Arbs Close, Infrastructure Remains

In this issue:

• Arbs Close, Infrastructure Remains—One way or another, arbitrage doesn't usually turn out to be an infinite free money machine. Usually the problem is with building the free money machine, though if that's solved then "infinite" becomes a big issue. And yet, many successful organizations start their hypergrowth phase by identifying something that looks suspiciously like an arbitrage!
• Abundance and Realignment—Sometimes, adding a little veto power is a good thing.
• AI Undeath—Celebrities compete against the living and the dead.
• Basic Capital—More leverage where it works best.
• Tariffs—What has actually been accomplished by tariffs so far?
• Insurable Risks—The early days of figuring out the right economic exposure for using AI.

Arbs Close, Infrastructure Remains

In a sense, starting a truly new startup is at the exact opposite end of the spectrum from arbitrage. In one case, the goal is to identify some completely new product, based on a unique insight about the needs of customers, the possibilities of technology, or some combination of both. In the other, arbitrageurs can be completely agnostic to all facts about the world other than the economic equivalence of two or more assets that trade at different prices. You don't have to believe in these prices: it's perfectly possible to do cross-exchange arbitrage between two crypto exchanges you think are scams, evening up the price of a token you think is worthless.^[1]

Arbitrageurs implicitly assume that the world is full of people who are rational, but imperfectly so, and that the big money is in keeping the expressions of their views that take place through financial markets roughly in line with one another—taken too seriously, it's a form of nihilistic optimism that every important problem will be figured out by somebody else, but that they won't perfect the last few details of their approach before moving on to the next thing. Startups often assume that the world is missing a critically important big idea, and that only they can bring it to fruition. It's a form of realist pessimism holding that nobody else will pick whatever the low-hanging fruit happens to be. At the same time, believing that there are secrets left in the world and low-hanging fruit left to pick is a form of optimism in itself.

That's taking both viewpoints a bit too seriously, of course. Realistically, most companies are operating in a gray area, and there's a way to interpret each in light of the values of the other:

• Take arbitrage, an activity that could literally amount to having two pricing screens up on two monitors, and clicking the right button on each when they get out of whack. It usually isn't that simple, for the very obvious reason that the obvious arbitrages have all long ago been arbed away. You might find one if you're lucky, but if you're looking it's better to be smart, diligent, and, increasingly, to have lots of infrastructure for both identifying signals and executing trades. In an efficient market, the equilibrium return for arbitrages is mostly determined by the fact that finding and exploiting them is a full-time job that competes with other similar full-time data-oriented jobs. Over time, the "arbitrage" part of arbitrage-style businesses starts to wash out: the real driver of business outcomes is whether a given opportunity is worth the time it takes to exploit, not the question of whether or not it's free money ignoring transaction costs.
• In the opposite direction, even though many startups do launch very original things, there's relentless pressure for evolutionary convergence: the consumer-facing ones will almost certainly find a way to interrupt your otherwise-delightful user experience with ads, unless they opt for a mix of ads, subscriptions, and/or marketplace fees (which are really just compensation for helping with customer acquisition, so also ads). If the tool you're using makes money, there's a good chance it's going to get paywalled (and, interestingly enough, the specific features most directly connected to making money are the ones that get locked behind a paywall!). Depending on who they sell to, they'll grow to the point that they have a roughly typical mix of coders, operations people, customer-facing personnel, etc. There just turns out to be a lot of products that are worth somewhere between $10 and $50 per month, even if each one's quite unique in other ways.

One of the ways these worlds collide is when a company starts with one model and slowly switches to the other. Netflix was originally a pretty simple arbitrage: the big studios underestimated how much viewing was moving over to streaming, and by striking a long-term licensing deal you could enjoy the demand-side economics of the ubiquitous-streaming present with the cost structure of the pre-streaming past. Costco's spiritual ancestors, FedMart and Price Club, were partly born as workarounds for the Miller-Tydings Act that let manufacturers set retail prices (a membership-based company turned out not to technically count as a retailer). Amazon got a nice boost from Quill Corp. v. North Dakota, which prevented states from collecting sales tax from businesses that didn't have a physical location in their state. Airbnb benefited from the fact that what it was doing didn't quite fit into the legal categories established to cover hotels, and Uber was very fortunate that so many cities had been regulatorily captured by the taxi industry, so there was plenty of demand for a better and cheaper service. In finance, there are a few funds that started out in convertible bond arbitrage and later expanded into other strategies—Elliott Associates is better known for activist investing, and Citadel's known for doing just about everything, but both originally started as convertible arb shops.

What all of these have in common is that the arbitrage sounds straightforward on paper, but there are reasons nobody had done it before. Anyone could look at estimates of the idle time of cars (95%+) and imagine a more liquid market in transportation—the anarcho-capitalist economist David Friedman imagined a similar service in The Machinery of Freedom, well before the smartphone. It turned out that Uber needed some closely-related third-party infrastructure (iPhones), some more distantly-related infrastructure (GPS), and even then it had to dump a lot of spending into a market to actually get to scale. It would be very naive today to ascribe Uber's current business to regulatory arbitrage—no issue could be more important to ride-hailing interests than determining exactly which rules Uber has to follow.^[2] And Uber itself gets a bad rap for this: they were actually complying with commercial vehicle rules and working with existing black car companies, until they opened up their business to a wider range of drivers in response to a similar move by a small competitor, Lyft. The minimum infrastructure necessary to build a good regulatory-arb business turned out to be significant enough to support a more regulated, higher-taxed business at scale. Similarly, Netflix got good enough at delivering streamed video content with its early streaming efforts that originals were worth doing as a high-margin (over the long run) complement.

In many of these cases, the long-term vision actually preceded the arbitrage-style angle, even though more arbitrage-adjacent business models dominated these companies' early days. That doesn't mean it wasn't causal: finding an exploitable arbitrage somewhere is basically the economy's way of offering non-dilutive financing to the first company to spot it.

Over time, most companies get a bit less arb-y. The big opportunities they could exploit at small scale become smaller, more priced-in opportunities that can't support them as they scale. So finding some way to buy X and then resell it for slightly more just doesn't describe their business: they have to transform their inputs in a significant way in order to have a durable advantage. Arbitrages don't scale forever, but arbitrageurs can sometimes turn a trade into an actual company.

Diff Jobs

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

• A hyper-growth startup (10x growth in 9 months) that’s turning customers’ sales and marketing data into revenue is looking for a head of deployments who is excited to work closely with customers to make the product work for them. Experience as a forward deployed engineer (Palantir, etc.) and/or leading enterprise deployments preferred. (NYC, SF)
• A Google Ventures-backed startup founded by SpaceX engineers that’s building data infrastructure and tooling for hardware companies is looking for a software engineering manager with 7+ years experience building large scale distributed systems. (LA, Hybrid)
• A multi-stage, fintech focused investment firm is looking for an investment associate to support thematic opportunity identification, diligence, and execution. Investing experience OR high-growth operating/investment banking/consulting experience and demonstrated interest in fintech required. (NYC, London)
• YC-backed company that’s turning body cam footage into complete police reports and automating other cumbersome law enforcement workflows with AI is looking for a junior backend engineer to operate in a forward deployed capacity. Entry-level ok, ability to pass an FBI background check required. (SF)
• YC-backed startup using AI to transform how companies quantify and optimize engineering productivity is hiring formidable full-stack and AI engineers. Experience with React + Typescript, Go, or Python on the ML side a plus. (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.

Elsewhere

Abundance and Realignment

Pro-growth is an interesting constituency, because almost any industry-level lobbying group would be willing to trade lower aggregate growth for better conditions for their specific industry. Whereas the benefits of economic growth on the whole are far more diffuse. But it's also a good technocratic rallying cry, and now there's a bipartisan caucus in favor of it. One irony here is that a big part of the Abundance critique is that veto power has been distributed too widely. But in practice, a caucus like this is more likely to block something than to push it forward, even though what it's blocking is usually worth opposing.

AI Undeath

What does it do to the economics of fame if celebrities with high name recognition accumulated over a long career can continue competing from beyond the grave? This is a genuinely useful question right now: an AI avatar of Agatha Christie is teaching a mystery-writing course, and licensing the likeness from the writer's estate. One thing it means is that celebrities are actually handing down an additional asset to their heirs, in the form of future royalties. It also means that there are higher barriers to entry—your highest-quality competitors are most likely to have an AI-generated afterlife, so the minimum skill bar keeps rising. On the other hand, some kinds of media are more about parasocial relationships with the creator than about the content itself, so if AI commoditizes some kinds of content, that makes the complementary parasocial piece more valuable. So the net result is that the payoff for celebrities gets even more skewed: there's a longer tail but also a more extreme payoff for the winners.

Basic Capital

Like many countries, the US structures its tax laws and financial rules to encourage saving for retirement. This takes two forms:

• Various tax-deferred schemes for making unlevered investments into risk assets, primarily equities, in order to take advantage of their long-term compounding.
• Extraordinarily cheap leverage, and some tax benefits, but only to people who make a concentrated bet on residential real estate where they live—i.e. doubling down on the unhedged risk that something bad will happen to local labor markets.

Both of these gesture in the right direction, but they lead to a weird situation where households, who are best positioned to weather short-term declines in the prices of risk assets, end up getting the best access to leverage if they make the one investment they should be underweight. Fortunately, financial engineering often finds a way: Basic Capital is a newly-launched startup that gives investors access to leverage within tax-advantaged retirement accounts. One way to look at this is that a young saver without dependents is probably underlevered financially, because their big asset is the future income they'll receive from working. But if they just try to lever up on their own, going 5x long equities through futures or something, they'll inevitably blow up. What they need is some kind of structured savings product that lets them scale up their exposure to higher-return asset classes that don't further concentrate their economic portfolio, without the risk of a margin call.

Disclosure: Basic Capital uses Diff Jobs.

Tariffs

The US and China are cutting their tariffs rate by 115 percentage points, to 30% and 10% respectively, for the next 90 days while they continue to negotiate. One view of the tariffs is that extreme numbers are the start of a negotiation, not the desired end goal, but it's worth asking what the US has gained, from a protectionist standpoint, over the last month and change. The companies that did the best in a tariff environment are the ones that thought tariffs were coming, stocked up in advance, and then ignored them on the assumption that they'd come back down. Anyone who onshored production, or even scoped out non-Chinese suppliers, seems to have wasted their time. One of the problems with any big change in economic policy is that its real impact comes from real output, i.e. the long-term effect that matters isn't what Americans pay for smartphones and socks, but what shifts there are in global capital expenditures to move production to new locations. And the more extreme the policy move, the more people will anchor to the previous status quo rather than to some midpoint. The US could work around this, perhaps by promising cheap credit or other subsidies to domestic manufacturers that would offset some of the losses they face if tariffs decline, but that policy, too, needs to be credibly sticky.

If tariffs do go back to a moderate level, closer to the pre-Liberation Day consensus expectation, it has a moderate effect on actual trade, but it does have an effect: it means that more companies will have inventory-centric backup plans, i.e. they'll stockpile a little more than they otherwise would, and perhaps aim for more easily-canceled orders when they do business with countries that might be hit with tariffs in the future. This probably does make the global economy more resilient, but it's a very roundabout way to achieve a secondary aim.

Insurable Risks

Insurance companies are writing policies that cover the cost of AI errors ($, FT). This is a very interesting thing to insure, because there are some fun tradeoffs with correlated risks:

• If prices are set in reference to particular models, it selects for a monoculture, where there are fewer errors, but more catastrophic ones.
• On the other hand, it's expensive to underwrite every single model, and they all seem to develop emergent properties, so the underwriting has to be continuous.

This is, on net, a very good thing, and it's another illustration of how well AI blurs boundaries: it feels more insurable to bet on the capabilities of a piece of software than on a team of people, but there's a different shape to the risks at hand. And insurers have some tricky motivations: if they do end up implicitly betting the business on a single model, their incentive is actually to magnify that bet as much as they can get away with: in a disaster scenario, there's no difference to the insurer between losing 100% of their money and losing 500%, even though the ecosystem as a whole cares a great deal about this. Having one company offer this kind of policy is a great start, but the market won't be mature until there are a lot of them.