughouse chess.
Of all financial markets, crypto in its current state is the closest to a pure game. A crash in Solana isn't going to mean that factories don't get built or that new companies don't get funding; a spike in ethereum doesn't directly translate into real-world wealth creation, except in a very indirect way. Cryptoasset fundamentals are strongly tied to other crypto, and less to the rest of the economy.
And games can be a powerful source of intuitions that apply in other circumstances. The League of Legends analogy linked above is useful, and might be harder to see without a game as a reference. But game analogies also create some bad mental habits.
Playing a game optimally means treating everything as a resource, including bad things with positive tradeoffs. If a game gives characters bonuses and maluses, the question is not how to get every bonus and never get any malus, but how to get the best return on each. So a Magic player doesn’t try to end every game with the 20 life they started with, but to get the highest victory probability from each lost point of life. A chess player isn't hoarding their pieces, or even trying to ensure that their opponent can only take a piece by making an equivalent sacrifice, but trying to get the maximum advantage; losing a queen and getting a checkmate is a victory, and when the game ends the sacrifices don't matter.
How would this apply to real-world decisions? You might stop thinking of honesty as something inviolable, and start thinking of it as one more source of mana to be earned and burned. You might min-max in order to speedrun through, knowing there are some things like properly-labeled bank accounts that an optimal player can skip in order to gain some crucial speed.7
Games have a determinate world. Average performance matters, and when the game ends you can just play again. That condition doesn't hold in the real world.
So you kind of want to just be risk neutral. As an individual, to make a bet where it’s like, “I’m going to gamble my $10 billion and either get $20 billion or $0, with equal probability” would be madness. But from an altruistic point of view, it’s not so crazy.
—Rob Wiblin, 80,000 Hours interview with SBF
FTX and Alameda have blown up. There will doubtless be a long legal process figuring out exactly what happened and when. And we may not know when, exactly, it started. If I had to guess: Alameda CEO Caroline Ellison drastically reduced her tweeting frequency after this thread on how Effective Altruism compels a seemingly-absurd focus on specific outcomes, with her subsequent tweets being 1) a callout to a crypto and philanthropy discussion, 2) announcing the departure of Alameda's co-CEO, and 3) saying that Alameda's balance sheet was fine. As Paul Graham put it, "the main indication of impending doom is when we don't hear from you."
SBF was not just a financial and software entrepreneur. He was also a regulatory entrepreneur, who was the second-largest donor to the Democratic party and the sixth-largest political donor overall in the US this cycle. He had some Effective Altruist reasons for doing this, but also some pragmatic ones, like proposing standards for digital assets which industry participants believed was intended to harm truly distributed financial products to the benefit of centralized ones like FTX. We can expect a regulatory backlash (which will have an easier time hurting onshore and legally compliant crypto operators who are within the reach of the law—oh well!). We can also expect crypto lobbying to be met with extreme skepticism.
Effective Altruism has been hurt, but will likely continue. And they certainly don't endorse fraud, or even the stochastic fraud of repeatedly taking risks that will blow up. (Here are some relevant pre-commitments—all good utilitarians know that the way to maximize utils is to follow virtue ethics!)
But the collapse raises important questions about how to look at risk. Some theories that work well at small scale are catastrophic at larger ones, which is equally true of scaling a levered trading strategy in a small and illiquid market and of scaling positive expected-value decisions from millions of dollars to billions.
And there's risk in ideological homogeneity; it's a hedge if the risk-tolerant utilitarians are right, but a catastrophe if their math is off. There are many people running the exact same moral source code. If that code has some sort of flaw, a Heartbleed-for-bleeding-hearts, then the fact that it's widely used is a problem.8 Effective altruists have done interesting work on the question of existential risks, like pandemics, and one of their favorite topics is the risk of "unfriendly AI"—that powerful computers trying to do what we want them to do will end up destroying humanity. There's a similar risk that sufficiently powerful utility-maximizing entrepreneurs will also take risks that turn out badly and have negative externalities: we need to think about Unfriendly EA.
FTX also raises essential questions of trust. Centralized intermediaries have accelerated crypto adoption, but they've added a layer of untrue-to-the-spirit-of-crypto opacity and leverage to the system. The pace at which crypto evolves means that seemingly advantageous systems can get huge quickly, leading to correspondingly big blowups.
Given a truly one-time loss, that can be temporarily covered by shifting a small amount of customer liquidity from the exchange to the fund, the probability of total failure declines. But once fraud is a possibility, it becomes a habit, and once it’s a habit p(failure) approaches 1. There's path dependency in following unusual beliefs to their conclusions: a strictly logical, rules-based moral system can get someone into a situation where the calculations are overwhelmed by emotional reactions. Plenty of quants smash keyboards when something the model says will happen 1% of the time actually happens, even though this is known to be negative EV.
Thinking probabilistically is a powerful framework. But in the end you’re stuck with whatever specific event actually happens in the real world, with a probability of 1.
Further reading: The Diff has an earlier piece on FTX, and this piece on parallels between the crypto financial system and emerging markets that go through booms and busts. Good SBF interviews include this one with 80,000 hours, this one with Tyler Cowen, and this one with Dwarkesh Patel. And a big thank-you to @AutismCapital for providing excellent real-time coverage. If you were involved in FTX, or if you found the "Maximize EV even if you take blowup risk" argument appealing, this is pretty good on the psychology of maximizing total good rather than adhering to a strict moral code.
Disclosure: long a little BTC. (Unlevered, not in FTX.) Also in the interest of full disclosure, I have lots of Effective Altruist friends and have enjoyed sparring with them. They take ideas seriously and are taking the collapse of FTX seriously, too. I was telling utilitarians that Hayekian concerns about limited knowledge imply that they should be deontologists instead before it was cool.
A Word From Our Sponsors
Taming the wild west of data
Information is one of the most valued commodities when it comes to most businesses, but the way it’s stored isn’t standardized. Are dates month first or day first? Do zip codes have 5 digits or 9? The answers to these questions are never the same. Inconsistencies become a real problem when companies need to share and exchange their data. This can cost businesses lots of time, energy, and money.
Flatfile is the product that takes this complexity away and makes data exchange painless and easy. Learn more about how Flatfile can help with your data problems and book a demo today.
Paul Allen's art collection was auctioned off last week for a total of $1.6bn, and overall, the collection produced annualized returns of 6.2%, below what the S&P did over the same time. And that's omitting transaction costs and storage costs that shave off some total returns. Art sometimes gets talked about as an asset class, but it's an idiosyncratic one with some interesting forms of survivorship bias (returns will look a lot better if you track artists from when they start out obscure to when they get famous—unless you somehow track all the artists who start out obscure and stay that way forever). Part of the appeal of art is that it's an inflation hedge, part of the appeal is that it's a crisis hedge. And, for the rich but not superrich, one thing it does is hedge against higher income inequality in the .01%; art, high-end real estate, and philanthropy are basically the only forms of consumption that the super-rich can spend a material share of their marginal income on, so buying art as a rich person is partly a bet that someone else will get even richer and decide they'd like to have it.
Where Inflation Hits
Walmart is telling suppliers not to raise prices too much ($, WSJ). In one sense, this is a non-story; it's not as if Walmart would come out and say "if you try to jam a price increase through, we're gonna have to take it." But it's also a decent signal about the company's view of consumer sentiment right now. If they're telling suppliers that prices need to come down, they're also hinting that they expect the end customer to either buy less or be more willing to trade down.
The Best Job In Tech
The Apple/Google relationship is a fascinating one: Google can earn more from a given searcher than Bing can, so Google is the natural high bidder for any search placement.9 But Apple can decide who the default search engine is on iOS, and most users stick with defaults. So, every time the search engine deal comes up for renewal, both sides have to put forth very strong signals that they could take or leave the deal, before they both conclude that they should take it. And one element of this is that Apple will sometimes hire search engineers from Google, put them to work on search projects, and use the existence of these projects as a way to extract concessions from Google. Google just hired one such engineer back ($, The Information). Being a pawn in a big negotiation is not a high-agency role, but being a symbolic way for each side to negotiate the price of a ~$10bn/year or so deal probably makes it easy to name whatever price you want.
Sitting it Out
J.P. Morgan avoided the losses some other banks have faced from buyout loans ($, WSJ), in some cases by avoiding deals entirely and in other cases by advising on deals instead of financing them. The theoretical benefit of being a big bank with many lines of business is that there's always useful information percolating through the organization. Loan officers can provide market color on how aggressively other banks are lending, credit cards can track consumer spending in real time, and trading desks can see financial risks that can reverberate through the real economy. And another advantage is that a big balance sheet isn't burning as much of a hole in the company's pocket if there are alternative investments to be made at different levels of risk: a bank that doesn't do much in the way of mortgage lending or traditional bank loans might be tempted to chase bad deals because otherwise their assets aren't getting any kind of return, but JPM can afford to let some groups earn subpar returns for a while in order to wait for better opportunities later.
The natural gas business is moving from a regional, pipeline-bound business to a global business in which cargoes can be shipped around the world ($, Economist), a process that has been going on for a long time (a gas liquefaction process was patented in 1915), but which has accelerated in the last decade due to fracking and in the last nine months due to Russia. Natural gas has historically been volatile, and the countries that import it most eagerly have government backstops protecting importers with fixed-value contracts against price drops. As it gets more global, prices may stabilize—but they may, like oil, be more tied to global geopolitics, too.
Companies in the Diff network are actively seeking talent! If you're interested in exploring growth opportunities at unique companies, please reach out. Some top current roles:
- A company building ML-powered tools to accelerate developer productivity is looking for a mathematician. (Washington DC area)
- A fintech startup that basically lets investors trade any theme as if there were an ETF for it is looking for a senior backend engineer. (NYC)
- A company that helps emerging tech businesses identify and hire top talent—and that gets paid in equity—is looking for new team members. A great way to build a startup portfolio and a great network without an in at a VC firm. (US, multiple locations)
- A company that's building a new hyperlocal bricks-and-clicks business seeks a market launcher who can close deals with small businesses and handle rapid scaling. (If you've ever wished you could have launched markets for Airbnb or Uber/Lyft in the early days, this is your shot...)
- A crypto data company is looking for experienced DevOps people who can build systems to quickly ingest and analyze on-chain data. (US, remote)
If you’re looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.
This even extends to how moneylike cash itself is. You can use money to prevent yourself from getting mugged by moving to a safer neighborhood, but the cash you’re carrying is worth $0 to at times when you’re most acutely interested in trading money for safety, i.e. when you’re in the middle of getting mugged. ↩
There’s probably more alpha right now in founders who wear well-tailored suits. Either they’re hardcore contrarians or they’re sufficiently out of the startup mainstream that the round won’t be competitive. If they're not memorizing Paul Graham, they're probably not doing YC. ↩
This is an especially important point because many ideologies lead to bad leaders because the process of getting a new system in place selects so strongly for risk-tolerance and sociopathy. The skills necessary to win an election or execute a coup correlate only loosely with the skills needed to run a country afterwards, and if either of those paths to power are competitive, there may be a range restriction effect where the people who select into seeking rather than wielding power are actively bad at the latter. ↩
There are precedents for someone as young as SBF to be in charge of something as big as FTX. At that age, Napoleon was the commander of France's Army of Italy and was running a successful campaign against Austria; Alexander the Great had conquered Turkey, Egypt, and Persia. Like SBF, Napoleon got his training in a legacy system before switching sides; like SBF, Alexander benefited from having parents with relatively high socioeconomic status. ↩
Even if you're doing full Kelly betting while playing poker, you can blow up, because there's some unknown chance your opponent is cheating, and the optimal cheater will play against someone whose bankroll has been built by optimal non-cheating play and who thus has a lot of winnings to take and a lot of confidence that their play is optimal. ↩
Perversely given the stated risk tolerance, it's worth pointing out that willingness to commit fraud radically expands the bankroll. Now, the most the company can lose on its bet is not 100% of its equity but 100% of equity plus accessible client funds. Early on, this could be done in a modest way—even converting clients' USDT into USD and investing the proceeds in short-term treasury bonds would generate some carry. But if that isn't enough to get the hedge fund back to breakeven—and if the fund's status as emergency liquidity provider means it's taking lots of toxic flow from other FTX traders when stablecoins blow up—then that's not enough and the bets get bigger. ↩
To be clear, I don't buy this explanation, both because of the alleged accounting backdoors mentioned earlier and because of the sheer implausibility. Alameda says their initial edge was operations! Arbitraging price gaps in Japan and Korea means being very good at opening and keeping track of different bank accounts. Additionally, the amounts in question seem to be too large for an innocent mistake. If you find several billion dollars under the couch cushions, you don't assume that you made a few billion and forgot about it. ↩
In some ways it’s hard to generalize about utilitarians, because thinking from first principles creates such high variance. For example, there’s widespread disagreement on whether the optimal number of human beings is much higher than we currently have or zero. But they are more likely than the average person to believe that ~8 billion is the wrong number. So we can't always generalize about exactly what utilitarians will do to make money, or what they'll do with their money. What we can say is that if most rigorous risk managers maximize log(wealth) but utilitarians view the utility of wealth as linear, whatever behaviors they engage in will be more extreme than the average person's. ↩
Bing still gets some search traffic from being the default for Microsoft's built-in browser, and for being the default fallback when people try to search for something with the start menu and it doesn't exist on their machine. This is actually very valuable to Google, since it demonstrates that they're not a monopoly. In fact, it would probably be worth it for Google to pay Microsoft to keep Bing alive rather than redirecting searches to Google, taking their US market share from 87% to 94%. ↩