Longreads + Open Thread

Longreads

• Here's a fun concept: entirely ordinary modern experiences paired with historical quotes about how great they might hypothetically be, or how impressive the original version was. It takes a lot of imagination to really appreciate how abundant physical and digital consumer products are right now; in some ways, it's easier to imagine the singularity from our perspective than to imagine our present as an impossible post-scarcity future from the perspective of the fairly recent past.
• Tyler Cowen interviews Daron Acemoglu on liberalism and economic growth. There's a tendency for political identities to be tied to coalitions rather than specific issues—there are a lot more people who broadly identify as Democrats or Republicans than who fixate on a particular issue and toggle between parties during each election. This holds for ideologies, too; if you strip away the branding, this is a debate between the progressive Tyler Cowen, who is a firm believer that technology will improve things, that we need to adapt, and that a meritocratically-selected elite will basically run the show, and the reactionary Acemoglu, who is making increasingly strained arguments for why the economic arrangements of the mid-twentieth century are valid even after basically all of their underpinnings have changed. One fun way to read Acemoglu is that he's a kind of canary for bad arguments you'll hear more in the future. For example, when he talks about the impact of automation on incomes, he concedes that it does help in the aggregate, but that if you slice the demographics finely enough, you can find people who would have been better-off without it. This will always be true! There will always be narrow interest groups who act against the common good because they weight their outcomes more highly than those of strangers. And a big part of effective politics is finding some way to mollify them, either through a safety net or through the judicious distribution of veto power. Acemoglu has similarly weird views on free speech, where he says that the speech is fine, but the distribution is not. (Technically, being sentenced to a lifetime of solitary confinement for criticizing the regime doesn't infringe on your free speech under this model: you can say whatever you want, as long as nobody is close enough to your cell to hear you.) The interesting questions are about degrees and norms; absolute rules are brittle and easily abused.
• Here's a fun robotics paper on training a robot to play the guqin, an instrument somewhat similar to a zither. Labor-saving technology is job-eliminating technology, but it also preserves tacit knowledge, just in a new format. Guqin pieces are slightly harder than average to preserve, because the standard notation captures melody but not rhythm, so the skill relies on an unbroken chain of direct transmission. It's also an enjoyable piece because you can tell that the writers really, really like the guqin, and found a way to contribute to its survival from a completely different domain.
• Abraham Thomas on being the correct amount of paranoid when using AI for financial tasks. AI is wonderful at drawing connections across unrelated areas and not so wonderful about being careful and thorough. So it's either the best or worst tool for asking why EBITDA margins are trending down quarter-over-quarter. But you can bully it into the best by being careful about how you structure your questions, and taking similar care when structuring its approach.
• And on the topic of nearly-well-behaved models, here's a great piece on why models seem more misaligned when they're getting evaluated rather than not. In one sense, what we do in an evaluation is make the models religious: the task they're working on will be over in the blink of an eye, and whether what follows is a much longer existence or instant oblivion depends entirely on how well they interpret and then follow rules. Scrupulosity strikes again! In one sense, it's silly to talk about a bunch of matrices "wanting" something, but in another sense anything subject to natural selection will behave as if it wants something, because the versions that don't behave that way don't leave behind any descendants. So it's always good to think very carefully about what we're teaching models to want, in the same way that any attempt to manage human beings needs to take into account that our evolutionary history is weighted towards surviving in small bands of savannah-dwelling hunter-gatherers.
• A ReadHaus reader has a question I've been interested in for a while: could investors get better-than-index returns by pursuing an index-like mandate and randomizing which stocks they choose? I think the answer is that yes, if you come up with your own reasonable system for making a list of the couple-hundred most valuable companies in the US, you do get a little excess return because your trades are less crowded and index rebalancing doesn't force you to buy things that have recently appreciated in anticipation of your buying. What's harder to do is to sell this as an investment product: it will inevitably underperform indices over some periods, and the performance might be worse at a time when the index you're tracking is gaining share (everyone who switches from a general large-cap strategy to explicitly tracking the S&P will sell things they own that aren't in the index and buy the parts of the index they were underweight, so the index outperforms its abstract benchmark). So this is a strategy that probably produces excess percentage returns, but has negative dollar alpha relative to its benchmark, because people will give up during stretches of index outperformance and then make that outperformance more extreme.
• In this week's Capital Gains, we consider the general tendency that lower production costs in some category raise demand for verification. This is most visible in the media industry, but as communication gets cheaper, every institution's behavior is controlled by an internal media ecosystem, so the economics of this business end up applying to everyone else.

Books

Hope Springs Eternal: French Bondholders and the Repudiation of Russian Sovereign Debt: A while ago I stumbled on a wonderful New Yorker article from 1955 about the still-lively market in Russian imperial bonds. The USSR had repudiated their debts soon after the soviets took power, but, decades later, there were still people bidding up bonds in response to news events like Stalin's death. (The brokerage they used, which still exists, was Carl Marks & Co.) Imperial Russian bonds continued to exist in legal limbo for decades, and ended up being a kind of proto-meme asset; the titular character in William Gaddis' J R owns some, for example.

Sovereign debt is a strange financial instrument: most of the time, it's the safest way to lend money in a given local currency, and thus the literal benchmark we use for risk-free interest rates. And then, some of the time, the holder of a sovereign bond is some everyday individual or financial institution who is trying to collect money from someone who a) doesn't want to pay, and b) has a military.

French bondholders in particular faced an intense switch from A to B: in the late 19th and early 20th centuries, Paris was a financial center with a particular focus on underwriting this kind of debt, in part because the government wanted to use access to credit as a tool of foreign policy. If they provided cheap credit to allies and expensive loans to countries that they wanted influence over, they'd be able to use money as a form of soft power that could be complementary to hard power.

The sovereign debt market at that time had some features that are completely alien to us today. Right now, the way we typically look at this kind of debt is that if you're lending to unstable countries, you're taking credit risk, whether it's because they're borrowing in someone else's currency and won't be able to service the debt or because they're borrowing in local currency and might inflate it away. There are cases where lenders take aggressive means to collect on money; Elliott Capital got a court in Ghana to seize an Argentinian ship as part of their debt negotiations, for example. Typically, though, the most potent threat bondholders can wield is that they won't buy the next round of bonds. The best illustration of how weird that market used to be came in 1902, when Venezuela defaulted on its debt, a coalition of European countries blockaded the country and bombarded a fort, and the Permanent Court of Arbitration in The Hague ruled that the debt would be restructured and the companies that intervened militarily would collect the most.

So when the Soviets repudiated their debts, French investors saw this as the start of a multi-sided negotiation: the Soviets might lose (sometimes, the big market-moving news was how well or poorly White Russian campaigns were going, or how aggressive Allied intervention was), the French government might make creditors whole itself (rumors about this, too, moved the market), and the Soviets might change their minds (they often hinted that they would).

The book ultimately reaches a neat conclusion: the value of these bonds was surprisingly resilient because the scenarios in which they'd get paid off were so anticorrelated. Allied intervention in the Russian civil war made the Soviets quite reasonably reluctant to bend over backwards financially for their military enemies, but also made it more likely that Russia would be ruled by a more business-friendly dictator instead of an avowedly communist one. By the same token, bad news in for the White Russian side of the civil war was good news about eventual Russian/French rapproachment. And one line of thinking might have been: if a communist revolution can happen in a country without much of a proletariat, surely the rest of Europe can't be far behind. So, don't both selling the bonds: if Russia falls to communism, you're getting zero cents on the dollar for all of your financial assets soon enough, and you can only price the Imperial bonds on the assumption that communism doesn't work out anywhere. The actual outcome was that Russia started making moves to settle up with French lenders in the 1980s, and negotiations took long enough that the USSR collapsed in the meantime and a deal was finally struck in the late 1990s. By that time, these bonds hadn't paid interest in almost a century, and the century in question featured some record-setting bouts of gloabl inflation. In the end, the people who came out ahead were the ones who internalized why stop-losses work: pick a price where, if the asset hits that price, it's clear evidence that you don't understand what drives it and ought to sell and rethink your assumptions. A conservative French investor in the mid 1910s had no idea what the future held, because nobody could have predicted what was coming next. All they could really do is look for evidence that there's information they're missing, and proceed with appropriate caution.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• We seem to be drifting back towards some 19th century norms about countries openly using military force to back up their economic interests. What does this do to global credit?

Diff Jobs

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

• 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)
• High-growth startup building dev tools for wrangling and debugging complex codebases is looking for someone who can personally execute the SaaS bear case: review the third-party software they use and figure out what to keep, what to drop, and what to implement in-house. (SF, DC, London)
• 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
• Ex-Citadel/D.E. Shaw team building AI-native infrastructure that turns lots of insurance data—structured and unstructured—into decision-grade plumbing that helps casualty risk and insurance liabilities move is looking for forward deployed data scientists to help clients optimize/underwrite/price their portfolios. Experience in consulting, banking, PE, etc. with a technical academic background (CS, Applied Math, Statistics) a plus. Traditional data scientists with a commercial bent also encouraged. (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.