Longreads + Open Thread

Longreads

• Daniel Kolitz reflects on two episodes of economic disempowerment in The Baffler: first, he lost his job as an SEO copywriter because LLMs could do it better. And then, he got addicted to payday loan apps, and kept finding himself deeper in debt. I also did some SEO copywriting long ago; it was never exactly fun, but was a kind of universal basic job for people who could type fast. The payday loan piece is interesting, with all of the usual pathologies of extremely-subprime small-dollar finance. But the most striking part is this: "As I write this, I owe $359.96 to Cleo, $31.99 to Albert, $62.99 to Brigit, and $86.12 to Dave." These companies have to spend a lot per dollar of average balance to acquire customers, and a lot per dollar of interest to collect. So the broader economic point the piece makes is weak: the author is, as a customer of a financial institution, worth a lot less than someone who has a mid-five figure credit limit and always pays their bill, or someone who takes out a mortgage. These payday loan companies are not part of some broader scheme to hook everyone on being part of the permanent economic underclass; they're a tiny corner of the financial system that makes small sums in depressing and operationally annoying ways.
• John Psmith of The Psmiths reviews Inga Clendinnen's Aztecs. I cannot caution you enough: if you're squeamish, exercise caution when reading this, and if you see a blockquote you should slam your laptop shut. The Aztecs were an advanced, well-organized society, and what they were organized around was nihilistic torture and human sacrifice. They are, as the piece points out, basically what you'd get if your society's dominant religion was the worship of Lovecraft's elder gods. But, in a way, there's a positive message here: technology and learning are not, as the twentieth century illustrated, a cure for barbaric behavior. But the empires in question tend to rise and fall fast, or, if they stick around, it's because they've forsworn their previous blood-soaked approach. (Also, be sure to read the footnotes, particularly the last ominous one about Fable finding a way to basically ritually purify itself after OCRing gory material.)
• Steve Newman at Second Thoughts on why robotics is hard. One thing he points out early is that robotics demos are particularly misleading, because it's more straightforward to make a robot superhuman at one task than to make it good at tasks in general, but the robotics bull case rests on the idea that we'll be able to replace many different kinds of physical labor with robots. Another big issue is that their world models need to be so comprehensive, and that this also applies to training—the closer a task is to a real-world one rather than an incremental one ("bake a cake" vs "crack an egg") the more expensive verification gets. And we still have unanswered questions on form factors (one possibility here is that much of the world will use robots with legs, but the US will have more robots on wheels—the Americans with Disabilities Act means we are a nation of convenient ramps).
• Data Colada has an investigation of a Dan Ariely paper with ridiculously good evidence of manipulation. The paper purported to study whether or not students' grades were affected by setting their own deadlines, and further by how those deadlines were spread out. It doesn't seem to show that. This kind of statistical sleuthing is delightful: they were able to reverse-engineer the exact manipulations made, and then go back and find reviewer comments asking for precisely those results! One reason psychology-adjacent fields are so vulnerable to fraud is that their outputs are so flattering to the reader. Imagine being the only person who knows that drawing little eyes on a piece of paper makes people more honest—you're wise, and everyone around you is just a meat machine running silly little programs. It's very hard to pull back from that and ask if the study replicates. (No.) The other reason is, of course, that most people in academia don't pause to ask if the paper they're reading is based on fabricated data. (There's actually another issue with the data: the study looks at granular measures of student grades, including participation. These are expressed as percentages. The scores for participation include numbers like 92.3%, 89.6%, and 80.77%, the last of which is truncated, and to get these numbers out of a fraction with an integer numerator, it needs to have a denominator of 3,000. The class met fourteen times, and if we assume there isn't a participation score on exam day, this implies that students' class participation was rated on a ~230-point scale. Meanwhile, if you look at the full data and focus on the subset of students who set their own deadlines, there isn't much of a correlation between students' scores on different components of their final grade. So one compelling possibility is that these scores, too, were made up.
• Tyler Cowen interviews Michael Moritz. As someone who wrote about tech and then started doing venture capital, I'm obviously interested. (As it turns out, we also have similar taste in unpleasant but compelling art; Moritz owns some Otto Dix art, and I, at a much broker point in my life, had to settle for using his Dr. Hans Koch as a Facebook profile image for a while.) Moritz is a great example of the media/venture fusion: he got into the VC business because he covered Silicon Valley for Time, and wrote an early book on Apple. He’s also a mild corrective for the view that tech journalists are inherently conflicted because of this career possibility—he was responsible for publicly revealing that Steve Jobs had had an illegitimate child, which apparently didn’t burn enough bridges to preclude a long and successful career investing in other tech companies. 
• In Capital Gains this week, we're on the market history beat. Specifically, the legendary Resorts International short squeeze of 1978. There are many writeups of this one out there, but one thing some of them miss is just what a shady company Resorts was: one of their promoters had friends who kept dropping dead, another bribed a judge, Rolling Stone accused them of being a CIA front (and then retracted it), and they may have been advised by Meyer Lansky. The hedge fund manager Robert Wilson is most famously associated with being on the wrong side of the trade, but there's some fun circumstantial evidence that Michael Milken was, too.
• A Read.Haus reader asks if the AI trade is causing companies that are further from pure AI but still connected to the datacenter buildout, like Caterpillar, to be more volatile. There are two arguments that these stocks should be even more volatile than pure AI names, one fundamental and one behavioral. On the fundamental side, the more disconnected some company is from the AI supply chain, the less likely it was to invest ahead of bottlenecks—someone selling GPUs has a very good sense of how big AI is, but for someone selling construction equipment, it starts as a minor boost to sales and then just keeps on growing. If they're producing at close to maximum capacity, they can ration with some combination of longer backlogs and higher prices, i.e. through some combination of higher margins and more predictable earnings. But this also means that a marginal drop in demand falls straight down to them. The behavioral reason is more fun: some traders in these stocks are AI-pilled, others think of them as cyclicals going through a particularly big cycle. And these two groups will anchor to completely different P/E ratios. So there's a big air gap between where a pure industrials analyst gets excited about a late-cycle cyclical and where an AI bull gets excited about a critical component provider in the early stages of an unprecedented buildout. We actually saw some of this when DeepSeek hit: the GPU stocks did badly, but the gas turbine stocks got hit harder.

Books

American Scoundrel: Roy Cohn's Dark Journey from Joe McCarthy to Donald Trump: One feature of technological and economic complexity is that it requires more detailed rules. When people are settling the frontier, it's pointless to enforce some kind of zoning; you only do that once they've clustered in cities and everyone's close by. Similarly, smaller communities need simpler laws around how workers get treated and how contracts are enforced, because social censure is enough to keep people mostly in line. As things get more complicated, you need more rules. This sometimes leads to characters who were born for a more freewheeling atmosphere, and adapt its norms to the letter of the law in a more settled state. And that describes Roy Cohn: anti-communist activist, lawyer, wheeler-dealer, white-collar crook, AIDS victim, and Trump mentor.

Trump is the main reason there's a big Cohn biography, and also the reason there were two movies in the last decade (Where's My Roy Cohn? and The Apprentice)) about someone who died three decades earlier. But, reading this book makes Trump's status as a Cohn protégé feel less significant, if more revealing.

If there's one goal that explains most of Cohn's behavior, it's this: when someone says "I know a guy who knows a guy," he wanted to be the first guy for as many guys as possible. In particular, he liked being a go-between who helped connect legitimate businessmen with the kinds of people who would protest that they were merely legitimate businessmen. It's a useful role. If some parts of the economy operate based on personal loyalty and favors, and other parts are more focused on institutions and formal obligations like laws. Someone who knows enough of the law to know what they can get away with can intermediate between these groups. That's basically how society negotiates the pace at which things get formalized.

It's hard to come up with other motivations for Cohn. He first achieved public prominence as a lawyer and investigator for Joseph McCarthy, where he worked alongside his very close friend David Schine. He and Schine were close enough that there was a fair amount of innuendo about their relationship, and it didn't help matters at all when Schine got drafted and Cohn started pushing the army to give him a desk job, ideally close to D.C., and then started investigating the army for communist infiltration. This could have been an almost touching story of forbidden love, except that Schine went on to marry Miss Universe of 1955; they had six kids and were married for 39 years until they died in a plane crash.

(Edit: Corrected the spelling of Schine's name.)

Cohn himself had an energetic social life, and seems to have carefully calibrated his sexuality so he was closeted enough to avoid public knowledge about his sexuality, but out enough to know who else was in the closet. Cohn, by virtue of being a political pariah who'd dodged numerous charges for assorted white-collar crimes, had less to lose than any of them. Apparently Cohn could also be charming, even to people who were primed to loathe him, which makes him either an effective sociopath or someone very good at playing an evil character in public while being a little more normal in private.

Cohn had an undeniable impact on politics spanning decades. It's hard to tell exactly why he did this, or what he really got out of it. Which is probably just the way he liked it.

Open Thread

• Drop in any links or comments of interest to Diff readers
• Roy Cohn is a great example of a fixer. Are there any good books on more benign versions of the archetype? (It's funny to think that agentic AI also functions as a fixer, at least for problems that are purely digital. And given what some agents have been getting up to lately, maybe Cohn's more flexibility is representative of a critical requirement for performing that role at a world-class level.)

Diff Jobs

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

• A top prop trading firm is looking for people who combine exceptional quantitative ability with the judgment and relationship instincts to build critical financial partnerships and lead high-stakes negotiations. Candidates may come from markets, finance, or further afield; what matters is this combination of abilities, strengthened by the perspective that comes with experience. (NYC)
• A top prop trading firm is looking for people with exceptional strategic thinking and quantitative skills to help the firm model and manage its financing risk and strategy. Open to candidates from a range of analytically demanding fields.(NYC) 
• Ex-Bridgewater, Worldcoin founders using LLMs to generate investment signals, systematize fundamental analysis, and power the superintelligence for investing are looking for machine learning and full-stack software engineers (Typescript/React + Python) who want to build highly-scalable infrastructure that enables previously impossible machine learning results. Experience with large scale data pipelines, applied machine learning, etc. preferred. If you’re a sharp generalist with strong technical skills, please reach out.
• Series A, Thiel-backed team building full-stack software, hardware, and chemistry to end water scarcity, is looking for an ambitious product engineer to help build operating platform and internal tools that weather modification operators can use to make weather, geospatial and flight data useful to plan, monitor, and review campaigns. If you’re excited to own product outcomes through deployment and are strong at turning messy problems into good product (using Typescript and React), this role is for you. (Los Angeles)
• Ex-Citadel/D.E. Shaw team building AI-native infrastructure to turn lots of insurance data—structured and unstructured—into decision-grade plumbing that helps casualty risk and insurance liabilities move is looking for a data scientist with classical and generative/agentic ML experience. You will develop, refine, and productionize the company’s core models. (NYC, Boston)

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.