Longreads + Open Thread

Longreads

• A classic story: the accidental HFT firm. This blog post tells the tale of launching and running a prop trading firm from 2005 to 2011. What's notable about the story is how full-stack it is: the strategies range from using machine learning to predict futures prices to physically fiddling around with T1 lines to get data slightly faster. A good model here is to be a specialist in terms of outputs and a generalist in terms of inputs, because only a generalist will know whether the smartest tradeoff involves writing code, negotiating a better deal with a broker, or soldering something.
• James B. Stewart has a story in the NYT on undercollateralized futures trades, intermingling of customer and corporate funds, and a financial death spiral involving ever-slower and less reliable customer withdrawals. (It's about a wine store.) It's also a sad story about a well-run business that was later run into the ground. The store in question had been selling "wine futures" since the 1950s, and would sometimes cover its naked short position in promised wine by buying it at retail and delivering it to customers. Leverage can be powerful, and for a business with significant inventory needs and low turnover, it's essential. But too much of it is toxic.
• Jens Nordvig has a brief history of dollar hatred. Part of what this piece emphasizes is the counterintuitive economics of reserve currencies: the financial crisis was centered on the US, and hurt the US economy, but caused the dollar to rise, both because global borrowers were struggling to service debt primarily denominated in dollars and because American economic weakness reduced US imports.
• Brian Chau looks at the pace of ML improvements and the likelihood of diminishing marginal returns. The core argument is that improvements in underlying hardware have occurred alongside low-level optimizations. Linear progress usually breaks down into a series of S-curves, but when one S-curve is steeper—like a closer match between the tasks at hand, compilers, and the hardware used to address them—there can be a one-off spurt of productivity followed by a deceleration.
• In Politico, the publisher of McSweeney's, who grew up working class, finds out that she's the owner of a one-eighth interest in the mineral rights of some Marcellus Shale land, and spends a while exploring the idea of unearned mineral wealth, the local and global disruptions of the hydrocarbon economy, and the question of whether to take the money upfront or roll the dice on long-term royalties. In a way, it's a metaphor for the first world experience in general: we've all inherited far more abundance than we've personally worked for, and inherited some share in the mistakes that led to that wealth.
• In this week's Capital Gains, we look at the mechanics of short squeezes, and how they're not just restricted to financial markets: "The world has periodically gone through oil short squeezes, for much the same reason: if oil prices go up 5%, it's extremely hard to respond by consuming 5% less oil."

Books

• No Country for Old Men: Beautifully written, impossibly grim. No Country for Old Men is a story about evil narrated by someone who is wrestling with the problem of evil: a man stumbles on the aftermath of a drug deal gone wrong, leaves with two million dollars in cash, and suffers endless consequences for it. Part of the point of the book is the unpredictable consequences of a single decision. But part of it is the all too predictable escalation that can ensue (the money is discovered by a character who was out hunting at the time; we later find out that he was a sniper in Vietnam. Violence is contagious!)
• White Shoe: How a New Breed of Wall Street Lawyers Changed Big Business and the American Century: this book talks about the rise of big law firms in the late 19th and early 20th centuries. The firms it talks about are top firms to this day. There are two good ways to read this book: one is as a story about the late gilded age told mostly through the legal conflicts that ensued, some of which went quite far (the "Cromwell" in Sullivan & Cromwell was intimately involved in the coup that separated Panama from Colombia). It's also the story of a craft business becoming institutionalized; law firms got bigger and more effective, but lost something in the process.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• A broad category where AI performs at parity with humans, but with lower cost, is in dealing with slightly messy data. An LLM is far better than a regular expression at handling typos, for example. And physical goods are much messier than digital ones. What physical products will be the first to be affected by AI?

Reader Feedback

Last week's piece on whether or not central planning can be accomplished with modern computing resources (with big tech companies' use of third parties as the main evidence against) got lots of interesting responses in several media. One of the main questions that came up comes back to legibility. The map is not the territory, but a map does help with navigation. The determining factor is not the quality of the map, but an accurate perception of its accuracy.

A Word From Our Sponsors

I’m proud to announce this week’s newsletter is sponsored by Compound.

Compound is personalized wealth management for people in tech. It’s how people at companies like Stripe, OpenAI, and Discord make smarter decisions about their money.

I actually know Jordan, Compound’s CEO, and wrote a piece for their site a while back about equity and options. The company is built and run by people who deeply understand the financial decisions you’re likely thinking about — and who are equipped to help you make those decisions.

Compound focuses on your finances and taxes, so you can focus on what matters most.

Diff Jobs

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

• A vertically integrated PE-backed cannabis company is looking for a production analytics manager to optimally allocate plant biomass. Excel wizards preferred. (Little Rock, AR—no remote, but relocation assistance is possible)
• A VC backed company reimagining retirement wealth and building a 401k alternative is looking for a product manager with fintech experience. (NYC)
• A company that helps investors use alternative data to make better decisions is looking for early-career data scientists and business analysts. (Remote)
• ​​A well funded seed stage startup founded by former SpaceX engineers is building software tools for hardware engineering. They're looking for a UX/frontend engineer interested in designing and developing software collaboratively with satellite, rocket, and other complex machine engineers. (Los Angeles)
• A profitable startup is looking for sales reps to market its AI-based services that help small companies accelerate their growth—especially people who are excited to use AI tools to accelerate some of their own work. (SF)

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.