If you have a few seconds, we'd appreciate you taking this quick one-question reader survey.
- A brochure marketing Long Term Capital Management. It's good to know what big blowups look like before they succeed, which is—almost exactly what successful companies look like! The brochure says all the right things: that LTCM is smart, careful, and hiring. If there's a hint about what's to come, it's on the last page, where they show a tiny excerpt from a 1996 Institutional Investor article profiling the firm's co-founder and lead partner, John Meriwether: "John is a born risk-taker." The next paragraph is obscured in the image, but it's about how he refused to cut a losing trade because he was convinced that it would bounce back. That’s potentially the optimal choice if you’re trading with a small part of the balance sheet of a big bank, but as a standalone hedge fund, it doesn't work. As the book ($, Diff) notes, volatility yesterday correlates with volatility today, but returns yesterday do not, so when markets get crazy, people with leverage will either dial back their exposure or become a sad case study.
- In Slate, Nitish Pahwa asks: what happened to Quora? It used to be an amazing resource for getting real answers from people who were either experts on a topic or had relevant personal experience. Back in 2010, it was full of oral histories of major startups, but now the quality of questions and answers has trended down so much that the product is almost unusable. (There are, even more annoyingly, still some interesting tidbits!) This is part of the fate of any user-generated site: the thing that makes it grow is that the existing userbase produces better content that attracts a wider audience, and that audience doesn't contribute the same quality. AI can theoretically slow down this glide path if you train on an older cohort of users, but those quality restrictions also limit how much training data there is.
- Matteo Wong at The Atlantic catches up with Neal Stephenson about the latest concepts he floated in science fiction that are now being implemented, including the metaverse and AI tutors. In The Diamond Age, AI tutors exist but are inferior to AI-augmented human tutors, and, as Stephenson notes, some of the same problems he worried about in the 90s have finally happened: "What they do is generate sentences that sound like correct sentences, but there's no underlying brain that can actually discern whether those sentences are correct or not." The right question with many of these tools remains: compared to what? Human teachers hallucinate, too, and one of the meta lessons education imparts is that authority figures are fallible and will sometimes repeat a claim without having either investigated it or considered whether or not it's true. So an AI tutor can't replace a great teacher, but can, on some dimensions, be a satisfactory substitute for a bad one.
- Sometimes it's good to review how commonplace ideas were expressed when they were fresh. In that spirit, W. Brian Arthur wrote a Harvard Business Review piece in the mid-90s on how competition in tech is different from other industries. In more physical industries, there are generally diminishing marginal returns: the next acre you plant and next oil well you drill is probably not as productive as the last one—because if it were, you would have done it first! But the next user of a software platform is often more valuable than the previous one, because they increase the size of the network. Even in 1996, it was clear that this didn't automatically mean everyone should maximize virality: he contrasts Apple's closed approach, which made them too niche, with the IBM PC's more open one, which ended up creating a massive market that IBM couldn't control. (Via the FT's Unhedged ($).)
- Sam Kriss has a damning review of Walter Isaacson's Elon Musk biography. This piece is worth reading in part because it's genuinely funny, but also because it's a good case study in overreacting. Elon Musk has yet to send people to Mars, ships promises faster than he ships self-driving, and tweets intemperately. But there are lots of people who overpromise on multiple timescales, run their mouths too much, and haven't played critical roles in founding two of the most valuable manufacturing companies of all time. At some point, even a relentless critic has to concede that if Musk isn't extraordinarily talented, his luck is good evidence that a) the universe is a simulation, with Musk as the player character and b) the simulation has numerous cheat codes. Occam's Razor suggests a more boring story, where Musk is talented but flawed, and where those flaws are interesting by virtue of being attached to someone who did, in fact, popularize electric vehicles and reverse a generation-long stagnation in space exploration. It's okay not to like the tweets.
- In the fourteenth episode of The Riff, we talk about media history, building a media company, and the distribution/content tradeoff. Listen with Spotify/Apple/YouTube.
- And in Capital Gains, we're covering the true size of the "bankroll" in discussions of betting strategies. It's hard enough to measure edge and odds, but you also have to know how much you're actually willing to lose.
Planning for Empire: Reform Bureaucrats and the Japanese Wartime State: The early twentieth century was a Cambrian Explosion of political ideologies, most of which did not survive. That change can be read as partly a reaction to the collapse of monarchies within Europe and the decline of colonialism outside of it, but there was significant variation in how different places and institutions responded. Sometimes the family tree gets messy: the "reform bureaucrats" were an odd coalition of Marxists, fascists, admirers of the New Deal, and amorally ambitious political operators.
The odd coalitions did not just show up at a high level, but were fractally present at every level. For example, Japan's economy was dominated by zaibatsu, large networks of holding companies, often centered around a bank and trading company. The older zaibatsu were largely in favor of free markets, because they made so much money importing and exporting goods. The newer ones often cozied up to the government, both because they could get state subsidies and loans for heavy industry and as a competitive differentiator.
This era also showcased a bizarre form of federalism: Japan tested out some of its industrial policy and government plans in the puppet state of Manchukuo. It's easier to impose reforms at home if they've been demonstrated elsewhere, but no one was pursuing quite the model Japan aimed for, so they incubated their own policy laboratory. This was in one sense a decentralized governance model, but in another sense was an invitation to experiment with looser conceptions of human rights.
There's surprising continuity, both in terms of personnel and policy, between pre- and post-war Japanese economics. In both periods, the country wanted to secure access to raw materials and a market in which to safely sell finished goods. It turned out that this was not really possible with the Japanese military, at least once they a) didn't have access to American oil imports, and b) were, in fact, fighting America. But the US Navy was able to accomplish what the Imperial Japanese Navy was not, and to secure Japan's status as a modern, advanced economy.
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- A company building the new pension of the 21st century and building universal basic capital is looking for a frontend engineer. (NYC)
- A fintech company using AI to craft new investment strategies seeks a portfolio management associate with 2+ years of experience in trading or operations for equities or crypto. This is a technical role—FIX proficiency required, as well as Python, C#, and SQL. (NYC)
- A successful crypto prop-trading firm is looking for new quantitative developers with experience building high-performance, scalable systems in C++. (Remote)
- A well funded seed stage startup founded by former SpaceX engineers is looking for full stack engineers previously employed by Anduril or Palantir. (LA)
- A data consultancy is looking for data scientists with prior experience at hedge funds, research firms, or banks. Alt data experience is preferred. (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.
- Drop in any links or comments of interest to Diff readers.
- So far in 2024, the investment story has lined up with 2023: Magnificent Seven outperforming, smaller AI stories hit-or-miss, less excitement elsewhere. What changes this? Will the Mag 7 reverse together, or will we revise them to the Magnificent Six?