Longreads + Open Thread

Longreads

• Vincent Bignon, Benoit Mojon and Miguel Ortiz Serrano on the liquidity squeeze of 1805 after Europe was cut off from Spanish silver. This is a very fun piece, because it has some of the elements we'd associate with a more modern financial crisis—a fragile shadow banking system that faces a sudden liquidity shock, leading to a long period of scarce credit and low growth—but in this case it happened in a financial system that was based on precious metals. Not only that, but the English financial system survived wartime stresses in part by being more willing to sever the link between paper money and metals, which is exactly the kind of policy that we'd associate with financial systems prone to this kind of collapse. Different financial systems can formalize things in different ways, but the fundamentals are surprisingly consistent.
• John Psmith reviews 50 Years of Text Games. Fun from the very beginning, where he points out that the earliest games were graphical because simple graphics and logic require less memory than even a fairly short series of texts, and that it took a while for computers to be able to fit games (one of the games mentioned, Adventure, outgrew the PDP-10 it was built on, and the game developer had to come up with a compression tool). This piece also cites a design pattern that's getting increasingly common: having a state machine, expert system, or other fixed graph representation of the logic of some system, and then using LLMs as an interface to find the right node of the graph and to summarize what happened when it was traversed. This applies to both practical applications ("AI-for-X" where X is a category where mistakes are expensive) and to games. But games will probably explore the possibility space faster.
• Jay Caspian Kang asks if AI makes college obsolete, with a so-far ambiguous answer. What usually happens in cases like this is that we have to think about different elements of the bundle. Talking to an LLM is more convenient than going to office hours, and most of the time students won't be asking questions that push the boundaries of their professors' skills (this is a general statement about how to use LLMs to learn, not a complaint about modern students—there is very little I ask an LLM that would stump an expert on that topic). On the other hand, LLMs won't be a good substitute for the lived aspects of college, whether the salient ones are career networking, partying, or being in a more exclusive dating pool. But it's also hard to separate that from the rest of the bundle: selling college as four years of debaucherous partying, with the expectation that you'll do the learning on your own, is not going to get priced at anything like current tuition. And in practice, LLMs have negative value in credentialing, in that they're exceptional Dunning-Kruger accelerators absent deliberate effort to steer them in a more critical direction. Unbundling is a hard problem to manage for institutions with high fixed costs, especially if those institutions are already under demographic pressure (total births in the US peaked, oh, just over eighteen years ago...). The optimistic outcome is that the next decade will be a period like the era of land-grant universities, or when Gilded Age industrialists started putting money into schools that would follow the German research university model. We might get a few more names added to the short roster of generally-seen-as-great schools, though this time it will be at the cost of many less remarkable ones disappearing.
• Virginia Postrel tells a story of a bold bet on R&D that led to a product that was initially not cost-competitive, but that could achieve incredibly low cost with enough scale: the disposable diaper. Modern disposable diapers are a materials science miracle; they keep getting smaller but more absorbent, and while I've transferred a lot of wealth to Procter & Gamble over the course of four kids' diaper-wearing tenure, they've been a pretty good deal.
• A fun Read.Haus question: insurance companies will sell you a policy when, statistically, the present value of what you pay them is more than the present value of what they might pay you, and so do options sellers. What about markets where the opposite holds true? Such markets do exist, in Vegas, Macau, and thanks to some recent changes in regulations, possibly in your pocket right now. Insurance and gambling are really the same business, and the only difference is which side of the bet you have to pay money to seek out. Which means that what defines that boundary is what kind of risks are fun to think about, and what kinds are unpleasant. Very few people get a pleasant adrenaline rush from imagining how inconvenient it would be if their house burned down, or if someone t-boned them in an intersection. But it's very easy to think about what you'd do if you won the Powerball. You pay to take fun risks, you pay to avoid less-fun risks, and if you think there's a discrepancy in how these risks are priced, you pay in some mix of effort, ego, and returns when you try to exploit it.
• In this week's Capital Gains, we look at when people are buying a stock for reasons other than maximizing returns. Institutional investors do this in a way that's very legible to other institutional investors—the payoff from owning a gold mine, or a high-cost oil producer, is pretty easy to understand in terms of what will will be doing badly when they're doing well. But other retail investor trends are more inscrutable.

Books

Steve Jobs in Exile: The Untold Story of NeXT and the Remaking of an American Visionary: The weird thing about Steve Jobs is that he was the perfect frontman for Apple when it started, the ideal CEO in the 2000s, and, as Apple's board recognized, a liability for the company in between. Apple outgrew Jobs' management skills when he was in his twenties, but somehow when he came back he was able to preserve Apple's position in PCs, move into music players, and define the category of smartphone. And the years he spent learning how to do this were at Pixar and NeXT.

What's strange about this story is that, at least in the early NeXT years, Jobs had exactly the same problems he had at Apple: grandiose ambitions, ludicrous timelines, scope-insensitive perfectionism, etc. With the passage of time the NeXT plan looks cooler and cooler: a powerful computer with an ominous cube form factor, running Unix, with lots of built-in libraries that made it easier to write software for it. NeXT made the interesting aesthetic choice to ship with great audio support but only a grayscale display, right when color was getting more common. Unfortunately, it was also a very insistent system. It had a high price point ($6,500 in 1988, or about $10k including a printer and decent storage), and those nice software libraries required users to learn Objective-C, which was not widely used at the time.

NeXT was a good technological bet, but a terrible organizational one: once Jobs was unfettered by legacy businesses and an unfriendly board, he could run the company how he wanted. And it turned out that he wanted to run it pretty irresponsibly. NeXT did everything a bit weirdly, with exactly two salary bands, a constantly-shifting product roadmap, big joint ventures that Jobs would suddenly blow up for no particular reason, etc. It's actually a struggle to see what caused Jobs to evolve as a manager, other than the fact that his absolute control over the company meant that there was no one to blame when things didn't work.

The weird thing is that NeXT worked as a business, in part because in retrospect it was built as the target for a strategic acquisition by Apple. It had a great operating system, but a small installed base of machines running it; it had a CEO who was great at sales but limited in what he had for sale. And, by the time Apple bought them, Jobs seems to have figured out management by a process of elimination—alienate enough cofounders, partners, and investors, and you'll learn your limits.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Has anyone written anything good on how companies and founders coevolve? Jobs is an extreme example, but there have been other founders who basically built a company that was beyond what they could run, and either leveled up or left.

Diff Jobs

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

• Series A, ex-Navy defense firm building AI-enabled drone defense systems is looking for an electrical engineer with range: schematic design, electrical simulation, printed circuit board (PCB) design, and firmware development in high performance languages (C++, Rust, etc.) If you’re interested in power and control systems for mission critical technology, this is for you. (Austin, TX)
• A startup building a new financial market within a multi-trillion dollar asset class is looking for a data scientist with commercial financial experience. (if you’ve been an investor but are newer to the data side, that’s great too.) (NYC)
• Lightspeed-backed team building the engineering services firm of the future is looking for founding members of technical staff excited about working alongside civil engineers to translate their domain expertise into the operating system that powers the next era of great American infrastructure. If you’re an engineer with strong product intuition, who's energized by access to users, and excited by the prospect of transforming how we design and construct our built world with frontier AI, this is for you. (NYC, SF or Remote)
• AI Transformation firm with an ambition to build an economic world model to run swathes of the private, unstructured economy is looking for FDEs, Platform Engineers, and business generalists who understand how to solve problems.  
• Well-funded, frontier AI neolab working on video pretraining and computer action models as the path to general intelligence is looking for researchers who are excited about creating machines that learn from experience, not text. Ideally you have zero-to-one pre-training experience and/or are a high-slope generalist who’s frustrated that the big labs aren't doing this. (SF)

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.