Previously: Sadly, FTX. Audio version of this post is here. I doubted whether it would be a good use of time to read Michael Lewis’s new book Going Infinite about Sam Bankman-Fried (hereafter SBF or Sam). What would I learn that I did not already know? Was Michael Lewis so far in the tank of SBF that the book was filled with nonsense and not to be trusted?
> The idea was simple, and also very much something I would have tried. There were a lot of different exchanges trading a lot of cryptos at lots of different prices. Sometimes the prices were different. When that happened, you could do various forms of arbitrage (and statistical arbitrage). Sam, being a trader and also being Sam, was a big fan of taking the free money.
This bot was completely insane to run in 2017. There was barely any way to hedge Bitcoin then, let alone *all shitcoins*.
I think you probably would have been talked out of it in 5 minutes.
Your story is "because that's who he is as a person", but I'm kind of still skeptical that there are people like that.
Some other people were speculating that it was because he was on such high doses of stimulants that his judgment was borked. That would make sense, and we do have suggestive evidence from the trial reporting that he's on an unusual amount of Adderall. (Entry-level ADHD dosing is not usually several times a day.)
Patrick McKenzie's story is, as I understand, "because he was a bagman for money launderers; the missing money wasn't lost in bad trades but funneled back to his secret crime clients." That too might make sense. Normally I'd be suspicious about a theory that posits a crime for which there's no direct evidence, but Patio's not a wild-speculation guy in general, and money laundering in crypto isn't an unusual occurrence.
But "he just likes to gamble, lots and lots, to an irrational degree?"
I...guess?
Is it plausible that anybody likes to gamble *this* much?
It's such a red flag to start a startup and have the founder retain all of the stock. Either you only hire inexperienced people, or you unethically mislead them. Tricks like, the offer letter only describes the number of shares, and it turns out theres 1000x more shares outstanding than they verbally told you. Either way is setting your company up for disaster.
One lesson here that I wish more people would take is, don't join a startup where the founder is being secretive about how much stock you're getting. Every time a friend of mine has been in this situation, a year or two down the road they discover that 1. they were only given a tiny amount of stock, not worth anything near as much as the founders hinted, and 2. the general quality of people at the company is really low.
Obviously in a tech startup the stock might be worth nothing. But you should know what % of the company you own (with some flex for things like size of option pool, amount of future dilution, etc).
For ModelBot, I wonder if the trade was some kind of market manipulation (e.g. pump and dump or similar) rather than arbitrage. That would explain the difficulty of backtesting it and also of running it with a smaller stake.
> In elementary school he’d read the Harry Potter books over and over. By the eighth grade he had stopped reading books altogether. “You start to associate it with a negative feeling, and you stop liking it,” he said. (505)
I found this amusing due to doing a similar thing myself. Harry Potter grabbed me as an early understandable long-form book with lots of content to chew through. Then I reread it multiple times over the next years.
Picking out books to read became harder and harder. I did end up associating going to the school library with a negative feeling, of having to use cover art and limited statements to discern what a book might be about and whether it would be any good. I was picky, and in reality none of my tastes were being properly appealed to (Harry Potter was just familiar and dense enough along some dimensions). I read other fiction, but at times I just gave up and picked something I knew I liked, and once you reread the first Harry Potter book it is far easier to continue onto the next. That would then make the next week or two of reading during class free-time a breeze, though a predictable one.
Being a complete layman financially, my guess for the reasoning to not activate ModelBot is that it would have potentially leaked the details of how the bot worked to competitors, based on transaction histories. If it barely works, someone has the opportunity to refine it before you (you have one 100x programmer, they have 101 programmers) or simply copy you. No chance to test it in public without a loss, even at small volumes. Otherwise I agree, it makes no sense not to set expense limits on the bot and calll it a day.
Minor nitpick: do you *actually* believe that electron suffering is something EAs care about or was that a tongue-in-cheek remark to prove your point? Because there is a forum post on it, but it is an April Fools joke.
Thank you for writing this. More than anything else recent (recent enough you can send to a normie and tell them to read it and they don't got "but it's from 2 years ago!") it's a fantastic argument in favor of "maybe people should actually do some thinking and thinking about thinking".
I would add, to the unanswered questions section: How is it impossible for someone so good at games/puzzles to be in Bronze in League of Legends? This is like hearing someone is a competitive-at-world-champs level chess player, but really struggles with tic-tac-toe and checkers. There's probably some good SBF mentality questions to mine here.
I was also, to some degree, friends with Sam when I lived in NYC. At least on the level of occasionally talking. We also talked about him leaving Jane street. I wish I wrote down what exactly he said. I don't remember him telling me he was going to run or help run anything. He just told me he was going to take a job. This seemed kind of dumb but 'take a job at an official EA org regardless of other opportunities' seemed like the kind of thing EAs do all the time. I just figured Sam's EA job quickly didn't work out for some reason. I'm still not sure if he was blatantly lying or the EA job just fell through. But I don't have any notes. Definitely feels like a bit mysterious to me.
It's more gonzo than gonzo. I knew the outlines of the story already, but the various cognitive train wrecks kept me glued to the entire review. And not just SBF. The entire cast is extremely smart in some ways yet makes dumb decisions that could be trivially avoided, throughout. And these people are all rationalist-adjacent at least. "The first principle is not to fool yourself – and you are the easiest person to fool."
I saw a trial quote that the thing where they bribed chinese officials wasn't in the charges. Why wasn't it? Seems like a blatant crime for which there's a lot of evidence, no?
Regardless of whether you want to call it a 'hack' the MOBL exploit is at minimum commodities/securities fraud. Its roughly the same exploit that landed Avraham (another rationalist) in jail and will almost certainly cause him to serve prison time. This sort of exploit is absolutely not ok even if some people's intuitions differ.
Appreciate this. What a fantastic review! I also read it and agree that it is an awesome book.
But I never believed Lewis was earnest. He was SBF's guest and Lewis surely knows that the crimes he described alone are enough to convict SBF. Emphasizing this in is unnecessary and imo turns the book into something less interesting -- a polemic on SBF. By writing the book in this way he:
1. Ensures future subjects will be happy to let him tag along and speak candidly with him
2. Keeps the focus on the big picture that you end with (although he never specifically states it kind of like the Alchemist)
3. Produces what's probably the first widely-read description of how powerful EA is becoming.
I have to wonder if Lewis is touting SBF's innocence in the hopes of selling more copies of the book to people wondering what the proof of his innocence is.
> The idea was simple, and also very much something I would have tried. There were a lot of different exchanges trading a lot of cryptos at lots of different prices. Sometimes the prices were different. When that happened, you could do various forms of arbitrage (and statistical arbitrage). Sam, being a trader and also being Sam, was a big fan of taking the free money.
This bot was completely insane to run in 2017. There was barely any way to hedge Bitcoin then, let alone *all shitcoins*.
I think you probably would have been talked out of it in 5 minutes.
I am still confused.
Why was SBF betting so aggressively?
Your story is "because that's who he is as a person", but I'm kind of still skeptical that there are people like that.
Some other people were speculating that it was because he was on such high doses of stimulants that his judgment was borked. That would make sense, and we do have suggestive evidence from the trial reporting that he's on an unusual amount of Adderall. (Entry-level ADHD dosing is not usually several times a day.)
Patrick McKenzie's story is, as I understand, "because he was a bagman for money launderers; the missing money wasn't lost in bad trades but funneled back to his secret crime clients." That too might make sense. Normally I'd be suspicious about a theory that posits a crime for which there's no direct evidence, but Patio's not a wild-speculation guy in general, and money laundering in crypto isn't an unusual occurrence.
But "he just likes to gamble, lots and lots, to an irrational degree?"
I...guess?
Is it plausible that anybody likes to gamble *this* much?
It's such a red flag to start a startup and have the founder retain all of the stock. Either you only hire inexperienced people, or you unethically mislead them. Tricks like, the offer letter only describes the number of shares, and it turns out theres 1000x more shares outstanding than they verbally told you. Either way is setting your company up for disaster.
One lesson here that I wish more people would take is, don't join a startup where the founder is being secretive about how much stock you're getting. Every time a friend of mine has been in this situation, a year or two down the road they discover that 1. they were only given a tiny amount of stock, not worth anything near as much as the founders hinted, and 2. the general quality of people at the company is really low.
Obviously in a tech startup the stock might be worth nothing. But you should know what % of the company you own (with some flex for things like size of option pool, amount of future dilution, etc).
For ModelBot, I wonder if the trade was some kind of market manipulation (e.g. pump and dump or similar) rather than arbitrage. That would explain the difficulty of backtesting it and also of running it with a smaller stake.
> In elementary school he’d read the Harry Potter books over and over. By the eighth grade he had stopped reading books altogether. “You start to associate it with a negative feeling, and you stop liking it,” he said. (505)
I found this amusing due to doing a similar thing myself. Harry Potter grabbed me as an early understandable long-form book with lots of content to chew through. Then I reread it multiple times over the next years.
Picking out books to read became harder and harder. I did end up associating going to the school library with a negative feeling, of having to use cover art and limited statements to discern what a book might be about and whether it would be any good. I was picky, and in reality none of my tastes were being properly appealed to (Harry Potter was just familiar and dense enough along some dimensions). I read other fiction, but at times I just gave up and picked something I knew I liked, and once you reread the first Harry Potter book it is far easier to continue onto the next. That would then make the next week or two of reading during class free-time a breeze, though a predictable one.
Being a complete layman financially, my guess for the reasoning to not activate ModelBot is that it would have potentially leaked the details of how the bot worked to competitors, based on transaction histories. If it barely works, someone has the opportunity to refine it before you (you have one 100x programmer, they have 101 programmers) or simply copy you. No chance to test it in public without a loss, even at small volumes. Otherwise I agree, it makes no sense not to set expense limits on the bot and calll it a day.
Have you read Number Go Up? If so, how does it compare?
Minor nitpick: do you *actually* believe that electron suffering is something EAs care about or was that a tongue-in-cheek remark to prove your point? Because there is a forum post on it, but it is an April Fools joke.
Thank you for writing this. More than anything else recent (recent enough you can send to a normie and tell them to read it and they don't got "but it's from 2 years ago!") it's a fantastic argument in favor of "maybe people should actually do some thinking and thinking about thinking".
I would add, to the unanswered questions section: How is it impossible for someone so good at games/puzzles to be in Bronze in League of Legends? This is like hearing someone is a competitive-at-world-champs level chess player, but really struggles with tic-tac-toe and checkers. There's probably some good SBF mentality questions to mine here.
I was also, to some degree, friends with Sam when I lived in NYC. At least on the level of occasionally talking. We also talked about him leaving Jane street. I wish I wrote down what exactly he said. I don't remember him telling me he was going to run or help run anything. He just told me he was going to take a job. This seemed kind of dumb but 'take a job at an official EA org regardless of other opportunities' seemed like the kind of thing EAs do all the time. I just figured Sam's EA job quickly didn't work out for some reason. I'm still not sure if he was blatantly lying or the EA job just fell through. But I don't have any notes. Definitely feels like a bit mysterious to me.
It's more gonzo than gonzo. I knew the outlines of the story already, but the various cognitive train wrecks kept me glued to the entire review. And not just SBF. The entire cast is extremely smart in some ways yet makes dumb decisions that could be trivially avoided, throughout. And these people are all rationalist-adjacent at least. "The first principle is not to fool yourself – and you are the easiest person to fool."
I saw a trial quote that the thing where they bribed chinese officials wasn't in the charges. Why wasn't it? Seems like a blatant crime for which there's a lot of evidence, no?
> At this time, they were borrowing at 50% (!) interest in order to trade
here’s a Patrick McKenzie thread explaining why this statement from Lewis is most likely wrong
https://x.com/patio11/status/1709271113962852747
Regardless of whether you want to call it a 'hack' the MOBL exploit is at minimum commodities/securities fraud. Its roughly the same exploit that landed Avraham (another rationalist) in jail and will almost certainly cause him to serve prison time. This sort of exploit is absolutely not ok even if some people's intuitions differ.
Appreciate this. What a fantastic review! I also read it and agree that it is an awesome book.
But I never believed Lewis was earnest. He was SBF's guest and Lewis surely knows that the crimes he described alone are enough to convict SBF. Emphasizing this in is unnecessary and imo turns the book into something less interesting -- a polemic on SBF. By writing the book in this way he:
1. Ensures future subjects will be happy to let him tag along and speak candidly with him
2. Keeps the focus on the big picture that you end with (although he never specifically states it kind of like the Alchemist)
3. Produces what's probably the first widely-read description of how powerful EA is becoming.
I have to wonder if Lewis is touting SBF's innocence in the hopes of selling more copies of the book to people wondering what the proof of his innocence is.