Longreads + Open Thread

Longreads

• Tyler Cowen interviews conductor David Robertson. Of great interest to the tiny number of people who like contemporary classical music, but also notable because being a conductor is the purest distillation of the task of managing a skilled team in a time-sensitive task with hard-to-quantify results. Another good detail is Robertson's claim that good composers have incredibly sensitive pitch. If you're writing a B-flat followed by a B, it's not strictly necessary to know what either of those notes sounds like exactly 5% more sharp or flat. But it's another instance of a low-level skill being surprisingly correlated to a higher-level one that doesn't directly require it.
• In ChinaTalk, a translation of Liu Shaoshan's essay on how China can take the lead in AI. Looking back at some previous national-champion transitions, one shorthand for the auto industry is that cheap, abundant oil built the American car industry and expensive, uncertain oil built the Japanese one. If it turns out that China's comparative advantage is in building models that run on older, cheaper hardware, at some point there may be an overshoot in investment that means that the majority of the world's GPUs are exactly that. The piece claims that the US used its control of Internet protocols as a way to project its culture, which isn't strictly true—TCP/IP does not have some notion of which packets do or do not properly reflect Xi Jinping Thought—but it is true that the US's relatively early adoption of the Internet meant that American cultural norms and the English language became more of a global default. If the best model you can run on a handset in a developing country is unaware of certain events involving Tiananmen Square, but quite well-informed about the Battle of Blair Mountain, that's going to have an effect over time.
• Nilay Patel interviews SharkNinja CEO Mark Barrocas. There's just no end to the stories about how well modern economies reward scale: SharkNinja makes consumer durables, a notoriously tough business because the deployment cycle keeps getting faster and the replacement cycle keeps getting longer, but they can make it work because they're constantly shipping newly-tweaked variants of existing products—there's a funny bit about two thirds of the way through where they talk about how every single time they came up with a new variant on the air fryer, their customers realized they wanted something slightly different so they had to create another one. Once it's a recurring purchase, if not a contractually guaranteed one, SharkNinja has a stronger incentive to support its products, keep updating their apps, make the apps actually useful rather than a gimmick (if your Smart Appliance has an app you actually use, its manufacturer has more information on how you actually use it), etc.
• Austin Vernon has a great breakdown of what contemporary logistics costs look like, and how that can change. Drone delivery, long a Diff interest, is one of the biggest contributors, but we're not looking at a static system. One of the best things about this piece is that it looks at some of the second-order effects: last-mile delivery is incredibly expensive per ton-mile compared to the rest of the process, which means that lowering the cost of the last mile actually raises the relative contribution of the rest of the shipping process, making manufacturing viable in more places.
• Colin Gorrie on why English doesn’t use accents. As it turns out, the reason we don’t have accent marks like the French is that we were copying the French, but copying them before they’d started using accents. It’s a fun way to trace the linguistic lineages. 
• In Capital Gains this week, the question of whether companies should diversify, or their owners should do it for them. Activist hedge funds turn out to be a strong force against empire-building and in favor of more specialized companies and executives.

Books

Lord Jim: It's pretty common to read a book that was published a century or more ago and have it feel shockingly contemporary, so Lord Jim is a bit of a treat: it was published in 1900, but speaks most clearly to the concerns of a few years ago. Specifically, the book is about getting canceled, feeling that you deserve it, and trying to figure out how to make amends to society while preserving your own sanity in the face of crushing guilt.

Our hero is first mate on the Patna, a ship that, around the time of publication, leaves some unspecified place in Asia (probably Singapore) with hundreds of Muslim passengers who are making the hajj. The ship collides with something, the (Western) crew slowly works up enough courage to make the cowardly decision to abandon ship, and at the last minute Jim joins them. They're rescued soon after, and the rest of the crew invents an alibi. And then the ship itself is also rescued, the rest of the crew skips down, and Jim stands trial alone for abandoning his charges.

The rest of our story is Jim bouncing through various ports in Asia, bouncing to a new port whenever he's reminded of his shame, until he settles in a remote island where he can be isolated enough from the rest of the world to actually accomplish something.

Lord Jim is, fundamentally, a book about boundaries. There's a Westernized world of steamships, regular mail delivery, a common language, currency, code of laws, and set of norms. And the ports Jim visits are apertures that connect this world to smaller and more exotic ones, with a mix of languages, expected behaviors, exotic wildlife, and miscellaneous adventure. (One character relates the story of how he survived an ambush and then, in mid-ambush, spotted a particular exotic butterfly hovering over the corpse of a man he'd just shot, so he takes a break from the slaughter to catch it.) There's also the harder-to-permeate border between the narrator, who's trying to understand Jim, and whatever is going on in Jim's head. And the whole story sets off with a very literal instance of a permeable boundary, when the Patna starts taking on water.

This would be a challenging story to tell today. There are plenty of exotic locations you can run off to, but in most of them at least a few of the villagers will have WhatsApp. Ironically, the place where you can flee to reinvent yourself is online. If you were writing the story today, Jim would be the one member of an altcoin team who didn't want to rugpull, did it anyway, got doxxed, and had to spin up a new pseudonym and take another crack at things.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• There was a recent study arguing that when developers use LLMs, they feel more productive but are actually less productive. This clashes with anecdotes I've heard, and with my limited personal experience. (For one thing, my 9-year-old was able to vibecode a little game, and given that she doesn't know how to code at all, it affected her productivity. Then again, hopefully the median full-time software engineer is a better coder than an elementary school student.) But I'm curious to hear from readers about either a) specific ways they found that LLMs helped their work, and b) cases where LLMs flat-out failed to deliver net productivity gains.

Diff Jobs

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

• A Series B startup building regulatory AI agents to help automate compliance for companies in highly regulated industries is looking for legal engineers with financial regulatory experience (SEC, FINRA marketing review, Reg Z, UDAAP). JD required; top law firm experience preferred. (NYC)
• A company building the new pension of the 21st century and enabling universal basic capital is looking for a full stack engineer to help build and scale complex backend financial systems, including trading systems, asset management systems, and ledgers. Experience in fintech or early-stage technology companies preferred. (NYC)
• YC-backed startup using AI to transform how companies quantify and optimize engineering productivity is hiring formidable full-stack and AI engineers. Experience with React + Typescript, Go, or Python on the ML side a plus. (SF)
• Ex-Optiver/DRW quants with over a decade of experience in HFT and AI are reimagining time series forecasting from first principles. They are building a research lab, initially monetized via derivatives trading. The team is hiring a founding engineer (Python/C++/Rust; distributed compute, ML infra) and a founding AI researcher to rethink how machines model the future. No finance experience needed. (SF)
• A well-funded startup that’s building the universal electronic cash system by taking stablecoins from edge cases to the mainstream is looking for a senior full-stack engineer. Experience with information dense front-ends is a strong plus. (NYC, London, Singapore)

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.