The Racehorse Who Gambled
Expected Value, Discounted Future Cash Flows, and Stealing from the Future (a Rough Draft)
Boy, Substack’s LLM image generator will NOT generate a horse with Sam Bankman-Fried’s head, no matter how I try. Still, interesting results.
This week’s draft excerpt from Stealing the Future is extremely drafty - it is, in fact, more of a sketch, based on my recent reading of Liliana Doganova’s excellent and exciting book Discounting the Future: The Ascendancy of a Political Technology. The book both analyzes and recounts the intellectual history of “discounting” as an accounting and finance practice.
Discounting, particularly given its role in venture investing, is obviously central to the FTX story functionally. It also has a clear relationship to the view of the future embedded in Effective Altruism. But exactly tracing those connections - and more important, contradictions - is complex. So what you’ll find below is an ‘essay’ in the most conventional sense - a series of thought experiments and half-starts towards some conclusion that is never quite reached.
The below will be go through many more iterations before it makes it into the forthcoming book (Repeater Books, October 2025, baby). That’s why these previews are reserved for my small circle of paying supporters - if you want to read more like this, consider joining us.
Expected Value, Discounted Future Cash Flows, and Stealing from the Future
Sam Bankman-Fried and his cohort exist in a strange, dual and divided relationship to the future.
On the one hand, most overtly, Effective Altruists profess belief that the future has equal weight to the present, in their core argument that future human lives have equal weight to present human lives. While the outward emphasis is on an ethical equality between present and future humans, this formulation in practical terms inevitably implies and implements a functional equivalency. This equivalency invites - in fact, demands - an equalization of investment spending between that focused on the present and that focused on the future, without accounting (in several senses) for either the inevitable downside risk of failure, or the pure existential fact that the future is not where we live.
But while Effective Altruism as an ethos suggests that we throw away the logic of discounted future cash flows, many of the individual members of the Effective Altruist and broader Rationalist and Longtermist movements, and the overwhelming majority of the major funders that have backed them, are the fruit of venture capital. Venture capital's function is to imagine a future where a present technology, business line, or individual company is worth much more in financial terms than it is now. Venture seeks to buy that future revenue in the present - but at a heavily discounted price, adjusted above all for the risk that this bountiful future doesn’t come to pass.
This "discounting" of the present price of a future thriving business is a widespread, in fact nearly universal form of financial analysis. It is used to decide whether stocks, bonds, or even cryptocurrencies are undervalued or overvalued, relative to their "discounted future cash flow." In a more conventional transaction like a corporate bond sale, discounting is used to set the interest rate collected by lenders, based on a combination of the riskiness of a company’s debt and the time horizon for paying back the loan. A number called the “discount rate” is used to reduce those future dollars in a consistent way and arrive at the “net present value” (NPV) of future cash flows. The discount rate reflects a company’s risk profile, and its potential upside, relative to wider economic conditions.
In venture capital, discounting and the calculation of Net Present Value becomes much more of a vibes-based exercise. The underlying data isn’t public, but more to the point, there’s usually no real data at all, because most venture-funded companies are still at the stage of investing in turning an idea into a business, rather than generating revenues from a fully-functional business. A venture-funded company is initially just a series of ideas to be attempted, with estimates of the possible revenue if things work out.
There are attempts to nail down this logic through metrics like "total addressable market," a tally of all the money being spent in a company's field, which establishes a maximum for estimating its eventual value. But venture capitalists are very open about the reality - sayings like "you invest in a founder, not an idea" are widespread, making clear that valuation is generally about intangibles like personality, relationships, credentials, and sheer talent as much as metrics.
Largely because it's so much based on intangibles, discounting and valuation in venture capital is a material locus of power struggle between the founder of a company and its venture investors - and if things go well, later, the investing public joins the fray. Venture investors want to put their money in an idea when its valuation is as low as possible, allowing them to buy a larger percentage of company equity. Founders, at this stage, instead want a high total valuation, because they want investors to pay more, in real cash, for a smaller percentage of the company.
If a venture-backed company “goes public” with a stock offering, those investors switch sides: they want the valuation to be as high as possible, so they can get the maximum return on those early investments. The public, including journalists and analysts who at least notionally represent them, generally want neither low nor high, but an accurate picture of revenues and costs, because those numbers have a relatively direct impact on the price of a stock over time, and so weigh heavily on the decision to buy an IPO or an existing stock, and at what price.
One deep problem with this system arises because venture capital markets are private, meaning that startups don’t have to publicly disclose their actual costs and revenues. This creates a very clear motivation for founders to create as much froth and hype around their company as possible.
This became epidemic during the "cheap money" era of 2010-2020, when interest rates hovered around zero percent and capitalists were looking for any even hypothetically profitable way to deploy their money. The most notorious example is WeWork founder Adam Neumann, who was far more concerned with valuation and funding rounds than actually building a business, leading to WeWork quite literally falling apart as soon as public markets and reporters started paying attention to whether it actually made any money.
Adam Neumann was also notable because of the huge amount of real cash he was able to extract from WeWork in the course of the various funding rounds. Through a mix of WeWork stock sales folded into fundraising rounds and loans backed by his (wildly inflated) WeWork equity, Neumann became a real-money billionaire. This was the dividend of Neumann's ability to evoke a vision of the future that people wanted to own, not any actual ability to make it come true - after its failed attempt at an IPO, WeWork's target valuation of $47 billion plummeted, with Wile E. Coyote swiftness, to less than $10 billion.
When the company finally went public, it did so not through an IPO, but through a less-transparent (and therefore itself more vibes-based) process known as a SPAC, or reverse merger. These became briefly popular during the early COVID pandemic after a few savvy operators, most notably Chamath Pahaliyapatiya, discovered them. Almost all SPACs issued during this period, including most of Pahaliyapatiya's deals, have lost immense value, and U.S. regulators have since closed the loophole that made them appealing. After going public via SPAC in April 2021, WeWork lasted a bit over two years before declaring bankruptcy in November of 2023, $4 billion in debt and with a business model rendered nearly unworkable in the post-COVID era of work-from-home. Adam Neumann, though, remained a billionaire - he had, entirely legally, monetized a future that then never manifested.
On Discounting
Discounting as a practice of placing valuations is the underpinning for loans and investment premised on future growth - money that can sometimes take quite a long time to “return”, with interest, to its original lenders. It is borrowing against the future through the mechanisms of investment and lending. Inflated earnings multiples - that is, the ratio between the open-market price of a financial instrument, and the actual projected future earnings underlying the instrument, amount to a license to borrow MORE against the future.
This inherently creates a bias towards founders with great narratives, for no reason more complicated than their ability to directly charm investors, who want to become part of their story. The most notorious instance of this is likely the head-trip WeWork founder and kaballah surfer Adam Neuman supposedly put on SoftBank head Masayoshi Son. Sam had a much different, weirder type of glamour, but it was just as tailor made for bringing the future nearer, and putting it into his pocket.
The clearest way this manifested was in a single number known as a startup’s “Valuation” - the notional value assigned to it by expectations of its future potential earnings. These valuations are arrived at in private markets among venture capital funds, investment bankers, and a small number of other players. For the most part, these valuations aren’t particularly real money - even if you’re an insider or early investor who owns shares, you likely can’t sell them. So the degree to which neutral market forces set these numbers is debatable - and sometimes, as in the case of WeWork, the bias and error built into a startup’s expectations of its future earnings lead to public scrutiny that completely annihilates the narrative insiders have been reciting to themselves.
These blowups have certainly become more dramatic, if not necessarily more frequent, over the past decade. And they have overwhelmingly taken the outer form of a “tech company,” in two incredible cases as a misdirect that fueled their scam. Theranos CEO Elizabeth Holmes publicly mimicked the habits of Steve Jobs in pursuit of her grift, and Adam Neuman sold WeWork to the market as a tech company, when it was really just another subleasing middleman.
This and countless other tech-adjecent scams tap into the assumption that technology-based businesses “scale” incredibly well, with the important result that small investments can produce huge returns. This association between high speculative valuations and software startups is in part grounded in reality - successful software really does allow smaller teams to make a lot of money. But it's worth noting that there's ideological connectivity here too: software attracts both founders and investors more likely to understand the world in a mechanistic way, perhaps increasing faith in the accuracy of valuations based on projected future cash flows.
This is at least as fundamental to the market for cryptocurrencies themselves, in a way that's worth detailing. Contrary to common understanding, many cryptocurrencies have clear revenue sources that could in theory be used to perform some equivalent of a discounted cash flow calculation to arrive at a reasonably grounded "valuation" for the asset or system. However, the intensely speculative nature of crypto markets makes this close to useless - imagine if every company on the New York Stock Exchange was Tesla, its stock value juiced by hype and founder charisma to many, many times anything that could be reasonably connected to its future cash flows.
This, too, became a specific weak point in Sam Bankman-Fried's thinking about finance, projection, and the future. The speculative premium is huge arguably because cryptocurrency as an entire category is so novel that it still hasn't had much contact with reality. But as first and foremost an asset trader, and with little interest in history or social context, Bankman-Fried saw the appreciating value of all crypto assets from 2020-2022 as a continuing underlying trend. In fact, the runup had nothing to do with crypto, but amounted to an economy-wide bubble in financial assets and internet activities (including day trading) directly juiced by COVID lockdowns and pandemic assistance.
He was far from alone in this mistaken thinking. Notoriously, an investor named Su Zhu, of a fund called 3 Arrows Capital, argued that cryptocurrency had entered a "supercycle," and its prices would continue rising indefinitely. Of course, that didn't happen, and Su Zhu got blown up and washed out nearly as badly as Bankman-Fried did. Zhu stayed out of jail, in part because he fled to the non-extradition territory of Dubai. But he was widely reputed to be in hock to Chinese gangsters, having, much like Sam, gone extremely long on his own inevitable good fortune.
The Horse Who Thought He Was a Gambler
One way of understanding the root of Sam Bankman-Fried's failure is that, unlike Adam Neumann, he took the logic of discounted future cash flows literally.
He behaved as if the equity valuation of his company was simply a true and accurate and inevitable index of what would happen, instead of a complicated product of his own ability to evoke one possible future and convince investors to back it. This gave him a logical, ethical basis - at least according to his consequentialist, utilitarian view - for freely borrowing customer money. While a private startup valuation is a projection of future earnings reflecting the sentiment of a very small circle of investors, Sam treated it as directly fungible with present real dollars. Along with similarly sincere gaps in his understanding of basic ethics, this technical misunderstanding of finance was also a key component that led him to commit fraud.
His literal understanding of DCF-based valuation also, though, kept him from ever feeling the urge to "cash out" on anything like the gargantuan scale of Adam Neumann. Though smaller moves like a notorious $10 million gift to his parents had a similar contour, Sam wasn't savvy enough to protect those outflows with any sort of legal rationale, and many have been long since clawed back.
The valuation of FTX - a number based on hypothetical future earnings, remember - was reflected in equity venture investment rounds and in the price of FTX's equity-like FTT token. Personal payoffs aside, a similar literalization of projected future value played into Bankman-Fried's use of FTT tokens as collateral in his business dealings, marked to a valuation based on an extremely low share of tokens "floated" in the public market.
Another way to put this is that Sam Bankman-Fried applied Effective Altruist thinking, which in its longtermism and its implicit certainty about prediction is in some respects the antithesis of DCF thinking, back to finance. That is, he equated future outcomes 1-to-1 with the present estimates reflected in the valuation of FTX, just as effective altruists equate future humans 1-to-1 with presently living humans. This thinking represents an inversion, or perhaps simply perversion, of the impact that Doganova finds discounting has on the locus of knowability and risk. In turn, by taking his startup valuation literally, Bankman-Fried created a kind of black hole of infinitely nested self-speculation.
Wrapped in this logic, Bankman-Fried saw himself not just as a person building towards the future, but a living present embodiment of that future. His inevitable success meant he was certain to be able to correct, in the future, any transgressions undertaken in the present. Every compromise was strictly temporary. This same logic, wielded by smarter and more disciplined people, has underwritten many transfers of wealth executed without crossing any legal red lines. This group includes Adam Neumann, who is noone's idea of smart or disciplined, but still managed to extract more than a billion dollars in cash from a company that wound up being nearly worthless, without going to jail.
Without the financialization of the future that structures investment capital, Doganova says, the risk of a new venture, such as the creation of a new drug, is unknowability as a lack of knowledge. That is, uncertainty of success. After financialization, such as venture investment, this unknowability of outcomes is transferred to the investor, while the 'valuation' assigned to the company implies certainty of success. (Neumann fell victim to this insofar as he mistook WeWork's valuation as a direct measure of his effectiveness, and paying more attention to it than operations.) The uncertainty of success is still captured in this new relationship, though, through a risk premium essentially baked in to venture allocations and valuations. Rather than being tasked with actually accomplishing something novel, an investor is tasked with balancing the risk of many such projects in a portfolio, and simply waiting for some of them to fail and others to succeed.
But Bankman-Fried, like Adam Neuman before him, made the error of thinking of risk from the viewpoint of an investor, while he was himself in the position of productive agent - a company leader and founder. This gave him seeming carte blanche to borrow against that future - and, in the case of his customer's money, to steal against the firm faith that he would be able to pay it all back. He lived his life as the embodiment of the net present value of all his own future earnings - but in his head, the calculation didn't include a discount for risk or alternatives. It was all available right here, right now, leading Bankman-Fried to the tragically misleading sense that FTX had "infinite money" - a misjudgment now enshrined in the title of Michael Lewis' book on SBF.
Bankman-Fried had been raised, like the horses on Leland Stanford's industrial breeding farm, and like all children of Palo Alto, as an agent of production, victory, and value. But in reality, Bankman-Fried, like Stanford's horses, was just one in a stable of hundreds, maybe thousands. He was the recipient of bets, and those betting on him by and large were only committed to the possibility of his success. The discount they enjoyed against the possible future value of FTX was driven in part by the real possibility of failure.
Like any competitor in a race, the best performers must believe in their own inevitability - but they also have to run the race. But Bankman-Fried himself did not think in those terms - as a productive agent, backed by capital to achieve a promised business goal. Instead, he regarded himself as the winner before the race even started, and began betting on that outcome using huge leverage. In short, he was a horse who came to think of himself as a bettor. But if every horse thinks that way, and none of them actually choose to run the race, the entire system is at risk of collapse.
What Discounting is Really For
The practice of calculating future earnings according to a “discount” based on their distance into the future, was the product of relatively recent evolution in accounting and finance, forged through a thoroughly modern professional discourse recently retold by Liliana Doganova. It is a fundamental building block of contemporary investing and finance - but it seems to fail a great deal. Doganova suggests that in light of this the goal shouldn’t be to refine the practice of discounting, such as by tweaking the numbers and ratios, but to acknowledge that at this point it clearly “works” - but works towards what goal, and for whom?
Doganova begins analyzing this work by approaching discounting as a way of managing uncertainty. Writing in the context of the pharmaceutical industry, she describes investment based on discounted future tax flows as a way to transform the uncertainty inherent to any research or development project, what she calls the uncertainty of a lack of knowledge, into uncertainty as the investor’s concern - that is, functionally, the risk an investor is taking of losing all of their money, and the interest rate of return they expect to be paid for putting their stake up. In pure accounting terms, this means it behooves any tech founder or other capitalist to disguise risk, hype up big ideas about cosmically huge future product reach and impact, and generally distract from the unproductive contemplation of technological, scientific, or industrial reality.
Sam Bankman-Fried, like Adam Neuman, understood this well, on multiple levels. He knew that valuation was about narrative, aesthetics, and scope. When he floated the bananas notion that FTX would someday buy Goldman Sachs,[ https://cointelegraph.com/news/billionaire-sbf-says-ftx-may-one-day-buy-goldman-sachs-and-cme] it was nothing but upside - even if rational market actors dismissed it, if even a few suckers let it influence their thinking to place tiny FTX in the same conversation as Goldman, it inched the valuation up a few ticks. It was dinner and a show, with leverage funded by his customers’ deposits.
"
This equivalency invites - in fact, demands - an equalization of investment spending between that focused on the present and that focused on the future, without accounting (in several senses) for either the inevitable downside risk of failure, or the pure existential fact that the future is not where we live.
"
It doesn't though. A supposition that future life and current life are equivalent AND that one can predict perfectly the future, might require equal investment in both, but then would immediately remove the "downside risk of failure." A supposition that future life and current life are equivalent AND that no one can predict the future even partially, would require zero investment in the future, as the downside risk is basically infinite. You're basically a bug at this point, anything you do to help your future, anything that works, is not an investment, but merely the result of serendipitous genetic mutations that passed on to offspring. Something in the middle regarding future and present valuation and predictability is sensible and what everyone currently does to varying levels of success.
I believe so much of life is down to cognitive dissonance and mental time travel. Mental time travel is being able to make internal models of the universe that include the past (remembering experiences and inferences based on information), the present (experiencing and gathering information), and the future (inferences based on past and present). But just as you are unlikely to remember the color of your pants last Tuesday, you are unlikely to predict your pants color next Tuesday, or, more importantly have seven models of the future for every color of pants you have, or for that matter seven times the models of future you have for all your types of shirts, or for that matter have these models properly weighted by clean vs dirty clothes and ease of access from your closet. Obviously, your mental models of the future just aren't going to include the color of your pants.
Something has got to give. You need to prune your models. Some models might include you hurting a lot of people. Like getting caught "investing" their funds though they never asked you to, and even worse, not being able to give them back immediately. That's not a very happy model. It means you are a bad person. You don't want to be a bad person (cognitive dissonance). And thus, snip, snip, there go those strands of Fate from your self made tapestry of the future.
That is the real danger.
"
One deep problem with this system arises because venture capital markets are private, meaning that startups don’t have to publicly disclose their actual costs and revenues. This creates a very clear motivation for founders to create as much froth and hype around their company as possible.
"
Being poor means less options. Startup investment is only for accredited US investors. Once IPO, these companies are usually already picked clean of any enormous return opportunities. Crypto is even less information than startups, but also less hurdles. No one wants less information, but desperation for a chance to be there "at the beginning" has people drawn to it. Poor people in US can't invest in (much less start) startups. People in other countries without much investment opportunities or whose job prospects are so low that "steady work" goes nowhere, see it as a giant risk, but still worth it.
Level the playing field so not only the rich have access to certain investment opportunities, and then crypto won't be (or at least seen so much as) a cesspool of memecoins.
"
He behaved as if the equity valuation of his company was simply a true and accurate and inevitable index of what would happen, instead of a complicated product of his own ability to evoke one possible future and convince investors to back it."
Maybe all those times CZ referenced "four" was regarding Biggie's Crack Commandments. SBF was getting high on his own hopium supply.
"
In turn, by taking his startup valuation literally, Bankman-Fried created a kind of black hole of infinitely nested self-speculation.
"
Particularly nice turn of phrase. A death spiral from hopium hodling over mercenary liquidity. Which is extremely ironic as he mercenary liquidity farmed like mad during the first DeFi Summer. Ponzi for thee, but not for me, it would seem.
"
But Bankman-Fried, like Adam Neuman before him, made the error of thinking of risk from the viewpoint of an investor, while he was himself in the position of productive agent - a company leader and founder. This gave him seeming carte blanche to borrow against that future - and, in the case of his customer's money, to steal against the firm faith that he would be able to pay it all back.
"
This assumes a definition where a founder is one who acts on behalf of his/her company, while an investor is cold, mercenary capital. And while more accurate, not absolute.
"Seeming" is doing a lot of work in that sentence. A good investor knows to ride the wins and cut the losses. FTX was winning, and yet he made investments that undermined its ability to win. It would be just as unethical if he stole from his clients and then spent it directly on moves that improved just this "winner," but it would be more sensible than putting it into unrelated investments (Anthropic, which was dumb luck as AI is probably going to crash too, just later than he would) or indirect, supplementary investments (celebrity and stadium promotions, which were just dumb, as DZM pointed out pre-collapse, particularly with the stadium, and I agreed).
For example, here's a scam I just thought of that would work for quite a while until it didn't. Have a centralized exchange that had "staking" programs in what you call "AI trading bots." the "trading bots" returns are higher for those investors who are newer and more active, versus those who are older and less active. The latter being less likely to withdraw. You do it by having opaque instruments that can't be unlocked right away, in this case AI trading bots, but it could be anything. It's still stealing and wrong, but you'd be much less likely to be caught and all the money is still there as long as you time the unlocking correctly. If any investor becomes more active, bump the return (of which you get a cut), less active, bring it back down. As they seem less likely to withdraw, increase exposure (more swings with activity), as they seem more likely to withdraw, reduce exposure. Determining on how likely they withdraw can be done by withdrawal size in relation to position size over history. If position just seems to go up, increase exposure to active = more returns.
Can you business patent a scam? Maybe with the new election it'll be both legal and encouraged. Consider this written proof for my IP claim. The ones currently running similar scams likely don't have it patented yet or have put it in public writing.
"
Sam Bankman-Fried, like Adam Neuman, understood this well, on multiple levels. He knew that valuation was about narrative, aesthetics, and scope. When he floated the bananas notion that FTX would someday buy Goldman Sachs,[ https://cointelegraph.com/news/billionaire-sbf-says-ftx-may-one-day-buy-goldman-sachs-and-cme] it was nothing but upside - even if rational market actors dismissed it, if even a few suckers let it influence their thinking to place tiny FTX in the same conversation as Goldman, it inched the valuation up a few ticks.
"
It's interesting, but it does kind of make sense that /this/ kind of bananas works, but stadiums-and-Katy-Perry bananas do not. It's ridiculous but when SBF made an outrageous claim about financial technology, before the collapse, well he was seemingly very successful at fintech, and since no one knew the real reason he was "successful" was boring old theft, it didn't feel like complete BS. When he started with celebrities and stadiums you start thinking, wait a minute, what does this have to do what he appears to be successful at? Does this guy really understand everything as well as he appears to? It just felt off at the time. It would feel just as off if there was an OpenAI Stadium. Some people are using AI, sure, but it's pretty niche and the visible effects aren't really there yet. Today, crypto still doesn't have much visible effects and still looks weird on a stadium.
What stinks is we're seeing this right now with Elon Musk, he makes bananas claims all the time with what he can and will do. Pretty sure he's going to say he'll come up with a better method for photosynthesis and try to replace all the plants in the world. I don't doubt SBF would probably end up like Musk after enough time had he not been stopped.
Thing is, Musk seemed to accomplish (by proxy) the highest ambition that SBF gets laughed about now: US presidency. If Musk had failed, maybe we would be laughing at his ambitions too, but he didn't fail. Someday, maybe we will learn how this recent election really ended up the way it did. So many thinking everyone is racist and I just think it's extremely easy with enough money to manipulate the online reality of people not privileged enough to cultivate critical thinking in college and regardless can't afford to spend time understanding complicated systems. They work 10 hours, doom scroll, and believe what money is trying to sell/tell you. Once you learn more about how SBF "succeeded," the less impressive it all seems what he did. I don't think we're going to hear the truth about how Musk "succeeded" in getting Trump elected. I just think in another reality there are people laughing at him for being such a tool aligning with Trump. They are both seen as very uncool losers. And there's a hopeful future toward a stable and functioning government.
Real talk: I'm worried. I think about this every day.