Longreads + Open Thread

Longreads

• Here's a 2000 Landon Thomas Jr. profile of Ryan Jacobs, who started managing an Internet-focused fund with $200k in assets, ran it up to $600m in assets under management two years later, and ultimately lost two thirds of his investors’ money. Which is a good reminder that these stories don't have to rhyme in any particulars other than the volatility: Jacobs got to run the fund because he was working at a small asset manager and had a lot of hustle, and the willingness to run what was then a failing strategy. He made the canonical 90s mistake of dreaming of total addressable markets without thinking of the mechanics of getting there. (Ironically, the bubble helped solve a lot of this problem by making distribution easier, and then un-solved it as the companies that controlled that distribution exercised their pricing power.) When Jacobs was plying his trade, it wasn't cheap to get exposure to particular kinds of beta, and you could do well by giving investors a high-fee vehicle to express a generally positive view on a sector. Today, it's still possible to sell a strategy that is structurally long some factor, but investors will want to know how you're adding alpha on top of that.
• Tuhin Chakrabarty, Xinyue Liu, Jane C. Ginsburg, and Paramveer Dhillon on how AI-written books are crowding out human-written ones, but at different paces in different genres. But one of the takeaways is that this is happening in the genres you'd already think of as long-tail, quantity-over-quality domains. There is some amazing genre fiction out there, but there are plenty of authors who are churning out one replacement-level sci-fi book, business profile, or romance novel after another. There are different audiences for different books, and in some categories books are more a scratch-the-itch phenomenon than something that's supposed to be life-changing. So this is partly a market efficiency story. There's a segment of the book-reading audience that liked certain topics but wasn't that sensitive to quality, and there's now an automated way to meet that demand. Serious literature will be safe for a while yet, though writers may face some harsh truths about what the average reader really wants.
• In the Center for Educational Progress Substack, Justin Rosentover writes about his experience with the Montgomery Blair High School magnet program. This is a fun piece because it's a great example of a local government running an experiment and getting good results: you can offer high school students more advanced coursework in parallel to the standard offerings, and you can do it in a way where there are students who default to more challenging coursework, and students who do it ad hoc. People vary tremendously in their academic ability and motivation, both in terms of average skill and in terms of subject-specific competence, but it's also helpful to be surrounded by peers who genuinely care about the robotics team more than the football team, and vice-versa. There's an opportunity for schools to flex a little and benefit their students a lot; hopefully more of them take it.
• Joshua Rothman in The New Yorker on lookism, i.e. discrimination based on personal appearance. Speaking strictly as a haver-of-takes, lookism is a great phenomenon to be long, because it's obviously a huge issue (not only do attractive people get paid more, but there's a lot they simply don't have to pay for), it's somewhat contrarian to explicitly advocate for People of Unattractiveness, and such people will be overrepresented in textual media due to selection effects. If you're in the business of reading and synthesizing people's views, this is an absolute gold mine: you can make the clear argument that it's deeply unfair that certain face and body proportions will give someone better jobs, lower rent, a lower chance of being convicted of a crime, etc., and you can also explore the question of how much "attractiveness" just refers to body fat, and whether this discrimination is somewhat rational—a landlord might indirectly make the judgment that someone who skips appealingly flavorful potato chips would also be less likely to write a check that's going to bounce. Within lookism, heightism seems like the easiest moral case to make (disclosure: 6'3", taken), since for people raised in the developed world it's almost entirely genetic luck, and it's strange that at least for now it's socially acceptable to discriminate on this basis. It's a topic that will generate endless takes once it gets big.
• Dylan Matthews in Asterisk on John and Mack Rust, who spent years trying to commercialize a mechanical cotton picker, as part of a political mission to empower workers. Picking cotton by hand is laborious and unpleasant—my grandmother mentioned having to do this as a kid, and the main thing she emphasized was that you'd be amazed at how long it takes to fill up a bag with cotton, and then amazed again when you learned how little it weighed. But that's part of the story of why it took so long to automate: if you're in a part of the world where kids need to do agricultural labor to support their families, it's a part of the world where labor is very cheap, and a machine needs to be very efficient indeed to displace that. But, at the same time, automation begets more automation: as output rises in other sectors, the friendly version of Baumol's Cost Disease shows up: employers in a rich country find that they can't afford not to automate drudgery.
• In Capital Gains this week, we look at why basically all statistics about the share of humans versus bots in some domain are bogus. It's not that they don't measure something, it's that they measure too many things at once: how much work can be automated, how much work people choose to automate, and the fraction of low-stakes interactions that can happen without a human in the loop. There are already fields where bots technically make 90% of the decisions but humans are actually in charge (like big online ad buys or big stock trades) and decisions where humans make almost all of the decisions but the bots are actually in control (like when people feed some algorithm data on what fraction of interactions with an app should have a particular call-to-action).
• A very fun Read.Haus question this week: is pricing more efficient for individual stocks or for broader factors. This one is really hard! On one hand: factors cover multiple assets, and if every asset is being priced reasonably well, factors should get some benefit from diversification and be even better priced. On the other hand, within organizations there's a correlation between thinking about factors and wanting to have zero exposure to them: there is probably more money in figuring out which AI or Iran war play is mispriced relative to some other one than figuring out whether AI and Iran are big deals or not, even though there's more dollar alpha in the latter category. It's entirely possible to go through life adding a lot of efficiency to the market by determining which overpriced company in a sector is least-overpriced, while being blissfully agnostic as to whether or not the sector in question is overpriced. And sector bets, or country bets, will have lower sharpe ratios than relative value trades within them, simply because there's a bigger sample size and more room to isolate risks that are worth taking. If you want to make a living picking stocks, be indifferent to broader trends and very attentive to which companies are over- or under-indexed to them. If you want to be a legend, focus on the trends; you'll have a better story. (But this, too, is part of the market-efficiency story. It's just less fun to say that you made your money betting on the gap between Hut 8 and Nebius, versus realizing that Nebius and all its peers will earn great returns on the GPUs they buy. So you have to choose whether you'll optimize for a great story about being a contrarian, or a less fun story about how you survived the grind.)

Books

Magnetic Mountain: Stalinism as a Civilization: For an infamously long time, Paul Samuelson's economics textbook kept arguing that Soviet economic output would overtake the US's, though later editions kept pushing back the date by which this was supposed to happen. But in his defense, there was a point at which the Soviet economy was growing quickly (it did even better in per-capita terms, Stalin being the kind of guy who recognized that this fraction had a denominator, too). And no project exemplified their approach to growth more than Magnitogorsk. Russia was, due to a mix of strange historical contingencies, the first country to go communist even though orthodox Marxist theory held that this stage would happen in industrialized countries. Like China under Deng, they realized that there just weren't many means of production for the workers to seize, and they'd better find a way to make some. Magnitogorsk was basically an attempt to build a completely new city, built around productive labor, and designed from the start to celebrate socialist values.

They had a nice head start on the steelmaking part, because the city was located near one of the richest iron deposits in the world. (The place name means "magnetic mountain," because early settlers noticed that compasses weren't accurate there.) They brought in an American engineering firm to design the steel works, and hired a German architect to lay out the city. But also, they started building the plant with incomplete instructions, and had to scrap and redo the preliminary work. Construction of the city started before the plans were done, so dwellings were a mix of intended-to-be-temporary barracks, tents, some actual apartment buildings (usually with semi-private living spaces, a few shared bathrooms, and one shared kitchen per building). There was also, inevitably, a ritzier and more exclusive neighborhood where managers and party elites lived.

In some ways, this feels like a book about a startup: the people in charge tended to be young, and to have mostly technical backgrounds. They had high morale, tried to be adaptable when their initial plans didn't work, etc. But it's hard to build a startup in a society that's also in the process of being completely reshaped: when they recruited workers from other cities, for example, the union those workers belonged to in their home city might refuse to give the workers permission to travel. When they ordered supplies, they tended to over-order, because they didn't know when they'd be interrupted, but that meant they had excess inventory which they had trouble tracking—and shipping was also unreliable, because railroads would write a given train car's destination on the side in chalk, so if it rained during the journey they'd lose their records.

There are some moments of black comedy. The most enthusiastic workers were members of the Komsomol, the communist youth league. Komsomol members were incredibly motivated to work hard building socialism, which sometimes meant building dams. They worked extra shifts, wrote about each other in Komsomol newspapers, got the dam done way ahead of schedule, celebrated by making busts of Lenin and Stalin, and didn't write any follow-up articles about how the dam wasn't deep enough and the city faced water shortages for years. It took decades for the media to finally portray this as it must have looked at the time.

This is, at many levels, an exhausting book—by the end, you will have ploughed through many, many details about what life was like in 1930s Magnitogorsk. But it's also frustrating to keep seeing so many wonderful plans fall apart as they collide with reality. The people who designed and built Magnitogorks really were trying their best, and the USSR was able to industrialize to an impressive degree. And yet, they weren't able to eliminate the chronic waste of capitalist competition so much as they were able to transmute it into even more wasteful processes. It's somewhat poignant to reflect on how much of this was the result of information poverty; you could write a version of Harry Turtledove's Guns of the South where a communist time traveler goes back to 1925 and gives the USSR a single laptop running Excel, and suddenly these calculation problems are more tractable. But every economic system expands to the limits of its information-processing capabilities, and once you have tools like these, the planning questions get even more complex and you need something better.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• What are some other good books about colossal engineering projects, both successful and not?

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)
• Lightspeed-backed team building the engineering services firm of the future is looking for founding members of technical staff excited about working alongside civil engineers to translate their domain expertise into the operating system that powers the next era of great American infrastructure. If you’re an engineer with strong product intuition, who's energized by access to users, and excited by the prospect of transforming how we design and construct our built world with frontier AI, this is for you. (NYC, SF or Remote) 
• Ex-Anduril, Ex-Abnormal Security, Ex-Bridgewater, fast growing startup bringing agentic cybersecurity to 99% of businesses via MSPs is looking for platform and machine learning engineers. Startup experience preferred; what matters most is that you've grown in scope and handled ambiguity over the last few years. (SF)
• Series A startup building multi-agent simulations to predict the behavior of hard to sample human populations is looking for a founding recruiter who’s able to attract and close the best research and engineering talent in the world. Experience building high-quality teams as a former founder, VC, or operator a plus. No formal experience in a “recruiting” function required. If you have experience communicating and persuading smart, disagreeable counterparties of your vision, this is for you. (NYC)
• AI Transformation firm with an ambition to build an economic world model to run swathes of the private, unstructured economy is looking for Systems Engineers, Platform Engineers, and business generalists who understand how to solve problems.  

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.