Longreads + Open Thread

Longreads

• Bloomberg's Nishant Kumar profiles D.E. Shaw. Investment businesses often start around one person, and try to scale them in as many directions as possible; in a discretionary strategy, you might have one portfolio manager who starts out mostly doing research, and then slowly adds analysts whose ideas they can vet faster than they could recreate them. This eventually runs into a mix of problems: it's hard to scale organizations indefinitely, and this implicitly locks a firm into a specific style. Other funds are run, from early on, as institutions, which implicitly means that their goal is to be able to survive 100% cumulative workforce turnover without falling apart. That seems easier to do in quantitative strategies, because 1) you can overlay many of these on top of each other (applying an existing signal to new asset classes provides some diversification, for example, and short-term signals help reduce the entry and exit cost of longer-term strategies), 2) hiring people out of academia means hiring people who've spent much of their life in institutions that are greater than the sum of their parts and are expected to outlast all of their key personnel, and 3) hiring academics at least used to mean hiring people who didn't expect to make a lot of money and are delighted that their puzzle-solving and puzzle-selecting skills are worth an order of magnitude more than they thought.
• Tyler Cowen interviews Richard Hanania, mostly on the topic of Joseph Conrad. (This was reviewed in The Diff during the era of very short Diff book reviews. It's much easier to read negative than positive judgments into Conrad's work: he likes looking at complicated people who find themselves in miserable situations and then telling us to look at how complicated and miserable everything is. That negativity means you can read him as a critique of late colonialism, but also as someone who's deflecting blame from it because the problem with human institutions will always be the humans. One interesting point they make is on state capacity and human rights: the countries that have high murder rates coupled with reasonably high GDP per capita tend to be democratic countries in South America, and the hypothesis is that in a weak state that still tries to respect human rights, cocaine traffickers will be able to hire better lawyers than, say, a guy who got in trouble for selling produce by the side of the road without a permit. So they regulate innocuous small businesses, but the really dangerous ones regulate themselves. A little less respect for human rights would mean a much shorter lifespan for drug lords, followed by higher economic growth, followed by a more gradual reintroduction of human rights. On the other hand, it could be the case that trying to do that just leads to the cartels taking over entirely, and the real problem might be that it's possible to run a lucrative export industry with relatively simple supply chains and only indirect interaction with the formal banking system. Conrad would have had a wonderful time writing about cartels.
• Johannes Boehm and Thomas Chaney have written a paper I vaguely hoped someone would write, using deposits of coins to measure changes in trade. The intuition is that coins get minted somewhere and then slowly radiate outward as they're repeatedly spent, so you can measure how closely different places were connected and how the shape of trade routes changes just by looking at which coins ended up where. There are many caveats to this (for example: coins tend to get buried by people who are worried that their wealth is going to be expropriated, but who also think that that's a temporary risk and they'll make it home. And the coins we have are the ones that belonged to people who guessed wrong). The results line up with the thesis that the early Muslim conquests cut Europe off from Asia; Pliny had complained that "by the lowest reckoning India, China and the Arabian peninsula take from our empire 100 million sestertii every year," so Europe had to develop more sources of domestic luxury items. So if you really squint, it's an example of a de facto tariff leading to a near-term reduction in economic output, followed (in this case centuries later) by growth in domestic manufacturing.
• Ryan Mac, Kalley Huang, Cade Metz and Georgia Gee in the NYT profile Dario Amodei. Sometimes you'll read about some historical scene like Vienna in the 1910s or Paris or Harlem in the 1920s and wonder if that kind of thing still exists. But the people who lived through it didn't necessarily know they were part of some eventually-famous scene. As it turns out, Dario and I were part of the same formative scene circa 2007, which was harder to see because it existed in Google Reader. The mid-2000s/early 2010s blogosphere turns out to have been incredibly influential on the world, contributing to political weirdness left and right and seeding the modern AI industry with much of its worldview. But this also illustrates why a scene alone is not enough. You probably can't name the fiftieth most-famous person who visited Café Central along with Stalin and Freud. Similarly, many people, myself included, read Overcoming Bias and thought it had some neat ideas. A few people read it and decided to dedicate their lives to building omnipotent AI before someone else did it the wrong way.
• In NYMag, Charlotte Klein profiles Charles Porch, who works for OpenAI but has the very un-automatable job of being friends with as many celebrities as humanly possible and convincing them that AI is good. A good neo-Malthusian outlook is that the more effective an organization is in one domain, the more it succeeds to the point that all of its problems are in some other domain that it doesn't have experience with, and it needs to hire someone who is from well outside its culture. It helps that Porch had previously had a similar role at Instagram, which was a little more closely connected to what celebrities care about, inasmuch as it let them solve the key problems of "I wish people were paying even more attention to me" and "I hate that picture of me and wish I could have picked a better one." But networks that get built up in one context can end up useful in another.
• In this week's Capital Gains: reporting news has almost never been a good business on its own, except for financial news. A big reason for this is that influencing the news pays off well in non-financial terms, whether it means getting invited to the right parties or being able to influence politics (both of which can confer large financial rewards as a second order effect).
• A Read.Haus reader notes that pre-IPO prediction markets on late-stage companies have a fair amount of liquidity, and wonders how early that will go. The boring answer is that we'll probably have some change in prediction market regulation that makes this less likely, and that the risk of this means that some informed investors have an incentive not to spend much time with these markets if they're not going to last. But, setting that aside: going public gives companies the advantage of raising capital, having equity as acquisition currency, and making employee equity liquid. The disadvantage is constantly leaking information and having morale-destroying share price declines. For companies with bimodal investor sentiment (which includes both big, bold ideas and total scams), being private means that short sellers don't impact the market, so their valuation is higher. So what will probably happen is that companies will try to restrict their employees and investors from betting on their equity value through prediction markets, and that smaller ones will be too illiquid to attract much attention. For a given company valuation, you probably get an order of magnitude more liquidity if the company regularly files 10-Qs, and you don't run as much risk that your counterparty when you buy is a well-informed seller.

Books

The Discrete Charm of the Machine: Why the World Became Digital: A few years ago I was thinking about film cameras, and about how over-the-top insanely nerdy it would have been to come up with them if digital cameras had been invented first. Converting things into bits isn't trivial, but once you have them in bit form, there's just no end to the tricks you can play and the other bit-flipping domains you can borrow from. But for analog processes, you're limited to what the physical world gives you, and you have to treat reality as an escape room designed by the demiurge where completely random materials you find happen to have indispensable properties for some application. (There's something similar in materials, where thousands of different use cases either get collapsed into "you should turn oil or coal into whatever kind of plastic has the right properties for the job" or "have you considered using iron mixed with carbon and a little bit of some other metals? Usually does the job for me.")

This book might be described as a love letter to analog, or even a kind of insensitive breakup letter. Reality is analog, at least at the level at which we experience it, but the analog world offers arbitrary precision in exchange for decaying accuracy—magnetic storage slowly loses its magnetism, records get slightly damaged each time they're playing, movie reels can accumulate wear and tear (or, if they're nitrate, they can violently burn). Digital data can suffer the same kind of decay, in the sense that solar radiation flips bits and the like, but it can be copied bit-for-bit, and checksums can be used to identify and repair damage. You don't have to worry that the YouTube video you watched dozens of times will start getting grainy. (Though one of the perils of aging is watching a YouTube video from, say, 2006 and realizing that your memory upscaled it and it used to look terrible.)

The natural thread for a book about why so much of the world is digital is to go step by step through how to make analog signals digital instead. And that's the approach this one takes: we look at how you can use arrangements of valves to do computations and store data, then find a particularly charming solid-state valve, the transistor. (The book gives some of the quantum physics background on why they work, without a lot of math—incidentally, one fun effect of AI is widening the skill range at which a book like this is useful. You wouldn't feel too bad about giving it to a smart elementary school student who knew how to send a screenshot to ChatGPT and ask for a more detailed explanation, but you could also give it to a STEM undergrad who'd use their chat app to formalize some of the qualitative descriptions and add some practice problems at the end of each chapter.)

It's interesting to revisit the sweep of information technology history as a long process of getting better at processing digital information, increasing the returns on digitizing everything. It's also a good way to make the informational world look as weird as it truly is: the author notes that you can "store" Beethoven's 9th by transmitting it in a continuous loop; it stretches from earth to Saturn and contains one byte per mile.

One last striking thing: the author mentions is right at the beginning: he was born in 1939, and the book was published in 2019. Throughout it, the author's enthusiasm bleeds through: this guy just really loves digital signal processing! When he provides a photo showing the progression of vacuum tubes technology over time, the source for the photo is "author's garage." May we all be that excited when we're eighty.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Please share all of your complaints about the Diff redesign. It is, of course, vibecoded, which means that a) it looks more like other vibecoded websites, and b) its cost to me, and my emotional attachment to it, are both in the low single-digit dollars range.

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 private equity investors to help lead leveraged buyout co-investments in healthcare services, financial services, and business services. If you’re excited about underwriting AI value creation and have experience leading traditional LBOs (financial,  commercial diligence, etc.) please reach out. (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)
• 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 (Remote)
• 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.