Closing Industry Frontiers
The US census used to include a "frontier line," showing the point past which the population of American citizens was 2 per square mile or less. This was a useful feature, and a sort of early instance of what would later become the neoliberal impulse—use the government's resources to produce something that helps people free themselves from government regulation by resettling somewhere else. This part of the census is best remembered today because Frederick Jackson Turner's influential essay The Significance of the Frontier in American History) opens by citing the 1890 census, which had finally dropped the "frontier line" since there was no longer a clean break between the settled and unsettled parts of the country. There were still dense places and less-dense places, but by 1890, if you wandered into a part of the country that didn't have many other people, there was probably a very good reason.
The frontier thesis is still an important part of how Americans are taught history, and it's an idea that translates well to other industries, too. For a long time, the software industry could view the US economy as a sort of unsettled region past the frontier, still under the protection of the US property rights system, but largely open to be claimed by anyone enterprising enough to homestead it. Any time bits are exchanged between companies or within companies in some inefficient, humans-in-the-loop way, there's room for a more cleaned-up software model to improve things. And any time some of those bits are the lucrative ones involving the transfer of money, there's opportunity for not just any business, but a lucrative one.
One of the reason's Turner's frontier thesis is so popular is that it explains a lot about American culture—a country with a frontier is one that needs positive-sum interactions in settled areas, because everyone has the option to simply leave. It's also a country that has to value individualism, because the people pushing the frontier have to run a full-stack economy: if they need something, they either need to bring it with them, or find a way to catch it and kill it, grow it, or make it. The frontier evolves in steps:
- There's an initial stage that's somewhat invisible to the rest of the world, and requires a high degree of self-sufficiency. This is a time when uncertainty is high, both in the general sense that the known distribution has a wide range of outcomes and in the specific sense that the distribution itself is hard to measure. Turner cites different archetypes that arrive at different times: "When the mines and the cow pens were still near the fall line the traders’ pack trains were tinkling across the Alleghanies, and the French on the Great Lakes were fortifying their posts, alarmed by the British trader’s birch canoe. When the trappers scaled the Rockies, the farmer was still near the mouth of the Missouri." (A bit later, he notes that Lewis and Clarke encountered traders who showed them mountain passes—before there's formalism and legibility, there are people with a risk appetite looking for high-variance opportunities.)
- Over time, there's settlement, and two forms of specialization: homesteads link together to towns, so farmers can sell cash crops and import some goods rather than requiring full autarky. And regions specialize based on their resource endowments.
- And then, the frontier dissipates. There's still variance and uncertainty, but more life decisions involve fixed and measurable tradeoffs. Someone can leave the city and go back to the farmstead—a phenomenon that makes 19th century unemployment rates incredibly annoying to measure, both because it's harder to count workers who have returned to the farm and because it's hard to know if they're actually working—but there isn't an escape valve.
It doesn't take too much effort to extend the frontier thesis to subcultures, artistic movements, or, of course, industries. Much like the first step of frontier evolution, the first wave of crypto companies had to figure out everything: their crypto infrastructure, regular financial infrastructure, user interfaces, marketing, and accounting (some of them did not get all the way to the end of this list). You can say the same for early SaaS companies—they spent a lot of their time and effort on tasks that are now something-as-a-service provided by some other startups. For example, stories from the early Facebook days often revolve around infrastructure complaints, but if you were to build a social app today, you wouldn't devote much time to figuring out where to put specific servers; cloud hosting abstracts much of this away.
Within software, there have been new frontiers. Software as a separate business, rather than as a free product bundled with hardware, was an early example of a new frontier. In one sense, Microsoft was a novel experiment with unbundling, but in another sense, it was an effort to identify and claim vast swathes of territory: if there wasn't much of a "software business" independent of hardware, the first company to try to build one was naturally the owner of the entire model.
The frontier is a continuous phenomenon when it's open, but a discrete one when it's closed. Turner points out that the frontier story in America is older than the frontier story of America, since the colonies started out as the far frontier of Europe. New industries end up being ways for the best practices of old ones to make inroads into the rest of the world; one way to read the success of Meta and Google is that they took the direct response mail industry's commitment to data-driven marketing and made it available to the rest of the world's ad budgets, invisibly, and at unprecedented scale. But that deployment phenomenon means that there's a limit to how much a new industry can accomplish. It's riding a steep s-curve, but growth slows and barriers to entry rise rapidly. Meanwhile, tighter economic integration between the frontier industries and the rest of the economy eventually means that there's less of an edge involved in focusing on either.
This is happening to the software industry. Those gaps of untapped data and unautomated processes are steadily shrinking, not just because startups are filling them, but because so many companies see software as, increasingly, a core part of their business. You know this is true because when big companies mess up their software, whether it's Boeing's 737-MAX or Southwest's scheduling bugs ($, WSJ), it's headline news.
Frontiers become invisible over time because the integration goes both ways. "The frontier is the line of most rapid and effective Americanization." It's very funny to bump into finance friends who have pivoted to startups. The economic frontier, to mildly paraphrase Turner, "Strips off the garments of civilization and arrays him in the [company T-]shirt and the [Allbirds]." Decentralization, egalitarian office norms, and compensation skewed towards equity are all ideas that apply more broadly than to just the industries that took them seriously first—but that also makes them a diminishing source of alpha. New industries lead to spillover effects in both directions, but the net result is that they're harder to distinguish from the legacy companies they've displaced.
When the real estate in question is virtual rather than physical, it's hard to draw a line on a map, but in a sense that speaks to the frontier thesis: it's hard to delineate software and non-software businesses when cars and heavy machinery are increasingly giant computers processing huge volumes of sensor input, and when service jobs can be augmented by machine translation and text generation models. And once you can't easily draw a line marking out the frontier, that frontier is gone.
Disclosure: Long MSFT, META. (Returns go down when the frontier closes, but on a risk-adjusted basis, settled civilization is probably the better deal.)
A Word From Our Sponsors
We're testing out something new with The Diff: a free mini-newsletter. Each week, we'll break down a single financial topic—whether it's the impact of interest rates, the meaning of capital allocation, or what actually happens when you make a trade.
The Slow Move to Subscriptions
A big component in the multiple expansion of software companies in the last decade has been the shift from one-time purchases to subscriptions. Some enterprise companies had indirectly made that switch before through large maintenance contracts, but it's generally easier to keep a customer from churning than to convince them to make another purchase. And for some products, this produces some reasonable alignment between customers and software vendors: the customer uses the product continuously, and it slowly improves, so it makes sense for the cost to be continuous, too. But UK-based Sage, which makes accounting software, is having trouble persuading customers to accept a price hike embedded in the subscription model, and is threatening to turn off their product entirely if they don't make the switch ($, FT).
One annoying effect of the rise of subscription software is that the economics are so good for sellers that they change incentives. Some of the most useful software products in history are Unix utilities that do one thing extremely well, and have done it roughly the same way since the 1970s. Turning something like that into a service that can justify recurring payments means trying to add new features, but the natural end state is for the product to always be 85% done, and for the definition of "done" to keep expanding. As one Sage client said, "Double entry book-keeping hasn’t changed much in a few hundred years."
When investors are making good money during bad times, it's probably through some kind of there-is-no-alternative trade: everything looks bad, but there's still money seeking returns, and that money will flow into the least-bad trade. During deflationary recessions, that usually means buying low-risk bonds at what feels like low yields, and then watching those yields continue to decline. (For example, the return on US investment-grade bonds in the 1930s was 9.9% annualized, better than the long-term return on stocks!) But during an inflationary and chaotic time, the trade can be different: Turkish stocks rose 110% last year in dollar terms ($, FT), despite high inflation and a policy environment that can be charitably described as unique.
Some of this is for sensible macro reasons: Turkish equities are somewhat hedged against currency fluctuations in the short term, because so many Turkish companies export to Europe and compete on costs. When the lira is down, they're more competitive, and this shows up in their profits. Longer-term, that effect washes out a bit—Turkish companies might get more money for their cars and washing machines, but they have to pay a steep premium to replace the equipment used to manufacture them, and workers will demand higher wages to offset the hit their standard of living takes as import prices rise. But the other piece of the puzzle is that other assets in Turkey aren't that much more attractive; the banking system is messy, and obviously cash is not ideal with inflation running above 80%. Property is a classic inflation hedge, but is tied to the domestic economy. And one nice thing about equities in unstable countries is that they're a way to band together to get a group discount on political influence; countries with weak governments often have strong companies, so investing in them means hedging against political risk.
Late last year, a startup announced that it had experimentally released sulfur particles into the atmosphere in order to offset greenhouse gas effects. If this sounds like something out of a science fiction novel, that's completely fair; they say they were inspired by Termination Shock, the recent Neal Stephenson novel that takes a rogue effort to do this as its premise.
The reaction to this story illustrates something useful about property rights. When Boulton and Watt designed their steam engine, they didn't think of themselves as interfering with some common right to a particular composition of the atmosphere, and now that their distant heirs have gotten so good at extracting hydrocarbons from the earth and burning them to produce useful energy, we find that this is a serious problem. So new efforts to change the composition of the atmosphere in climate-altering ways naturally earn some skepticism—even if their explicit goal is to reverse the effects of doing the same thing by accident.
Once a fraud is exposed, the narrative starts moving forward and backwards at the same time: we see how the aftermath plays out, but we also see where the warning signs were early on. This WSJ piece digs into the history of Alameda Research ($), and shows that the trading firm was less profitable than it implied and had serious risk management problems from the beginning. Part of what went wrong at the company was that its initial financing model was designed to quickly exploit a temporary opportunity: crypto prices were significantly higher in Japan than in other parts of the world, and in a situation like that it makes sense to borrow aggressively at high rates, with the expectation of winding down the trade when the opportunity closes. But the fund later faced style drift, since it didn't have an advantage in trading high-volume cryptocurrencies but was willing to make a market in more esoteric ones. In one sense, the worst thing that happened to Alameda was the crypto boom of 2020-2021, since that boom disproportionately benefited entities that were levered long the weirdest cryptocurrencies. Having one genuine winning strategy and then falling backwards into another one is a recipe for overconfidence and then disaster.
Elasticity in Electricity
The amount of Bitcoin hashing dropped by almost 40% in response to storms last week; crypto miners are an extremely inessential service during a power shortage, and happen to have favorable economics for deciding not to consume electricity (as long as they're either paying the spot rate or agreeing to curtail their use when demand is high). This creates some very interesting tradeoffs, since one consequence of crypto's higher demand for electricity is that it makes older coal power plants economically viable for a little bit longer. But it also means that the grid can more quickly adapt to changes. A decent economic framework is to look for shock absorbers: there are some kinds of variability that are unavoidable, but that different entities have different tolerance of. When governments subsidize fuel and food, for example, it's because they're better than households at managing a sudden deficit. These shock absorbers have a cost, of course, but supply and demand shocks all have to be absorbed by somebody if the market is going to clear.
Companies in the Diff network are actively seeking talent! If you're interested in exploring growth opportunities at unique companies, please reach out. Some top current roles:
- A successful crypto prop-trading firm is looking for new quantitative developers with experience building high-performance, scalable systems in C++. (Remote)
- A fintech company that brings systematic strategies to everyday investors seeks a senior backend engineer. (US & Canada, remote)
- A leading provider of market data to both retail investors and institutions is looking for a staff-level frontend developer. (US, remote)
- A company using ML tools to spot software bugs is looking for multiple roles in machine learning, systems programming, and deployment. (Northern Virginia)
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.