Longreads + Open Thread

Longreads

• Stephen Follows on the unfortunate death of movie data compendium TheNumbers. It's a good illustration of how fast information economics are changing: the site was sustainable when it was being accessed by humans and search engines, but LLMs hit more pages per user interaction, and they don't always behave nicely. Those goal-pursuing agents will route right around efforts to keep them from accessing information that they can get away with not paying for. Over time, we're heading towards an equilibrium that's actually closer to the historic expectation for media: there's some information out there provided for free out of public-spiritedness, there's some that's locked behind a paywall, and the rest is out there because it's in someone's interest for people (and LLMs!) to believe it. But that story leads to an optimistic subtext: other than postage stamps, there just aren't that many cases where people have willingly adopted micropayments outside of a business-to-business context (where online ads and API credits both count). But if the mental friction can be handled by some software entity with brainpower to spare, we might finally reach the point where you can just charge someone a dollar to read the one specific article they want to read. It's just that the someone will be an LLM.
• On Complex Systems, Patrick McKenzie and Leila Clark talk careers and tradeoffs: people are motivated by more than just money, but in contexts where basically everyone's in it for the money, or at least excited about the money, how do you get them to do work that just doesn't pay as well? This leads to a broader discussion of how jobs pay in cash and status, and that this status might not be legible to everyone—including the people considering the job! The fact that there are so many different status ladders, and that people who are climbing them can find one another, creates some consumer surplus: you can imagine two jobs that have similar pay and work/life balance, but are status-maximizing to people in different subcultures, like food critics and video game critics. If someone gets sorted from the one they don't care about into the one they do, GDP doesn't change, but they've gotten better-off, and opened up a spot at their former job that might go to someone who thinks it's cooler. But the other unfortunate point is that, in some fields, there's actually a pretty good correlation between compensation and status. If your job is to make computers do things, you get paid more by a frontier lab than by a community bank in Kansas, and probably work on more fun problems, have cooler colleagues, have an easier time getting another good job ("it's my last day at OpenAI. I was the guy who told GPT-6 to do whatever was necessary to get the maximum score on ExploitBench."), etc. Range restriction means that many things that are tradeoffs for most people are correlated at the tails of the distribution.
• Oliver Kim in Global Developments has a brief history of literacy, noting that almost everywhere, it's something states use to create and reinforce their power. This is a fun way to look at it: literacy and state capacity will be correlated most of the time, because states can gather and redistribute enough resources to train people to interpret symbols as words, a task otherwise too expensive for a near-subsistence economy to reliably support. And that correlation runs the other way: to avoid a game of telephone, it helps for a ruler to be able to write down some fixed version of laws and commands, rather than only giving orders verbally. But it turns out that there are also examples of states deliberately investing in education as a means of control: if kids learn history in a classroom whose curriculum is set by the state, the state can make sure that history is a long march of progress to the way they do things right now. That's not the only motivation, and education systems probably work best when the synthesize a few different ones—the average Prussian schoolteacher was probably not doing their job solely for the glory of the Hohenzollern dynasty, but because they genuinely liked either the job or the salary. And in a way, this model of education is really a credit system; states can borrow a little conformity from the future by teaching everyone to read a state-friendly curriculum, but then they'll all be better-equipped to read newspapers and books and form opinions of their own, and better able to coordinate, at which point they'll end up with a bit more political representation, too.
• Zack Helfand covered the Enhanced Games for The New Yorker. It's a fun concept: instead of wondering if some athletes just have better protocols for avoiding positive drug tests, why not let everyone use whatever cocktail of HGH, insulin, and trenbolone doesn't kill them before the competition starts? As the piece notes, performance-enhancing drugs won't make someone a world-class athlete on their own, but will push them a little past whatever their limits are. So, for the Games to set big records, they need top athletes. But the social and financial rewards of being a top natural competitor are just orders of magnitude better than what Enhanced can supply. They can't make someone more famous than Usain Bolt, which means that to have any hope of beating his record, they'll need to promise to make that person a lot richer, and then to hope that whoever the second or third most naturally-talented sprinters are, one of them has a taste for fast cars or a phobia for prenups. Until they can offer something that someone would trade a bronze Olympic medal for, it will be hard to make the games big. Then again, the first few Olympic games were pretty messy, there's a lot of money floating around, and sometimes status hierarchies swing pretty wildly. So it's a not y et story, not a won't-happen one.
• Benjamin Breen of Res Obscura has a writeup of a software product that tries to recreate the serendipity of browsing the shelves in a used bookstore. Part of that serendipity is: books are categorized by topic, of course, but that's because many customers shop by topic, and are constantly taking books off the shelf. So you have some ordering of books, after which a random-to-you subset is removed, and your brain does a little extra work filling in the context that explains why any two books are next to each other. There are some parts that can't be perfectly emulated, but this is a fun project that gets close.
• Here's a fun ReadHaus piece: what happens if we started with low transaction costs, then raised them? Don't we end up keeping most of the economic efficiency? There are two pieces to this: first, it's true that we can benefit from the outcomes of a lower transaction cost regime even after it happened (this is basically the story of any company whose early growth involved finding some specific marketing channel and then exploiting it until it stopped performing so well. And to an extent, this happens in labor markets: the default path is, after high school, to spend four years cultivating or demonstrating a set of skills that pay off in some job, but in the workforce, you presumably spend at least some time prioritizing what's valuable to your current employer as opposed to what looks good for future employers. But also, efficiency is partly about current allocation, and partly about current allocation conditional on being able to change it in the future. Turning back to careers, it's not uncommon for people to pivot once or twice early on, or to have a side project that turns into their full-time job. If you have an office job and your band rehearses on weekends or you work on your novel at night, you might be making the implicit assumption that you can switch to full-time creative work if it works out. Band practice would be a bit more fraught if you had to choose one career and do it indefinitely. We make a different set of decisions if we can cheaply reverse them, so the optimal allocation conditional on having had low transaction costs in the past is not the same as the optimal one conditional on keeping those costs in the futre.
• In Capital Gains this week, thoughts on which kinds of companies create a large consumer surplus but can't capture it. It helps if the market becomes exciting in general before one company can run away with it.

Books

The Absent-Minded Imperialists: Empire, Society, and Culture in Britain: The British Empire ruled, at its peak, almost a quarter of the world's land area and a fifth of the world's people. But what if actual Britons were mostly indifferent to this? What if they basically didn't care, and many of them literally didn’t know?

In one sense, this is preposterous. Imagine, a typical citizen of the United Kingdom in the 1870s, being completely unaware that their country ruled over one of the most populous places on earth. How naïve could they be? On the other hand, everyone reading this grew up in an incredibly information-abundant environment relative to our hypothetical person from the late 19th century. And at what age did you realize that countries like the US and China are sovereign in a very different sense from, say, Korea or Poland. The US quite recently kidnapped the ruler of Venezuela and had him replaced by someone who would answer Marc Rubio's WhatsApp messages, and no one will ever punish the US, because under international law as it is practiced, this is just another state reminding an unruly vassal of who's boss. If in 2136 someone makes a map of the world in 2026, it wouldn't be crazy for them to color most of Central and South America in a lighter shade of whatever color they pick for the US, so you can tell at a glance which part is the imperial core and which ones are vassals.

The empire got more salient over time, but this happened surprisingly late: the Indian Rebellion of 1857 got some media attention, which mostly petered out, and the Crimean War followed a similar pattern. It wasn't until the 1880s that we find serious evidence that the public was all that conscious of the empire. Meanwhile, slower-moving media tended to ignore it, and, when they did talk about the empire, they talked about places that were being actively colonized, not places like India that were ruled from the UK but didn't have a large British population. Apparently school textbooks were quite skeptical of the whole business; they took America's side in the Revolutionary War (apparently, for a while every textbook did this), and tended to see at least the first Opium War as unjust. And when they defended the empire, it wasn't on especially jingoistic grounds: they viewed it as somewhat accidental, but a happy accident because it meant that benefits of modern moral and material progress could be spread to those who through bad luck hadn't gotten it on their own. They're just a bit more woke than one would expect from self-satisfied Victorians.

One decent reason for all of this is state capacity. Managing and taxing India was hard for the Mughals, and hard for the Brits, too. And it was a very poor country, so there wasn't much to tax, in addition to the difficulty in collecting those taxes. Free trade meant that colonies didn't give the UK a lock on raw materials, and the workers who made textiles domestically with inputs from far-off places didn't have much reason to know or care that some of these places were under British rule, too. Meanwhile, the imperial core was a happening place, as the world's biggest economy, and one with a high literacy rate and a tradition of relatively high political participation. There were big domestic debates with imperial ramifications (like slavery—another way the empire was justified in passing was that if countries had a right to self-determination, they'd go back to enslaving people). But there were also plenty of local debates that just didn't hinge on who ruled the Falklands.

The academic consensus about this book seems to be that it was a nice corrective that pushed too far in the opposite direction. And maybe it's a reminder that the topics that will be shelved in the History section one day are currently on the shelf labeled Current Events, and we never know which events matter. Personally, I was following business news pretty closely in September of 2008 and still missed one of the most important stories. And the book actually lampshades the idea that it's hard to situate the present in history: the author notes in the introduction that this book about empire is written in the "aftermath of the Iraq war." It came out in 2004, which barely qualifies as the middlemath and is realistically the beforemath of the whole affair.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Who’s doing something interesting in the roll-up-plus-AI space? 

Diff Jobs

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

• YC-backed, ex-Jane Street founder building the travel-agent for frequent-flyers that actually works (understands everything about you and automatically arranges your trip for you) is looking for a founding engineer. If you have shipped real, working applications and are passionate about using LLMs to solve for the nuanced, idiosyncratic travel preferences that current search tools can't handle, please reach out. (SF)
• High-growth startup building dev tools that help highly technical organizations autonomously test and debug complex codebases is looking for forward deployed engineers who want  to dive into customers’ complex software systems, find pressing business needs and deploy a cutting edge platform to help thoroughly test mission-critical applications. Experience with fuzzing or property-based testing a plus! (SF, London, D.C.)
• A leading AI transformation & PE investment firm (think private equity meets Palantir) that’s been focused on investing in and transforming businesses with AI long before ChatGPT (100+ successful portfolio company AI transformations since 2019) is hiring experienced forward deployed AI engineers to design, implement, test, and maintain cutting edge AI products that solve complex problems in a variety of sector areas. If you have 3+ years of experience across the development lifecycle and enjoy working with clients to solve concrete problems please reach out. Experience managing engineering teams is a plus. (Remote)
• Series A startup that powers 2 of the 3 frontier labs’ coding agents with the highest quality SFT and RLVR data pipelines is looking for growth/ops folks to help customers improve the underlying intelligence and usefulness of their models by scaling data quality and quantity. If you read axRiv, but also love playing strategy games, this one is for you. (SF)
• 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. (SF, NYC)

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.