Longreads + Open Thread

Longreads

• John Psmith reviews Xenophon's Anabasis, the completely insane story of a group of Greek mercenaries who, in 401 BC, managed to: got hired for some minor military expedition, find out (after they've traveled too far to turn back) that they've actually been hired to help Cyrus the Younger seize control of the Persian Empire, and win a military victory on his behalf. At which point they learned that Cyrus had died, which meant they were now unemployed and deep in enemy territory. This is many problems at once, but the one that everything else depends on is organizational: determining who's in charge and what they're in charge of. Which leads to some interesting outcomes: he is, in some sense, a general, but he can't exactly order around 10,000 heavily-armed mercenaries. So he gets their vote on the big choices (often with a bit of gerrymandering) and then gets permission to handle the details, or he'll use divination to determine the next action, but reroll if he doesn't get the results he wanted. Anabasis is timeless, and this particular review makes it timely again.
• In The American Scholar, Jonathan Weiner writes about the nature of memory, particularly how he and his brother differ so much in the way they remember things. Having extremely detailed, specific memories that let you basically rewind to some point in the past sounds like a superpower, but the piece makes it also sound like it's basically adding an autoplaying-at-full-volume video feed to your inner experience; storing memories like that means not extracting and storing specific details. A very thoughtful piece on much more than the science behind variation in memory. (Via The Browser.)
• Brady Brickner-Wood in The New Yorker on men who lie about their height. It's a little strange that height is the rare immutable characteristic that it's perfectly socially acceptable to be a bigot about. And it's also not uncommon for people's revealed preferences to be less bigoted than their stated ones, which is fairly unusual. Once it's widely-understood that people are lying to themselves, it gets harder to preserve a norm against lying to them, and so the equilibrium is that people debate just how much lying, in either direction, is socially acceptable. It's a bit like having weak norms against tardiness, which can be a big economic drag, and hard to fix.
• Jonathan Chait in The Atlantic on the rise of the neo-Brandeisians. The main problems I have with the neo-Brandesians are that they aren't very well-informed, don't have good models to interpret what they do know, and tend to be pretty dishonest and annoying. (See here for a Diff review of a specific work.) Chait has a fun, brutal look at the movement, which is interesting in a meta sense because it gives people a little more permission to start moving away from the neo-Brandesians. Highlights include reading some of their earlier writing, from the 90s, where the movement developed from the observation that Walmart was hard to compete with and would end up controlling everyone's economic interactions. This did not turn out to be true. But the funniest point he makes is that the key insight of neo-Brandesians is that you can break up more companies if you're willing to do so even when it makes prices higher and makes consumers worse-off, and that the movement finally started getting real power in 2022, when inflation was hitting generational highs. This is more bad luck than anything they actually affected, but still kind of funny. (To be fair, the neo-Brandesian argument is that making consumers worse-off by breaking up efficient monopolies today saves us from being exploited by those monopolies in the future. But to make that case, you need to make it very, very clear that you deeply understand why those companies behave the way they do today. That has not happened.)
• Kat Rosenfield writes about being canceled in the early wave of cancel culture. This headline/byline combo made me think that the story was going to be exaggerated, because I think of Rosenfield as part of a cluster of writers who are open-minded and culture war-flavored enough that they'll always be making enemies, but also that they have plenty of publications that are looking for exactly what they have to offer. Imagine my surprise when I learned that that's a new career, and she used to be a young adult novelist before getting in trouble over something that was about 15% ridiculous and 85% deliberately fictionalized. With the benefit of hindsight, we can look at the young adult novel community in general as an instructively bizarre one, but it must have been bewildering and unpleasant to experience this when the whole situation was brand new.
• In this week's Capital Gains, a slight divergence from the usual fare: you never appreciate the sense in which we're all operating at close to 100% of capacity until you can't quite do that, sparked by an exciting injury. (Surgery yesterday, went fine.) It's a good reminder that we get used to surprisingly high levels of economic output, at many different levels, and that snapping one link in a supply chain (or a tendon) drastically limits what can be done.
• A Read.Haus reader asks under what circumstances we could get a glut of compute. The pithy answer is that the more this question is meant rhetorically, the more likely it is to happen. And in more detail: compute is a tricky product, because it's fungible with so many other things: it replaces lots of different kinds of human labor, enhances different kinds in different ways, but also, in the context where it's being used to build algorithms that improve efficiency for industrial processes and logistics, it's a substitute for energy. And in the context of generative ads, as long as there's some upside to using it, it the inference capacity owned by big ad platforms can stay at pretty close to full utilization. But, we could overestimate how much economic growth there will be, how much room there is to optimize—what if the first trillion dollars actually does help us exhaustively identify every possible product or service, the next trillion perfectly optimizes the creation and distribution thereof, and the third trillion is a waste? But because of this extreme fungibility, it will be more fruitful to look at which individual companies will run into problems. We've already seen some, like xAI, buy more compute than they could directly use, and have to sell some of it to other companies. For now, that's put a floor on returns. But that's something to underwrite, not to assume.

Books

The Dictator's Handbook: Why Bad Behavior is Almost Always Good Politics: As a general rule, if two people recommend the same book to me over the course of a week, I will almost certainly read it. I really wanted to like The Dictator's Handbook, because it takes a realist approach to political outcomes—the actors aren't states, they're people, and the goals are personal, not ideological. I wouldn't say that this perfectly aligns with my worldview, but I would say that I apply exactly these kinds of considerations often, and find them fruitful. Countries, companies, and movements are all made of people, and even if they have some collective interest, that collective interest is an approximation of an average, not its own distinct thing.

I have three basic problems with the book:

1. It has a model that feels a little more rigorous than it is. The high-level model is that the way power works is that there's a nominal electorate, who don't necessarily matter; there's a narrower selectorate, who actually decide who gets to be in charge; and there's an even smaller winning coalition, who expect rewards in exchange for their continued loyalty.
2. It leans towards a model where most participants in the political system are trying to maximize their money. The book does acknowledge that there are some austere dictators out there, but still tends to assume that their inner circle is in it for the money. But, especially for a first generation of leaders, that's often not true! The shared experience of winning against all odds—the Long March, the October Revolution, Singapore surviving the withdrawal of the UK and the split with Malaysia, etc.—these kinds of experiences engender a deeper kind of loyalty than getting your percentage from Mister Five Percent whenever the government builds a bridge or licenses a mine. You can somewhat rescue the model here by saying that living up to one's ideology is a form of consumption, in which case Mao was the most profligate person of all time—it would take a staggering amount of spending to get that many people to agree to go through what he put them through.
3. There is no polite way to say this, but: whenever the book talks about situations I'm not familiar with, like Samuel Doe's coup in Liberia, I learned a lot. And every time the book talked about topics with which I was more familiar, I felt like it was oversimplifying at best, and sometimes deliberately cherry-picking numbers to make a particular case. For example, when he's telling the story of Hewlett-Packard's entrenched management in the early 2000s, he compares their share price to the Dow. And, to be fair, HP was part of the Dow. But the period in question was uniquely bad for tech stocks, and money was rotating back into some of the old industrials. In the slice of time the book addresses, HP's shares fell 47%, compared to a 9% decline for the Dow. But the S&P dropped 18%, and the Nasdaq 100 dropped 35%. This doesn't change the thrust of the argument, but given that the book is built on what's supposed to be a systematic, rigorous, quantifiable theory, it's a bad sign that they'd fiddle with the data this way. In another case, the book compares two earthquakes in 2008, one in Chile and one in China, and ascribes the higher death toll in China to the Chinese government's weaker incentive to keep people happy. Maybe so! But, personally, if I were writing such a book, and using such an example, I would at least throw in a caveat about how China experienced the largest internal migration event in human history from countryside to cities over that time period, and that this affected both how high building standards were and how vigorously they were enforced. (Some of this could be mere laziness, not a deliberate effort to mislead; at one point the book uses the word "meretricious" in a context where it's clearly supposed to mean "meritocratic." Sloppy!)

The book doesn't apply its model as consistently as I would prefer. For example, it talks a bit about how undemocratic the electoral college is, and also how gerrymandering reduces the share of the electorate needed to win. But then it gerrymanders its own description of this, by noting that in a parliamentary system, you only need enough votes to win half of the seats by one vote, or a little more than 25%. Which is only meaningfully true if you wait until the votes are cast and then draw the districts accordingly. If gerrymandering is bad, one shouldn't do it in text, either.

But the book misses a more important point about elections, which is that there are many mediating factors between the will of the voters and the electoral outcome. For example, if you're an American, you recently had an opportunity to vote for either Donald Trump or Kamala Harris (or cast a protest vote, I guess). So, in a sense, the outcome was your choice. But, how did it come to be that those two people were your top two choices for who gets to control the nuclear codes? It's not like you made a list of all of your friends, ranked them based on various traits, and then said "I think Don's experience developing golf courses is pretty compelling, but my buddy Kamala worked as a lawyer, so I guess of all the people I know, I think probably one of them should be the most powerful person in the world." No, they got selected through two narrow-selectorate processes, that of political parties and of media attention. The book pays almost no attention to the media, other than mentioning Trump's efforts to restrict it, but in practice the media serve the same role that the electoral college originally did: you choose a small set of elites that you trust, and they will shape your worldview in a way that makes the right candidate clear. That's the only practical way to implement democracy; if we didn't have some system for narrowing things down to a small number of people who had institutional support, and actually asked everyone to just name who they personally thought should be in charge, you'd have a much wider distribution of votes, and also a completely dysfunctional system where the President would be a nobody to almost everybody. There's just no chance that 77 million people independently concluded that Trump should run things, and that 75 million of them thought that way about Harris. In essence, the presidential election is a plebiscite-veto where 75 million people said "I don't know who should be in charge, but it definitely should not be Donald Trump," and a few million more said the same thing about Kamala Harris.

All that said, the book does make some informative and interesting points, and realists, even if they aren't so realist when talking about systems they participate in, they still make some great points. For example: it's hard for authoritarian systems to handle succession: you need everyone to coordinate on taking orders from the same new boss, but not to have that new boss kick out the old boss prematurely. So, among other things, this means that authoritarian regimes are very cagey about the health of their leaders. For authoritarian systems to survive for a long time, they need some kind of institutional norm that succession happens at a predictable pace, and that the successor is made clear in advance but also has clear limits on what they can do. The CCP had this for a while, though Xi Jinping's behavior shows that this kind of setup is fragile. The strongest part of the book covers dysfunctional former colonies. When they were ruled from afar, they suffered from the classic problem that there's a narrow group in charge, and they can afford to be indifferent to the plight of the average person. After those countries achieved independence, that was still how they were run: a small inner circle that was extracting wealth from the rest of the country, sometimes in horrific ways (the book contrasts the Battle of Mogadishu, in which the US to extraordinary lengths to rescue a hundred US soldiers, with the Battle of Afabet, in which Ethiopian commanders, worried that their equipment would be captured, bombarded their own trapped soldiers in order to keep their equipment from being captured). So, I can paraphrase the apocryphal Gandhi line: my opinion of decolonization is that it would be a good idea.

Open Thread

• Drop in any links or comments of interest to Diff readers
• What are some books that take a similar realist view to how countries work, but that don't over-formalize the way The Dictator's Handbook does?

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

• Well-funded, frontier AI neolab working on video pretraining and computer action models as the path to general intelligence is looking for researchers who are excited about creating machines that learn from experience, not text. Ideally you have zero-to-one pre-training experience and/or are a high-slope generalist who’s frustrated that the big labs aren't doing this. (SF)
• Series A startup building multi-agent simulations to predict the behavior of hard to sample human populations is looking for researchers and engineers (ML, platform, infrastructure, etc.) to improve simulation fidelity and scale the platform to hundreds of millions of simulation requests. Problem-solving and genuine interest in simulation matter more than pedigree. Experience working with languages with an algebraic type system is a plus. (NYC)
• A Fortune 500 cybersecurity company with decades of proprietary security data is running an internal incubation with a pre-seed startup mentality and a mandate to build something new in AI. They are looking for a founding engineer who can ship fast, an engineer with a security background who’d be excited to contribute to OpenClaw’s security efforts, an AI researcher, and a generalist (ex-banking/consulting/PE background preferred) who wants to wear a bunch of different hats. Comp is FAANG+ and cash heavy. If you want to build something new in AI, but also need runway, this is for you. (SF/Peninsula) 
• High-growth startup building dev tools that help highly technical organizations autonomously test and debug complex codebases is looking for senior product managers who enjoy defining developer-facing APIs and abstractions. Experience with fuzzing or property-based testing a plus! (London, D.C.)
• A leading AI transformation & PE investment firm (think private equity meets Palantir) that’s been focused on investing in and transforming businesses with AI long before ChatGPT (100+ successful portfolio company AI transformations since 2019) is hiring experienced forward deployed AI engineers to design, implement, test, and maintain cutting edge AI products that solve complex problems in a variety of sector areas. If you have 3+ years of experience across the development lifecycle and enjoy working with clients to solve concrete problems please reach out. Experience managing engineering teams is a plus. 

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.

And: we're now actively deploying capital into early-stage companies through Anomaly. Our focus is on defense, logistics, robotics, and energy. If you'd like to chat, please reach out.