Longreads + Open Thread

Longreads

• In Texas Monthly Lauren Larsen has a delightful look at the ice industry. Every industry provides some kind of window into the modern economy and economic history. On the current-economy side, one point the piece makes is that while ice is cheap enough that people don't economize on it, ice consumption is basically a proxy for the frequency of public gatherings. It's the cheapest component of a cooler full of drinks at a party, but if you're priced out of throwing the party, you're buying less ice. And in terms of history, it's a fun example of an industry that needed to invest in infrastructure to harvest demand: in the 19th century, ice was a globalized business, with American ice getting shipped all the way to India—which meant getting shipped to a place that had no infrastructure for keeping it cool. Then the same thing happened again, at a more granular level, when the in-store ice bagger made grocery stores a viable distribution channel. Sometimes, the nature of the product—annoying to store and transport, free in some contexts and priceless in others—causes the same pattern to show up in completely different contexts a century apart.
• A while ago I had some questions for ChatGPT about the earliest days of the brokerage business, which led to this Jill Lepore article in the New Yorker, on America's long and complicated relationship with debt. One useful mental model here is that bankruptcy, or practical difficulties in collecting debt, are actually a form of credit expansion: many immigrants to the colonies were not so much going to America as they were fleeing debts in Europe, and that meant that they were inflationary to the US (i.e. they could spend all the money they earned instead of returning some to creditors) and deflationary for Europe (because they defaulted). US bankruptcy reform follows a cyclical pattern, where restrictions on bankruptcy get loosened after financial crises, and then get tightened up when times are better and more people are defaulting strategically rather than out of necessity.
• Tyler Cowen and Sonia Farrell Pearson on requiring AI agents to have capital or buy insurance so they have an incentive not to misbehave. It's like the Onion post-9/11 bit about the US giving Al Qaeda money to build a headquarters so we have something specific to bomb: it's very hard to retaliate against something that doesn't have some kind of tangible existence, and for now the closest we're likely to get is to give agents something to optimize for and then punish them by giving them a lower score when they misbehave. My personal view is that we'll likely solve this by linking every use of AI to a specific person who can then be punished if their AI misbehaves, sort of like the pater familias concept in Roman law, where a father can be punished for his kids' misbehavior, but also has a lot of freedom to punish them. The trouble with capitalizing agents is that what we're basically trying to do is put a price on the call option that is limited liability. When transaction costs are high, the cost of abusing limited liability is high enough that it happens about as often as is socially optimal—sometimes companies cause problems whose cost is multiples of their assets, but we also don't send the founders of failed companies to debtor's prison. But AI reduces transaction costs in significant but unpredictable ways, and the most powerful models will probably find ways to abuse this. In our current uncertain state, it's going to be hard to set up rules for capitalizing AI agents that don't either set the number so low that it's worth it to spin up lots of risk-seeking agents and profit from the ones that work, or so high that it's cost-prohibitive to use agents at all.
• Andy Masley has a nice piece on how people go ideologically crazy. It's a pretty simple model: if you develop strong opinions on some niche issue, you're likely to have thought about that issue way more than the average person, because there are so many issues people could care about and we all have a finite amount of time. Which means that when you ride your ideological hobbyhorse, you can absolutely steamroll almost everyone you interact with. It's hard to do this kind of calibration, because of course it's natural to assume that whatever quirky belief you have is interesting to other people because it's interesting to you. (Then again, sometimes the rest of the world really is crazy.)
• The Computer History Museum does some great interviews with tech luminaries. Here's one with Gordon Moore and Arthur Rock, especially fun on the early days of venture capital. It's interesting that Rock made so much money backing completely different kinds of companies: he was early in Intel, which at the time had wasn't a consumer brand; he backed Teledyne, which made its money from being smart about issuing and then buying back stock while diversifying across industries; and he did well backing Apple, very much a consumer brand. Rock's view was that he didn't have an edge evaluating technical details, and focused on backing the right people instead. Which, given the track record, seems to have worked.
• A Read.Haus reader asks why people don't run concentrated long-biased strategies with a hurdle rate. I think there are two separate questions here, one on what's worth doing and one the best way to convert investing skill into fees. When there was a less efficient market, both for assets and for talent, it was possible to offer an investment vehicle that bundled together stock picking with general exposure to the market. Now, those are separate because each one can be bought individually, and buyers vary in what mix they want. That said, there are some strategies where it makes sense to bundle alpha together with factor exposures: sometimes it's hard to find offsetting shorts, and sometimes, as with Situational Awareness, investors want the factor exposure, too. But Situational Awareness illustrates how hard this is to pull off in two other ways: first, obviously, taking more factor risk means having bigger potential drawdowns. And second, they were able to raise that much because they'd already published a manifesto laying out a likely path for AI. It's harder to do that in other sectors, but it will be interesting to see if anyone tries it.
• This week in Capital Gains, we look at why it's hard to take advantage of overvalued stock. Companies can get some benefits from exploiting this, but they'll lose a lot of market value in the process.

Books

April Fools: An Insider's Account of the Rise and Collapse of Drexel Burnham:When Michael Lewis published Liar's Poker in 1989, he kicked off a new era of wry financial memoirs. Every few years, someone else carefully reviews their NDAs, weighs their options, and writes their own banking memoir. April Fools was one of the first. It had the advantage that Drexel Burnham Lambert got itself in a lot more trouble than Salomon did, and the disadvantage that Michael Lewis was a tough act to follow. So the book tries to peg people with memorable nicknames, but just can't beat The Human Piranha.

Our protagonist, Dan Stone, worked for Drexel in the convertible bond group. The 1980s were a great time to be in pretty much any financial service business, and convertibles were agonizingly close to high yield bonds, both in terms of who issued them and how they traded. (As Michael Lewis points out, the lower-rated a bond is, the more its price responds to company fundamentals rather than to interest rates. Converts tack on some equity exposure by fiat, and, since that equity kicker means they can pay less interest, they're also popular with somewhat distressed companies.) But in Drexel, one of the sharpest divides was between the Milken group in LA and the rest of the company on the East Coast, and convertible bonds were an East Coast operation.

One thing that's striking in the book is that the two Drexels were so independent. At a high level, that meant that Milken was running his division with lots of autonomy, but it also meant that his direct reports basically treated the rest of Drexel as some external entity, even a competitor, and were sometimes cagey or outright dishonest to colleagues in other parts of the firm.

Drexel in the 1980s had an incredibly weird setup, where Milken's corner of the firm was not only underwriting and trading as it pleased, but there was also a whole ecosystem of special-purpose vehicles and funds run by Drexel people, sometimes with clients as limited partners. Even by 1980s standards, having the same person manage an underwriting, make a market in the resulting securities, and sometimes opportunistically invest in them via a hedge fund looked pretty aggressive. But it was also an extremely efficient way to accumulate wealth.

The book gives some other glances at 80s financial culture; at one point, an analyst gets chewed out for downgrading a client's stock. This chewing-out partly takes the form of paraphrasing some wisdom from Milken: "There is no such thing as an unhappy corporate finance client. There is such a thing as an unemployed analyst." And it also has some little details that were probably part of every investor's historical model of the world a few decades ago, and have since been forgotten, like the early-80s small-cap IPO boom, with a very SPAC-y set of gimmicky companies (apparently the peak happened around when investors in the IPO of "Muhammed Ali Arcades" realized that Mr. Ali himself was not a buyer).

Drexel is a strange case because it's much easier to prove that they were skirting as close to the edge as possible than it is to definitively connect them to insider trading. But that's why regulated companies that rely on wholesale funding tend to have strict compliance policies: they want it to be incredibly easy to prove that they're trustworthy, because the day their counterparties don't believe them, funding disappears. Drexel could have made more money a bit more slowly had it played by the rules. Instead, they basically created the lucrative business of high-yield debt, dispersed a lot of talent to the rest of the street, and, in the fire-sale days after their collapse, provided seed funding to the next generation of high-yield debt firms by giving them the chance to buy lots of junk bonds at deeply distressed prices. Pretty nice of them, but not the outcome they aimed for.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Some organizations specialize in rule-bending and eventually get in trouble, but there are at least a few cases where they cleaned up their act and turned into well-behaved corporate citizens. Are there any good writeups of this process?

Diff Jobs

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

• 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)
• Fast-growing, General Catalyst backed startup building the platform and primitives that power business transformation, starting with an AI-native ERP, is looking for expert generalists to identify critical directives, parachute into the part of the business that needs help and drive results with scalable processes. If you have exceptional judgement across contexts, a taste for high leverage problems and people, and the agency to drive solutions to completion, this is for you. (SF)
• Well funded, Ex-Stripe founders are building the agentic back-office automation platform that turns business processes into self-directed, self-improving workflows which know when to ask humans for input. They are initially focused on making ERP workflows (invoice management, accounting, financial close, etc.) in the enterprise more accurate/complete and are looking for FDEs and Platform Engineers. If you enjoy working with the C-suite at some of the largest enterprises to drive operational efficiency with AI and have 3+ YOE as a SWE, this is for you. (Remote)

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.