Book Review: On the Edge: The Business

Previously: The Fundamentals, The Gamblers

Having previously handled the literal gamblers, we are ready to move on to those who Do Business using Riverian principles.

Or at least while claiming to use Riverian principles, since Silicon Valley doesn’t fit into the schema as cleanly as many other groups. That’s where we begin this section, starting at the highest possible conceptual level.

Time to talk real money.

Why Can You Do This Trade?

First law of trading: For you to buy, someone must sell. Or for you to sell, someone must buy. And there can’t be someone else doing the trade before you did it.

Why did they do that, and why did no one else take the trade first? Until you understand why you are able to do this trade, you should be highly suspicious.

“Every single thing we do, I can point to, like, ‘here is the person who was doing a thing wrong,’ ” said Hall. “We build a map of all of the players in the world…who are trading for reasons other than they want to generate some alpha.[*15] You know, somebody got added to the S&P 500. So now all the S&P 500 ETFs out there have to buy this company. Or it’s a CEO of a startup, he’s now a billionaire. He has ninety-nine percent of his net worth in the company. He’s allowed to sell his shares on this day [and] he’s probably going to sell a bunch.” (4330)

One could argue that we used to live in a Moneyball-vulnerable world. Gambling lines were terrible, baseball teams didn’t go after players who get on base, traders didn’t even know Black-Sholes and there was treasure everywhere, this theory says. But then it says, now Everybody Knows all that. There’s tons of competition. You can’t get ahead purely with statistics, or at least the basic versions, anymore?

“There was [once] some place where statistical knowledge gave us an insight vis-à-vis the people who didn’t have access to it,” [Peter Thiel] said, mentioning innovations like the development of life insurance and the Black-Scholes model for pricing stock options. “But in the world of 2016, or 2022, if we are too focused on statistical or mathematical knowledge, we end up missing out on a lot.” (4541)

The quants won, in other words. We’re living in their world. (4551)

Were that it was true. A lot of financial prices and certain types of decisions much better reflect the quant prices, but so many other decisions very much do not, including many of our most important ones. The result was not technocracy.

It does mean that if you want to win by being a quant and doing quant-style things to find Alpha in various markets, you have to work steadily harder over time.

In a World of Venture Capital

I definitely believe The River is a thing.

I definitely agree that it includes the outright gamblers and casinos, the stock traders and the rationalists and Effective Altruists.

The weird case is Silicon Valley and Venture Capital.

The basic activity of an investor is to ask, how often will this business succeed? What would happen if it did? What would cause it to succeed or fail?

Cuban told me that Shark Tank is more like real investor meetings than you’d think; in early stage investing, you’re trying to filter through pitches quickly, and first impressions count for a lot. Cuban’s best heuristic is to look at the company from the entrepreneur’s perspective. “I tend to have a good feel for what a business needs to do to be successful. And so I can sit there and listen to their pitch, put myself in their shoes as if it was my company, and ask the [difficult] questions that I would need to deal with,” he said. (4051)

And that’s all very River, lots strategic empathy, much Bayes. Indeed, this can be a highly discriminating and useful process.

Thus such folks like to talk as if they have the River Nature. They definitely want to award the River Nature high status. And, pitchmen that they are, they sold Nate on this pitch.

When I first proposed this book, venture capital was slated to play more of a supporting role, paired off with hedge funds in a chapter that was mostly about Wall Street. Instead, it turned out to be the protagonist. (4461)

I did not see the book that way. I see Nate Silver as the protagonist. And if I had to choose a group I’d still go with the poker players, in a ‘poker players go into [X]’ or ‘people in [X] think like poker players’ kind of way.

But what about venture capital? Do they actually have it?

I’m not so sure. There are a lot of signs something is not right there.

Remember that list of what ‘successful risk-takers’ do and are, according to Nate Silver? In brief the items are:

1. Cool under pressure.

2. Insanely competitive.

3. Strategic empathy.

4. Take shots.

5. Raise or fold.

6. Prepared.

7. Selective high attention to detail.

8. Adaptable.

9. Good estimators.

10. Stand out, not fit in. Independence of mind.

11. Conscientiously contrarian.

12. Not driven by money.

As we’ll see, I think they talk a better game on many of these than they play.

I think that SV/VC, and especially VC, sort of turned this into a set of slogans and heuristics, to which everyone involved then tries to spout and conform to. That they systematized the whole process in the name of taking risk and being contrarian. Without that frame, VC is deeply conformist and risk averse, and even uses a weird form of risk aversion to justify its risks, based on FOMO (fear of missing out) on the big hit investment being the ‘real risk.’

And yet, for all their success, almost every VC I spoke with still lived with FOMO: fear of missing out. Specifically, fear of the founder who got away. “The mistakes of omission are much, much bigger mistakes,” said Andreessen. (4668)

How does venture capital justify taking risks? In large part through literally viewing the greatest risk as not taking (the wrong) one.

I had versions of these mantras repeated to me so often that I believe they’re the governing gravitational forces behind pretty much everything that takes place in the Valley—why it attracts so many strange and disagreeable people, for instance. Venture capital is a unique enterprise for two essential reasons: (4617)

It has a very long time horizon. (4619)

It offers asymmetric odds that reward taking chances on upside risks. (4622)

Are these principles the rest of the world should emulate? The first one is. (4625)

Both are good principles. Supposed contrarians all reciting the same mantras, however useful, should still make you suspicious. Yes, it is good that they have relatively long time horizons and seek big upside potential. And yes, even some of that is huge, even if this is neither as unique nor as pure as they would have you believe, or is often done via a sort of play acting - the symbolic version of the thing.

The result is something hugely (modulo AI existential risk concerns) useful and pro-social and productive, a wonder of the world, something I absolutely love that it exists, that the world needs more of. That is because People Don’t Do Things, especially new things, or take risk, or get the talent pointed at real problems, and all the neat stuff like that.

Nate’s theory of how this works out is a risk story. Books like this are hedgehogs.

So here’s my theory of the secret to Silicon Valley’s success. It marries risk-tolerant VCs like Moritz with risk-ignorant founders like Musk: a perfect pairing of foxes and hedgehogs. The founders may take risks that are in some sense irrational, not because the payoff isn’t there but because of diminishing marginal returns. (4750)

There is certainly something to that. The VCs are making a deal with the founders, where the founder commits to acting irrational in this way, and the way they do that is by credibly claiming they are indeed irrational in one of several ways that would do this - they value the mission enough, are sufficiently deluded about their chances, actually care about making the billions, and so on, or some combination.

Whereas I have a slightly different story. To me: The core activity at the end, of driven and talented people actually trying to Do Things, especially related to tech, is insanely valuable.

Even though everyone is kind of going through motions and playing Keynesian beauty concepts and social games and signing and imitating and so on most of the time, it doesn’t matter.

The process identifies founders and teams capable of Doing Things, and puts them in position where they get to and have to try and Do Things. It works. Even if 90% of the time and money and activity is to get to that point, and insiders capture much of the direct financial gains, the actual payoff makes up for it.

The San Francisco Bay Area—the counties of Alameda, Contra Costa, Marin, Napa, San Mateo, Santa Clara, Solano, Sonoma, and San Francisco, California—was home to 7.76 million people as of the 2020 U.S. Census, or roughly 0.1 percent of the world’s population.[*8] However, as of October 2023, it is headquarters to almost 25 percent of the world’s unicorn companies (defined as private companies with a valuation of at least $1 billion). (4573)

The pie is that big. We still all win.

Yes, they accomplish this in part by using rhetoric and multi-level signaling games and funding preferences to forcibly steal talent and companies from elsewhere, but it all still largely counts.

So three cheers, to all the smoke and mirrors. If that’s what it takes.

(As long as, of course, we ensure the world probably doesn’t end.)

Short Termism Hypothesis

A thing venture capital likes to say is that they are long term oriented, while the stock market is short term oriented.

Furthermore, the value of maintaining a long view may increase in a world where—and here I’m sounding like Thiel—data and analytics can sometimes lead to nearsightedness. It’s often easy to algorithmically optimize for a short-term fix—and much harder to know what will produce long-term brand value. (4632)

The catch is that both of them are:

1. Long term oriented, in the sense of trying to calculate the net present value (NPV) of future cash flows, and make good trades.

2. Short term oriented, in the sense that they extrapolate quickly from short term results to implications for those cash flows, such as expecting growth to beget future growth, the classic ‘hockey stick graph,’ and so on.

3. Partly Keynesian beauty contests. Everyone is trying to predict what the next trader is going to do. Except on Wall Street you can essentially hold to maturity and let the company earn the money if others disagree (mostly), whereas in Venture Capital, unless you are prepared to cover future rounds, the company’s success depends on additional fundraising.

There are frequently claims Wall Street is too invested in the short term. I think this is at least imprecise. What they are doing is not so much ‘neglect the long term’ as it is falling victim to Goodhart’s Law, and focusing too much on the numbers and what they can see, and not enough on intangibles - yes, like long-term brand value creation. Once you have good models, it’s hard to disregard them sufficiently or adjust them enough to account for what in context are intangibles.

Non-Determinism and its Discontents

Peter Thiel is one of the most interesting people I’ve ever met. There are downsides to be sure, we disagree on quite a lot, but I’d be thrilled to spend more time with him even if I had no agenda and no hope of convincing him to much change his mind.

If you ask someone like that a question, you can never be sure what you’ll get.

When I spoke with Thiel, this was the first question I asked him. “If you simulated the world a thousand times, Peter, how often would you end up in a position roughly like the one you’re in today?” This was intended as something between a softball and a curveball—an unexpected but relatively nonthreatening conversation starter.

Thiel delivered an answer that went on for almost thirty minutes. He began by objecting to my premise. “If the world is deterministic, you’ll end up in the same place every single time. And if it’s not deterministic, you would almost never end in the same place,” he said. (4527)

The question is indeed somewhat ill-formed. Which things are held constant, versus not held constant? When do we begin to roll the dice? If you begin one minute before conception, in an important sense a person probably won’t exist, or will be very different. If you randomize other events, who knows what happens. And so on.

In the common sense hypothetical of this type, where world events happen on schedule unless you in particular alter them but the personal stuff is randomized, I think the answer is he ends up in a similar position remarkably often.

The details will change a lot, and certainly there is a group of people who are proto-Thiels that greatly outnumber actual Thiels, but I don’t think the proto-group is that many zeroes bigger. Not every proto-Thiel becomes a Thiel, but a remarkable percentage do - I think such a person ‘creates their own luck,’ perseveres, ends up in the right places at the right times and finds a way. Not every time, not even a majority of the time, but way more often than you would think. And if you’re not a proto-Thiel? Then it almost never will happen.

That is exactly why the principle in the Valley is, you bet on the founding team.

The Founder, the Fox and the Hedgehog

As a reminder, in this metaphor, foxes know something about many things, and hedgehogs know a lot about one particular thing.

Those clever little foxes were the heroes of The Signal and the Noise. But do foxes make for good founders? I can think of exceptions to the rule (the polymathic Collison is quite fox-like, for instance).

But in general, VCs are looking for people with one big crazy idea that they’ll commit to for a decade or longer. (4693)

I think this is actually backwards. Silicon Valley looks for founders willing to commit to one big crazy idea for a decade.

That one big crazy idea is ‘found a venture capital backed Silicon Valley startup.’

They bet on the man (and yes it is usually a man), not the cards.

Everything along the way, until rather far along, is mostly evaluating the team.

If your first idea does not work, you shift it around. Or you fully pivot. That’s fine.

Indeed, there is a reason why Y-Combinator does not even require an idea for a company. The idea to found a company at all is the idea that matters most. Ideas are cheap without execution. Execution is everything. That’s how all this works.

To be a CEO, you need to be a fox of the first rate. You need to be able to raise capital, to lead and manage, to hire and fire, to negotiate, to determine strategy, to dealing with mistakes and crises one after another, to market and understand the customer, to balance signaling to investors with signaling to customers with signaling internally with building and shipping a product. Your realm is anything and everything. You have to give up any semblance of a life for years on end.

And you have to stay sane (enough) throughout, as most people are lying to you about how the world works or your business works in various ways, or have no idea about either while claiming otherwise, while you are driven crazy exactly by whichever element here is your weakest link.

Founder CEOs who are playing the game (relatively) straight up earn every penny.

The Team to Beat

What makes a good founder? What makes Silicon Valley eager to bet on you?

Nate Silver asks an important part of this question.

Are VCs Intentionally Selecting for Crazy Asshole Founders? (4918)

Yes.

I mean, sure, #NotAllVCs, and there are limits to how far they want you to go.

But yes, they absolutely will favor you if you are an asshole, because they think that other VCs will too, and because they think being an asshole is vital to success - you’re going to have to act like one on various occasions to Get Things Done, and you’ll need to credibly signal that you can act like one as a threat.

And yes, they absolutely want you to be the right kind and amount of crazy, because they are looking for you to do things that you would not otherwise be wise to do, but which help them and increase chances of success, including having absolutely no work-life balance whatsoever for years on end, and also again because they know the next level up will be looking for that crazy.

A lot of this is pure pattern matching. They’ve seen crazy assholes succeed, they know others have seen it too, so they want to fund more crazy assholes.

There’s more to it. And there are plenty of other considerations too, being a crazy asshole is only a small part of what you need to do. But there is no mystery here.

What are other signs and traits of a potentially successful founder you want to bet on?

Another thing you want is that they are self-made. Coming from money won’t work, or at least that’s conventional wisdom. You wouldn’t be hungry enough. Indeed, America today is highly unusual for how many of our richest people did not inherent.

Of the billionaires on the 2023 Forbes 400 list—the four hundred richest people in the United States—70 percent are basically[*19] self-made. And 59 percent came from an upper-middle-class background or below. (4919)

In 1982, only 40 percent of the Forbes 400 had started their own business; the majority were simply scions of inherited wealth. (4924)

Let’s say you inherit a $25 million trust fund on your eighteenth birthday. Are you going to start a business with it? Maybe you should. But it’s much easier to withdraw $1 million a year to live off, travel the world and have some wild parties, and put the rest in S&P 500 index funds earning 7 percent a year. (4929)

Even if you felt you could count on the world to ‘stay normal’ and the S&P to return 7% a year forever (5% after your peaceful 2% inflation target), what fun would that be?

A lot of people would answer quite a lot of fun. And, well, fair. But it’s not for everyone, at least not for all that long. Purpose is important.

Perhaps the problem is that the very rich are not brought up to see it that way, and that they don’t develop various skills that are vital to the path to founder success.

“Why are second-generation kids never that successful?” asked Social Capital CEO Palihapitiya, who moved with his family from Sri Lanka to Canada and worked at a Burger King to help support them. (4941)

It can also help to have something else: a chip on your shoulder. Josh Wolfe, of Lux Capital, is fond of the phrase “chips on shoulders put chips in pockets.” Feeling left out, excluded, or estranged can make you extremely competitive. Remember, VCs want founders who are willing to commit to low-probability ideas—ideas they think the rest of the world is wrong about—for a decade or more.

What motivates a person to do something like that? Wolfe, who grew up in a single-parent home in New York’s gritty Coney Island neighborhood, told me he thinks there’s a common answer: revenge. (4946)

You’ll show them by becoming a billionaire? Well, maybe. At least sometimes. Perhaps I simply didn’t want revenge enough. I like to think that’s not the main motivation, that you can want to do something big for the world, something you care about, or want to make a lot of money for various reasons.

“Who is my real customer? It’s a young, disenfranchised, disenchanted entrepreneur-to-be,” said Palihapitiya, referring to the sorts of people who might come to him for investment or mentorship.

“And I use those words specifically, because if you’re comfortable, and you’re happy,” he continued, “you’re not the kind of person I want to work with anyway because you’re probably not gonna be successful.”

This feels like a dangerous game. Successful founders may be disagreeable on average, because disagreeability is correlated with competitiveness and independent-mindedness. But the disagreeability is still a bug, not a feature.

If you start to select founders because they’re disagreeable, you may get the wrong ones. Especially if founders deliberately play into stereotypes that they think VCs will like, as Sam Bankman-Fried did (we’ll cover him in the next chapter). (5010)

Any time you are selecting for anything, you are selecting for the people able and willing to fake it. That’s always a problem. But perhaps also an opportunity?

A lot of the reason Silicon Valley works is that everything is a test, preparing you for future tests, and eventually, one hopes, the creation of actual value.

If SBF or someone else can play into the stereotypes that VCs like, the naive reaction would be that this is terrible. You’re falling for a fake. Except, what if what you are testing for is the combination of:

1. Figure out what the teacher’s (here VC’s) password is.

2. Be able and willing to give the teacher their password back.

If you can do that, then those are highly valuable skills. You can figure out what people want to see and hear, and you can provide that for them, and you’re not going to have any objections to doing that. Sounds like someone to invest in, no?

Yes, part of the problem with this is you end up with a lot of frauds. But, as Matt Levine and others remind us, the optimal amount of fraud is not zero. You want a founder who is not too worried about whether they might technically be committing fraud, or about they can cash the proverial checks they are metaphorically writing. You want someone who does what they have to do, not someone scrupulously honest.

Also, if you fake being someone, you start to act like them naturally, and you do things like them to keep up the facade. Almost as good.

Occasionally that all means someone steals all the customer deposits along with your investment capital. That is a known risk. But so what? Small price to pay.

What about the traits themselves?

It makes sense that, on average, you don’t want a happy entrepreneur, because they won’t be willing to put in the insane hours and effort and sacrifice, or take the risks in various senses, potentially including both moral and legal risks.

Silicon Valley Versus Risk

Do not get me wrong. Silicon Valley founders take metric tons of risk.

The catch is, they mostly do it by ensuring no one tells them the odds.

Elon Musk is an outlier, but less of one than you might think.

Thiel had once considered writing a book about Musk and their PayPal days. “The working title I had for it was Risky Business, like the eighties movie. And then the chapter on Elon,” he said, “was ‘The Man Who Knew Nothing About Risk.’ ” And yet, when it comes to risk, it’s Thiel who is more of the outlier in Silicon Valley. “Peter is not a risk-taker. There is nothing. He is a guy wired to protect his downside,” said Moritz.

Although that’s nothing compared to some others:

For instance, SBF was quite specifically insistent that people ought to be willing to risk having their lives end in ruin. There’s an idea, he said when I spoke with him in January 2022, that “the bigger you are, the more risk you can take without endangering what you have.” (6024)

I do think that even if you could have otherwise played it safe and been personally set, you need to be willing to risk ruin if the stakes are high enough, in the sense of being willing to make enemies, or perhaps face political or legal consequences or even potentially violence. But risking financial ruin in particular, through losing money, past some point? You’re simply never getting odds to do that.

Silicon Valley tells founders several things at once.

1. You will probably fail.

2. You have to believe you will succeed, or we won’t fund you.

3. You have to swear to everyone you will succeed, or we won’t fund you.

4. When you fail, if you do it the right way, we won’t hold it against you.

Venture capitalists do not want founders who understand how to take carefully calculated risks and make precise Bayesian estimates.

Venture capitalists want founders who will risk it all to try and hit is big, whether or not that makes any financial sense for the founder, or the odds are any good. Or whether or not it is even +EV for the company, because VCs think in terms of those huge wins.

They very much want you thinking like Musk here rather than Thiel:

But there’s also something more. The most successful founders like Musk succeeded despite what were ostensibly extremely long odds. Thiel, recalling the challenges Musk overcame to build SpaceX, thought that no one obstacle Musk faced made for an insurmountable barrier. But the quants had run the numbers—if you have 10 hurdles to leap over, and you have a 50 percent chance of tripping over each one, the odds of making it to the end of the course are 1 over 210, or just one chance in 1,024—and concluded that the venture was imprudent.

Musk had thought differently. “He was determined to make them happen,” Thiel said. “It was a matter of assembling the pieces and putting them together, and then it would work. We’re just in this strange world where no one does it because they all think probabilistically.” (4560)

That’s… not what I think of quants running numbers, but yes, if you think there are a bunch of uncorrelated medium-difficulty hard steps, all of which must succeed, the odds are very much against you.

(Similar arguments are often used against pretty much anything ever happening. For example, the classic anti-existential-risk-from-AI argument that demands you lay out the details of one particular scenario with all of its steps, then says ‘this only happens if all 10 steps happen, and each is not that likely’ or arguing against one particular step, ignoring that the argument does not rely on those particular steps or path. And the reason something like SpaceX still works is in part that they fail and they figure out how to fix or work around the problem, repeat as needed.)

Of course, Thiel’s calculation was obviously wrong. People who found companies like Tesla (also see PayPal) are almost never going to do something where the outside view is actually 1024:1 odds against success. No, Elon’s odds probably didn’t make any sense either, but any reasonable reference class wouldn’t have counted Elon out. Notice that other space startups like Blue Origin might not be doing as well as SpaceX, but the odds certainly aren’t 1% or less.

VC also wants you not to worry about the downside, and not to be tempted to try to take some gains off the table, no matter how insane it would be not to do that.

It shows.

I’ve played in a handful of high-stakes poker games against rich guys—including one game frequented by venture capitalists, Silicon Valley founders, and the hosts of the All-In podcast, who are friendly with Musk. I was sworn to secrecy about the particulars, although since one of the All-In hosts, Jason Calacanis, said this publicly, I can confirm that the first time I played in the game, I won enough money to buy a Tesla.

The other thing I can say—just in general terms—is that the higher the stakes, the crazier the action. The biggest poker games select for the players who want to take crazy, irrational, −EV risks.

Maybe not literally going all-in every hand—although I’ve seen strategies that aren’t far from it—but embracing variance, as the “Techno-Optimist Manifesto” would say. (4505)

Embedded deep in the VC psychology is the idea that they can sleep soundly at night because they are not only enriching themselves, but also making the world a better place. For some recent technological developments, however, the value proposition has been more questionable.

Social media may well have had net-negative effects on society. Crypto gave rise to a lot of scams and cons, like Sam Bankman-Fried’s FTX—heavily invested in by Sequoia and other VC firms—which cost cryptocurrency holders out of at least $10 billion. And with AI, the disruption could be profound, resulting in mass reshuffling of the economy even if it remains relatively well aligned with human values. (4656)

What a nice term for a degen making -EV risks, ‘embracing variance.’ They’re not embracing variance. They’re gambling, and not the advantage betting, trying to win variety either. The bet it all on black kind.

So is this my way of saying that the richest founders in the world are just degenerate gamblers who got lucky? No, I’m not saying that. I think they’re highly skilled degenerate gamblers who got lucky. (4996)

Almost by definition, the people at the very top of any leaderboard are both lucky and good. (5002)

Indeed, that is all a fair way to describe the ‘Techo-Optimist Manifesto.’ Rather than argue that benefits exceed costs and the risks are worth taking, the manifesto is a series of unqualified assertions of blind faith and absolutes. Technology is always good, humanity always benefits, there is zero danger, and so on. It is its own strawman, a proud announcement of being blind to downside risks.

Which is exactly the venture capitalist way. When you invest in a company, when you try out a new tech, what could go wrong? Losing all your money is, in their view, only a small mistake. So that investment went to zero, so what. It can’t go below zero. It shows you were bold. The real mistake would be missing out on the next big hit.

The VC mind, by default, assumes that the downside must be limited. That mind lives in a world without short selling and with limited liability. It simply cannot fathom the idea that their investment could have a negative left tail even bigger than the positive right tail. Marc Andreessen (the author of the manifesto) seems to toy with outright denying the idea that any such investment could individually harmful on net at all.

The second trait—the asymmetric nature of payoffs in Silicon Valley—is even more essential to understanding its mindset. But its implications are more ambiguous than the first one. (4647)

Thus, Marc Andreessen says just do go ahead, whatever it is, however risky in any sense, including everything from AI catastrophic risks to being another crypto fraud. His politics and stands on regulations, and his justifications for them, not only ignore but deny and laugh at tail risk. Simultaneously it argues that this time will be the same as our recent ahistorically calm history so there is nothing to worry about, and also he forgets his history. And he is not alone.

That leads you to do things like this, after Adam Neumann famously built up WeWork only to have it collapse:

It was [Adam] Neumann, as you’ve probably guessed. In a room full of Silicon Valley’s elites, a16z were showing him off—and sending a message. “Why would they run out and give a bunch of money to Adam Neumann after everything they’ve seen? Like, what in the world was that all about?” said Benchmark’s Gurley. “If I were asked to analyze what they were doing, they wanted to send a signal to everyone.”

The signal was that they didn’t care about reliability—they wanted founders who gave them upside risk. They were embracing variance. “If they’re that type of person, they’re open for business, the door’s wide open, and we’re willing to talk to you, no matter what.”

The signal mattered. A16z seems determined to shout from the rooftops that they are willing to be as reckless as possible, in all senses, and to pour good money both after bad and also the extremely risky, and to not worry about cost or downside, as a business profile.

Mostly, however, I think this was simply a good bet. They want to be in business with Adam Neumann because they absolutely have odds betting on Adam Neumann. That was a highly impressive ‘failure.’ Adam Neumann won, in that he proved he could build up something big, get people to believe in a crazy business plan, and was smart enough to take a bunch off the table before it all fell apart. You don’t get that far without real skills, and there was a substantial chance WeWork could have succeeded. It does not take that big a probability of success before this bet was +EV.

I don’t know the terms of the investment, so it is certainly possible they overpaid and didn’t have odds. What I do know is that the world wants Adam Neumann to go out and try to Do Things again, to build another empire. That there is a price where Neumann should be happy to do it, and investors should be happy to help him.

The Keynesian Beauty Contest

I’ve already gestured at it a few times, but the core way Silicon Valley makes funding decisions is a Keynesian beauty contest, counting on it being grounded enough to get to a good equilibrium.

The process is akin to what economists call a Keynesian beauty contest. (5107)

In particular, it is an iterated such contest. You are trying to predict what others will do both now and into the future. If you select for that, you win. If not, you lose.

You could get around this by being willing to fund the entire company, at prices that enable the business to prosper, without needing the social proof of other investors. Unfortunately, the VCs of Silicon Valley, even when they can very much afford the required funding, are psychologically or socially incapable of this. Hence the contest.

Is Silicon Valley Really as Contrarian as It Claims? (5083)

This looks contrarian because the others are the other VCs in Silicon Valley, and especially the larger ones that will be required for later rounds. The question is, contrary to who? What does it all mean?

If you are contrarian (in the good sense) because you are thinking for yourself, then that is indeed good. If you are instead contrarian because you are thinking what those around you are thinking, and trying to ponder what they ponder, then you are a conformist all the same.

Do VCs actually want to be contrarian?

Founders Fund—Thiel’s firm—has a reputation for it. But according to Rabois, there’s a delicate balance to strike. “When you initially invest, or found a company, you want it to be, like, ridiculous and contrarian,” he said. “But you want it to shift into consensus, because you need other people’s money, publicly or privately.

You need to recruit employees that are more normal than the founder-type class.

So the art is pulling a trigger and then shifting it. And if that gap is too long, then you have a problem.” (5097)

You need there to be some element in there that differentiates the founding team and its ideas, allowing them to potentially win big. You get, and must spend, a few weirdness points.

In most other ways you actually need the team to be checking off the same boxes as everyone else around them. And indeed, it is vital that you be pondering what those around you are pondering, whether or not anything involved makes sense.

So Rabois is always trying to calibrate to his friends’ preferences. “One thing that’s kind of a secret in the industry,” he told me, “is that most of the people I compete with, I’m actually pretty somewhere between real friends and very good friends with. So part of what I’m doing is mapping their brain. Like, will they or their fund appreciate this? Are they going to see the signals I’m seeing?” (5114)

Sebastian Mallaby, the author of an excellent book about Silicon Valley called The Power Law, thinks that in certain respects it is an exceptionally conformist place. “In some ways, venture capitalists are the ultimate herders,” he said. “You go to Sand Hill Road, and you see that they all have offices on the same road. And there’s kind of one good restaurant on that road, at the Rosewood hotel, so they all bump into each other at the same bar.

He maps their brains, they map his. Everyone is trying to get in exactly the right amount of sync. You do want to find overlooked opportunities, but what makes them overlooked opportunities is that they won’t stay overlooked for long.

Yet for all its Mean Girls conformity—Rabois trying to figure out what his five best frenemies will think—Silicon Valley’s opinions are still relatively uncorrelated with those of the outside world, Mallaby thinks. (5125)

I do believe that on many important margins. Of course there is overall a huge correlation with the outside world, but there are many key differences, and the VCs map much closer to each other than to the outside consensus.

I even heard the argument that as the Village becomes more conformist, Silicon Valley has more opportunity to profit by running in a different direction. (5131)

It potentially does grant an opportunity. You can pick off those who don’t want to conform to the Village, either because they are non-conformist in general or because they don’t want to conform to the Village in particular. Or you can take the opportunity to move your SV consensus somewhere else.

One part a Keynesian Beauty Contest is that by default it will be absolutely brutal with enforcement of stereotypes. Convincing others you have what it takes is, itself, what it takes. If I think everyone else thinks you don’t have what it takes, then that means you indeed do not have what it takes.

Which means that correlation is king. Any visible sign that anti-correlates with success is going to haunt you every step of the way, whether or not it is causal.

If that sounds like a recipe for de facto racism and sexism, whether or not anyone involved is actually either of those things? Well, you’re right.

The investor, like many others, turned her down. “It’s amazing, I love it, blah blah blah, but I’m not going to invest,” she recalls him saying. “Here’s why. As a Black woman you’re going to have a harder time fundraising, a harder time retaining talent, a harder time selling,” the investor said. “Every part of this process is going to be harder for you, and it’s already an impossible process.” (5148)

The Secret of Their Success is Deal Flow

What makes you a top successful VC?

It helps to already be one.

Is VC Success a Self-Fulfilling Prophecy? Andreessen Horowitz is a lot like Harvard. And Founders Fund is a lot like Stanford. Marc Andreessen and Peter Thiel would probably resist the comparison, between Andreessen’s dislike for the Village and Thiel’s skepticism of postsecondary education.[*27] But the similarities are obvious once you see them. The top venture capital firms—like the top universities—are exceptionally sticky institutions. (5179)

What links top VC firms and top universities is that they are recruitment-driven businesses. And recruitment can become a renewable resource. (5194)

The secret is deal flow.

The best deals in Silicon Valley are obvious.

This is in some sense inevitable in a Keynesian Beauty Contest. If you can be confident everyone else is going to like the deal, then everyone else can figure this out too, and already likes the deal.

The trick is that the price of the deal does not reflect expected future value, or the place where supply and demand would currently intersect. You might think it would do this, but you would be wrong.

Instead, VC in SV has set up a system whereby the best founders are warned not to use their market power to drive up the price, and also where borderline founders mostly don’t get to offer sufficiently large discounts to make deals happen.

A given startup is ‘supposed to’ trade at a given price, based on what stage of development they are in, certain ‘fundamentals’ and various heuristics. And yes, there is room for ‘how high does the founder dare go,’ but there are severe limits to that.

The key weapon is fear of the ‘down round.’ If a startup trades now for a valuation of X$, then in the future trades for Y$, you had better hope that Y>X, or at minimum Y=X, and they will use various tricks (‘structuring’) to try and make the headline number not go down if at all possible. A down round is very close to death. So startups want to price, they are told, sufficiently low that they won’t run into this problem.

This is of course economically nonsense. If I have the New Hotness, I am very likely to get to the next funding round in a place where I have a company forth funding. In some cases, I will succeed wildly. If today’s round is priced fairly, that means there will be plenty of times the company does okay, and is still a good investment, but the expected value of the company has gone down. There is nothing wrong with that. Yet the entire ethos sets up to basically kill you if that scenario happens. In particular, if the lead investor uses this as a reason to pull out, that too gets you into deep trouble.

The VC solution, which is strangely good for their business model, is that you should price such that if the business remains viable the price will definitely go up.

This is of course rather silly. Either those middle-scenario viable scenarios are worth worrying about, in which case a little decline in EV shouldn’t stop anyone from taking a +EV wager going forward. Or they are very little of the value, in which case founders should not care that things go badly in those cases. Even if you buy the whole argument, you can’t have it both ways.

The other weapon the VCs use to persuade is that if the startup raises Too Much Money (TMM) then they will expand too quickly, take on too much burn, pay too high salaries because they lose negotiating leverage, and so on. Having done that too early, they will fail. The solution of course is to Simply Not Do That. If you get Series B pricing on your Series A round, you can save the extra equity and put the surplus in a lock box, even if yes as Frog says you can open the box. In theory you could even take some of that surplus off the table for yourself, and still be ahead on equity, why not (the answer to why not is ‘because people would disapprove’)?

These clearly +EV plays being frowned upon is a sign you are not fully in The River.

The third argument, which has merit, is that if things are going this well, you don’t need to negotiate hard on price in order to be in great position to succeed. You can afford to give a discount to ‘market’ price in order to choose the VC partners you want to have around and to signal by involving, buy their goodwill and help and future inclination to fund, and save time so you can go build the business. That’s fair, but only up to a point.

That’s why top VCs can say things like ‘the price you enter at does not matter’ or otherwise not care much about price, and get away with it. The markets involved are not allowed to clear. And if a VC offers things as part of their package, that has big advantages in winning deals over ‘pay more money that is more valuable.’

“Basically ninety percent of the fight is over before it begins in venture,” Andreessen told me of the process of recruiting founders. “Like, at the point of contact with an entrepreneur, when we go in and we do our whole sales process and try to win the deal…ninety percent of that fight is over before it begins, because it has to do with the reputation that we’ve established, the track record—you know, the brand.” (5200)

“It’s sort of a self-fulfilling prophecy, the whole thing,” he said. (5205)

So the top VCs by reputation, by connection to opportunity, get the best deal flow. When they want the deal, they get it, and they mostly get the size they want. And those top deals are systematically way underpriced. Everyone else has to pick over the remainder, and deal with adverse selection.

Ironically, I know people who have turned a16z in particular down because of aspects of their reputation. But that only reinforces Marc’s point here, that the battle to fund a given great company is 90% over before it begins.

Does a16z do a good job of selecting the great companies and deals? Does Founders Fund? What about other top VCs?

I cannot know for sure from where I sit, but based on what I do know, I think no. I do not think they are especially good at picking winners, or at supporting winners. Indeed, based on various investment choices of theirs especially in crypto, I think a16z does a below average job of picking great companies.

But that does not ultimately matter anywhere near as much as deal flow. So they win.

I still say that my best guess is: More than all of a16z’s Alpha is in their deal flow.

How much Alpha is there? A lot.

Andreessen recited from memory data on a typical returns portfolio, which he later confirmed by email: 25 percent of investments make zero return. 25 percent produce a return greater than 0 but less than 1x. 25 percent produce a return between 1x and 3x. The next 15 percent produce a return between 3x and 10x. Finally, the top 10 percent produce a return of 10x or greater. (5217)

That profile is insanely great if you can get it. If you put everything at the midpoint of the range and the top 10% at 20x, that’s a 260% profit, without even accounting for the true smash hits that traditionally are the key profit center. Who wouldn’t want to invest in that?

Although it is worth noticing that a16z will often put vast amounts (nine figures) into relatively mature companies, where the chances of those 20x+ returns are lower.

Isn’t it grand that founders aren’t seeing that and demanding higher prices?

Again, this all only works with top deal flow.

It’s important to clarify that this data pertains only to what Andreessen calls “top-decile” venture firms like a16z—not just anybody renting a room in an office park on Sand Hill Road. The industry as a whole does not generate particularly attractive returns. But the top firms can be extremely profitable, targeting a 20 percent IRR and going north from there, perhaps even 25 or 30 percent or higher for high-risk sectors. (5223)

That return number very much lines up with Andreessen’s distribution.

Note that it does not imply that it would be worthwhile to raise more capital. If your advantage is deal flow, you are already taking full advantage of all the best deals you can find. Adding more capital to a16z would force it to, effectively, fight through the same adverse selection problem as everyone else, taking deals a16z would have otherwise already passed on. It can both be true that a16z returns 20% IRR, and that raising another fund for what remains would not be creating a good investment.

The Valley Beside the River

Throughout this book, I’ve dealt with some controversial cases as we conduct our census of the River. Is Donald Trump a Riverian? No, he’s not analytical enough, despite his history in the casino business. Elon Musk? Yes. (5416)

I continue to think Trump is not a controversial question. If you still have doubts, you could watch or think back to his debate against Kamala Harris. None of the key attributes are there. Elon Musk, along with Silicon Valley in general, is a more interesting case. I want to give him and SV in general partial credit, but not full credit.

I am willing to say that Silicon Valley is, at least, River adjacent. That they share some of the River nature, and its virtues. They think of themselves as kindred spirits, and that makes it at least partly true. They do celebrate accomplishing big things, doing that which works and figuring things out, and the taking of risk. Most of all, they want other things too, but they very much want you to go out and Do Things, build something people want, and ship.

As long as it’s the right kind of risk and right kinds of things to ship, aimed at the pattern matched big things. There is a lot of real underlying it all, and also more than a little cargo culting going on. Even though it is indeed the world’s best cargo, and thus perhaps the world’s finest cult. But it is a reality tunnel, an echo chamber, a realm of the zeitgeist and the vibe, and within its own tunnel it is deeply conformist to its tales of exactly how to be contrarian.

Which, again, is a great thing to which to conform. If you have to choose, it’s hard to do better, and very easy to do far worse.

All of that merely scratches the surface. Silicon Valley is not Nate’s world, and the shift from things Nate knows well to places he is exploring is clear. He got great sources, and those sources are in some ways highly honest and forthcoming. In other ways, they are very good at the hoodwink.

Another interesting note is that in the final chapters, Nate talks with Sam Altman and Paul Graham, as well as a bunch of SV-adjacent others. It’s not clear how much those conversations impacted the Valley chapters, but if I wanted to talk to one person to truly understand SV, Paul Graham would be very high on that list. And it’s interesting how little it feels like Graham’s perspective got mixed into this.

Is all of this meta-rationality? Is it the River taken to its logical conclusion, that you should take the calculated risk of taking uncalculated risks according to a set of heuristics that has proven to be successful?

Thus, shouldn’t we say that you absolutely should (from a fully amoral perspective) be betting on and investing in Adam Neumann and Sam Bankman-Fried, telling them to be even crazier, and being happy you are getting good odds on a huge payoff?

One could plausibly argue that. I mostly buy it. It’s not the same as the River nature.

Checkpoint Three

At this point, we transition from Silicon Valley to Crypto.

I am far more interested in the true nature of Silicon Valley and Venture Capital than in Crypto. SV and VC are vital to our future. Crypto… likes to claim it is the future. I find that claim rather doubtful, increasingly in a ‘I don’t think about you at all’ kind of sense.

Still, there’s some good stuff that one cannot ignore.

Fun With SBF and Crypto Fraud

If you want my take on What Happened with SBF, read Book Review: Going Infinite.

I will not be explaining again. Nor will I cover the crypto basics, IYKYK and if you don’t then there are plenty of other places to learn if you’re curious.

I understand why Nate Silver focused so much on SBF before SBF blew himself up, and why he kept that focus after SBF blew up, given his level of access. It’s a great story, and I do think it is a story that matters. I still wish that it wasn’t so centrally used as the pathway into Effective Altruism.

I do think EAs themselves need to heed the cautionary tale here, but SBF represents an extreme tail risk failure mode on many levels, and has been used to attempt to condemn anyone and any concept remotely related to EA. This has plausibly substantially decreased our civilization’s ability to sanely respond to AI. Presenting him this way to a casual audience unfamiliar with even the basics gives me a lot of ‘no one picks on my brother but me’ vibes.

Nate Silver did have the benefit of doing direct interviews with SBF, so he has some additional puzzle pieces, or at least fun anecdotes. Here are ones I loved the most, starting with another very good opener.

The room was getting darker, the power was going out on Sam Bankman-Fried’s laptop, and he was telling me increasingly unhinged things. (5399)

That sounds about right. My favorite part of this is the power going out on the laptop. What could be a better metaphor for failed risk management? Actually needing your laptop charged, and running out of charge, while sitting in your work area, really should be a Can’t Happen. There are any number of ways to plug in your computer.

Nate gives us more evidence that SBF ended the story rather full of delusions of hope, and that he had indeed fooled himself throughout the process.

The specific number he would later give me was 35 percent—a 35 percent chance that he’d emerge from the situation with something that would be “considered a win” by knowledgeable people. “I know a lot of people will say I’m crazy for giving this number. It’s an insanely high number. And, like, it can’t possibly be right,” he had qualified. (5444)

If you had the proper River Nature, and you noticed your number could not possibly be right because it was insanely high, you know what you would do? You would revise the number downward, at least until this was no longer obvious to you. Disagreeing with ‘a lot of people’ is fine, but ‘can’t possibly be right’ very much is not.

SBF learned to fake adjusting his priors for such situations when necessary - I can testify to that first hand. But SBF had no interest in doing this for real. About anything. Ever.

SBF’s continued embrace of the ‘it was all tactical errors’ argument is even sillier.

The way SBF framed things to me, these were forgivable, tactical errors—like a poker player playing a hand suboptimally in a challenging situation. “You put those together and it went from, like, significant but manageable to significant and not manageable,” he said. “I just sort of lost track of it,” he told me at another point.

In reality, of course, any of these would have been mission-critical mistakes, let alone all three put together. It was like a pilot saying, “Oh, the plane wouldn’t have crashed if only I hadn’t drunk three bottles of whiskey, punched out my copilot, and then told air traffic control to fuck off when they said the runway was closed.” It’s also clear that Sam’s story was almost entirely bullshit. (7032)

SBF kept saying this to Michael Lewis. Here he is saying it to Nate Silver. The difference is that Lewis treated it as a non-absurd statement, and Silver didn’t.

Nate notes that Sam had several different modes of talk. I saw that too. When you could get Talking-Shop Sam, as Tyler Cowen did in their conversation and Matt Levine, Joe Weisenthal and Tracy Alloway did on Odd Lots, it was great stuff.

Talking-Shop Sam spoke rapidly in run-on sentences, full of jargon about arbitrage and expected value. I liked Talking-Shop Sam. We spoke the same language and had good camaraderie; I could imagine us being friends. (5464)

Then there was No-Shits-Left-to-Give Sam—a more candid and seemingly more honest personality, a Sam who prefaced his responses by implying that now he was going to give you the real answer. No-Shits-Left-to-Give Sam could get dark and erratic—this was the version that I was a tiny bit worried might suddenly lunge for my digital recorder. Finally, there was Sneaky Sam. This iteration of SBF was pedantic and lawyerly. (5466)

NSLTG Sam is fun too, especially when he gets dark.

Sneaky Sam is fun in his own way, too. Why not play some poker with him?

“Um, what do you think the odds are that one year from now, FTX will be an operating platform? And what do you think the odds are five years from now that customers will have been made whole?” “To the first question, I’d say under ten percent. To the second question, thirty percent. I have no idea.” Oops! I had fallen for Sneaky Sam’s trap. (5483)

As it turns out, these were great estimates. Ten percent seems high, but there was model error and unknown unknowns to consider, it was not so crazy (at the ~10% level) that someone might see sufficient brand value, especially if they’d first bought up a ton of claims. And it’s hard to fault the 30% given it actually happened purely because Number Go Up once more - between Anthropic and Bitcoin, 30% seems high but not crazy high for ‘enough profits are made that FTX becomes solvent again, as measured by USD liabilities at time of bankruptcy, before paying the IRS and other fines.’

It was only possible because the liabilities were converted to USD, while the assets were not, and the assets were highly volatile. And yes, this implies FTX debt was trading super cheap for a while. The Efficient Market Hypothesis is false.

For those who think it is hard to spot the crypto frauds, it’s not all that hard:

I wondered about skilled crypto trading. What differentiated the more successful traders from the less successful ones? (5562)

[Mashinsky, head of Celsius’s] reply was darkly cynical, although served with a twist of humor, because Mashinsky tended to laugh at himself when he spoke. “Poker is about skill. This is not about skill. This is just about getting on the bus. (5563)

[We] lend it out to FTX. FTX needs liquidity,” he answered instantly. “You can give it [directly] to FTX. Are they gonna pay anything? No. You deposit through me, I’ll squeeze them to give me three, four, or five times more than they’ll pay you. Because I have a million and a half people that are all together, marching as one.”

“That makes sense,” I said, not sure it made any sense at all. “It’s very simple,” Mashinsky replied. (5574)

If that’s what the guy loaning out money to other crypto people at 18% says?

Yeah, he might be a fraud. And you almost know FTX is a fraud as well.

The whole story makes no sense. If a crypto exchange is borrowing money at 18%, effectively using a credit card, what, you think it is solvent and has all the customer deposits?

Nate agrees with me that, in at least some sense, SBF was in the end sincere.

Despite Bankman-Fried being an unreliable narrator, my best guess from talking to him and many other sources is that his interest in effective altruism was at least partly sincere. (6190)

Other Crypto Thoughts Unrelated to SBF

Crypto is many things. To many, it is only one of those things, which is gambling.

Here’s something I learned when writing this book: if you have a gambling problem, then somebody is going to come up with some product that touches your probabilistic funny bones. (5794)

And whichever product most appeals to your inner degen will be algorithmically tailored to reduce friction and get you to gamble even more. (5798)

In the future this will only get more extreme.

Nate draws distinctions between the characters and ethos of different blockchains. Some of it he links to different origin stories, some to distinct functionality. And some to the fact that Bitcoin’s core value proposition is as the ultimate Shilling Point: A form of originalism and uniqueness, taken to the level of religious fervor.

Some of this is because of the origins of the respective blockchains and the cultural baggage they carry with them: Ethereum was a quasi–Silicon Valley startup, while Bitcoin was a cyberlibertarian alternative to fiat currency. “Ethereum is an ecosystem for venture capitalists to make bets on the future of computing and, like, Web3 and DeFi and gaming and NFTs and all that shit,” said Levine. “Bitcoin is a place to make bets on future institutional adoption of an economic asset class.”

But that doesn’t fully explain the fervently anti-Ethereum attitudes that Buterin encountered. He likened Bitcoin maximalists—who often proclaim that Bitcoin is the only worthwhile cryptocurrency—to religious adherents. “Bitcoin is not really a technology project. Bitcoin is a kind of political, cultural, and religious project where the technology is a necessary evil,” he said. It’s common in the River, a highly secular place, to insult someone’s movement by comparing it to a religion. (For instance, you might say that wokeness is a religion or that effective altruism is one.) (5888)

To repeat, famous works of art don’t become famous for completely arbitrary reasons. Focal points usually have something going for them. But the traditional art world almost goes out of its way to intensify focal points by curating a sense of exclusivity and scarcity. (5939)

Statistically speaking, yes, as art gets more expensive it gets more likely to be high Quality in an abstract sense, but from the outside view, past a certain point it is mostly a social manipulation, bragging rights, one-upmanship, tax evasion and money laundering game.

In other news, have you heard about CryptoPunks?

Do you have a friend who owns a CryptoPunk? I have a few of them, and what I can tell you is it’s very likely that they’re going to show you their CryptoPunk when the opportunity arises. In some sense, that’s the whole point of owning one. (5957)

Checkpoint Four

Crypto took up a lot of the book, but that’s all I have to say about it here, so this was a relatively short section. It’s too juicy for Nate to pass on, but it’s mostly not too juicy for me to pass on.

What remains is the ‘relevant to my current interests’ sections. It’s rationalism, EA, AI and AI existential risk, which we cover in the final post.