Building the Oracles

Plus! Price Discrimination in Media; AI-Generated Propaganda; War Economies, Redux; Unbundling and Rebundling; Attrition

Byrne Hobart Byrne Hobart

In this issue:

Today's issue of The Diff is brought to you by our sponsor, Butcherbox.

audio-thumbnail
The Diff August 19th 2024
0:00
/708.075102

Building the Oracles

One part of technological process is formalizing and refining things that used to be informal interactions. This is especially visible in social networks, which tend to have socially-awkward founders who are explicitly deliberating about social interactions that most people do naturally, and who thus have a more detailed model of what those interactions are. But it shows up elsewhere, too, sometimes even taking the form of bringing an old practice back to life.

For example: right now, a fairly popular business/investment category is the search fund. It works like this: you a recent MBA or veteran of a couple years of private equity, will raise a small amount of money from friends and family in order to buy out a small business whose founder wants to retire. That business has a role in its community and in the broader economy, and if you do a bad job, whatever equity you put in and whatever time you invested will both be zeroed out. So this approach tends to really focus the mind. Hundreds of years ago, there was a nearly-identical path, but in the public sector: military commissions used to be bought and sold, and when they were bought, it was often by someone who had borrowed money. It's obviously not perfectly optimal for every general to be the highest bidder rather than the best leader, but this system actually had some surprising benefits. The Aristocracy of Talent notes that this solved two pressing problems: first, it meant that the government didn't need to provide pensions to officers, because those officers had an asset that they could sell to the next generation. Second, it actually created a financial incentive for officers to be more aggressive, since the way they'd pay their debt was by winning battles and looting. (There are, obviously, some negative externalities here for the lootee, but when this system was put in place they do not appear to have been consulted.) The problem that young people need third-party funding to inherit positions of responsibility is unchanged, and sometimes, in very different contexts, we'll reinvent the mechanics of this.

Let's consider another one: many ancient civilizations had rituals that allowed people to commit to arbitrary promises in a binding way. Swear to do something by Zeus in front of all of your friends and neighbors, and you're probably going to do it.[1] That's a useful social technology. We can do person-to-person and organization-to-organization commitments like this quite easily, but 1-to-N commitments are hard. There really isn't a way to risk opprobrium when you promise that this is the year you'll run a marathon or finish your novel, and then it turns out that it isn't.[2] Paper Belt on Fire points out that the technology of making a public commitment that's visible to everyone, and that carries penalties if it's violated, has actually been reinvented, mostly by Vitalik Buterin.

A smart contract is basically a modern implementation of an oath. It's a way to make a definite financial commitment based on some particular outcome, as judged by a third party. It's at least part of the technology stack for rolling your own version of Homeric reality, promising to either achieve great things or literally die in the attempt.

And it's overwhelmingly, though not exclusively, used for gambling instead. Crypto may be full of degenerate gamblers, but there's a separate reason that this happened. For a smart contract to work, it needs some kind of oracle: a trusted source that will verify that something did or didn't happen, or a way to measure the quantity of what happened. If you want to use smart contracts to create an esoteric weather derivative, maybe to transfer climate change risk between ski resorts and ice cream vendors, you and your counterparty need to agree on which metric to use, and you need to be completely confident that this metric can come from an agreed-upon source that will be consistently updated.

There are many sources that are pretty good, but not necessarily perfect, for this. But there's one category of problem where oracles are incredibly easy to find: finance. Creating a call option on ethereum, or some more esoteric derivative like a bet on the correlation between Solana and Dogecoin, is a matter of choosing which centralized or decentralized market's prices to use. Similarly, creating a smart contract that provides margin loans and automatically sells the assets in question is straightforward.

But how do you construct an "I will run a marathon this year" contract? You might plan to scrape the marathon's official site, but now you have both a broad process and an imperfectly deterministic one: maybe the site changes, maybe the person runs a different marathon from the one you expected, maybe they leave their house for a jog one Saturday morning and come back 26.2 miles later. Pretty soon, the optimal contract for this becomes a not-so-smart contract—either the absolute overkill option of hiring a lawyer to articulate exact terms or, much more realistically, running the whole thing on the honor system. It's hard to have high-stakes questions come down to personal honor in a culture that doesn't treat that as the preeminent virtue, so the higher the stakes the harder it is for either side to trust the process—which, as George Akerlof realized long ago, means the deal probably just won't happen.

Markets solve the oracle problem nicely: there's an existing incentive to have up-to-date data, the data in question has a built-in economic incentive to be correct—to print a bad trade for the oracle, you have to make a bad trade yourself—and the network effects of markets tend to be not just durable but more durable than markets themselves. (The default benchmark for the world's most important commodity is set by an organization that was founded as the "Butter and Cheese Exchange of New York.")

There's a version of the future where oracles are everywhere: the more separation there is between frontends that interact with consumers and backends that perform the actual services those users want, the more APIs there will be that have rich data that's worth betting on. So certain prop bets will be more easily implemented by smart contracts. And if more people are using wearables, logging their behavior, and being tracked online and off, it's at least technically more feasible to have smart-contracts-for-everything. This information is mostly controlled by a handful of big companies, which have strong economic incentives not to share their data, and in most industries the frontend and backend are provided by the same entity, which doesn't expose its APIs. Google is technically capable of letting people use its results for their own search products, but it's not a service they offer any more (they used to, but shut it down in 2011).

But of all of the plausible antitrust remedies that have been proposed, "open the APIs" is the one that most credibly weakens the biggest companies but preserves consumer welfare. It's incredibly challenging to figure out which functions should be broadly available and which should remain proprietary, but as a general outline it makes sense.

So there's a reasonable possibility that the world will have a lot more oracles in the future: many more ways to access one specific URL and get precise information on a real-world event, which you can then use to construct smart contracts based on it. Finding a market-maker is still a challenge, so for now this would mostly enable bilateral bets. But it would lead to something closer to a complete market. It's a weird substitute for swearing an oath by Zeus, but even if the temples are ruins, the human impulse behind them remains.

  1. Annoyingly, it's hard to track down the first version of this claim that I'd heard, which was that New Year's resolutions descend from a Roman New Year ritual held at the temple of Janus. There are many claims that this is the origin of New Year's resolutions floating around, but they don't seem to cite any specific ancient history. If you know whether this was an actual Roman ritual or fake history invented by a Victorian moralist, please let me know! ↩︎

  2. The closest thing we have to a one-to-many public commitment with social costs to the failure to uphold it is the two-to-many ritual of marriage. Even then, though, the stigma of divorce is weaker than it used to be, and the legal process is easier. Marriage, as it was once understood, is a useful social technology, and someone ought to legalize it again. ↩︎

A Word From Our Sponsors

Choose Your Own Offer (Free Ground Chicken, Ground Pork, Ground Sirloin) + $20 off each of your first three boxes!

Unlock the ultimate convenience and elevate your culinary game with Butcherbox! Imagine having premium proteins delivered straight to your door every month. With their top-quality meats and helpful recipe tips, you'll be creating mouthwatering, high-quality meals in no time. Butcherbox isn’t just about saving time and money—it’s about gaining access to exclusive products and special deals you won’t find at your local grocery store. From succulent filet mignon to wild-caught seafood, Butcherbox ensures you always have the best ingredients on hand. Plus, you can pause deliveries whenever you want, making it incredibly flexible to fit your lifestyle. 

And now, for a limited time, enjoy an amazing offer: choose between Ground Chicken, Ground Pork, or Ground Sirloin to receive in your first box! As an added bonus exclusive to our newsletter readers, you'll get $20 off each of your first three orders.

Don't miss out—transform your meals with Butcherbox today and savor the difference top-tier proteins can make in your cooking.

Try Butcherbox Today

Elsewhere

Price Discrimination in Media

The Minneapolis Star Tribune plans to rebrand to the Minnesota Star Tribune and to expand its coverage. In one sense, it's a counterexample to the long decline in the newspaper business in general and local papers in particular. But it also points to the meta-model of media: these companies have value as trophy assets far beyond the present value of their future cash flows; to some extent, thinking about the economics of the business is a bit like asking whether a Picasso throws off enough revenue to cover the cost of keeping it safe and in good condition. In that sense, a local newspaper has two kinds of subscribers, one of whom pays many orders of magnitude more than the rest, but who is still getting the same product, just with the option to take a bit more credit for it.

AI-Generated Propaganda

The Diff argued earlier this year that AI-based political propaganda would be an insignificant feature in elections for the forseeable future. It's hard to definitively prove this case, because we only know who got caught, but the getting-caught part is still instructive. OpenAI says they identified an Iranian group that was using ChatGPT to create election propaganda (targeting, or at least pretending to be part of, both sides—it's a lot more plausible for a bad actor to try to delegitimize the entire process and sow chaos than to support a particular candidate, though they will sometimes do that, too). This operation was using OpenAI's tools, but didn't appear to get any traction whatsoever; they might have saved time writing whatever agitprop they were writing, but it didn't actually agitate anyone. And this is fairly common: last month, the FBI announced that they'd identified a Russian influence operation, and actually listed all of the twitter handles involved. Spot-checking a few dozen, I wasn't able to find a single instance where anyone had ever mentioned or replied to any of the accounts.

War Economies, Redux

Despite sanctions and high defense spending, Russia's economy is booming ($, Economist). This is an instance of swings between populism and elitism as an economic stabilizer ($, Diff): the country had had relatively tight fiscal policy for a while, partly in order to store up reserves, but for a variety of reasons that's no longer a priority. So they've switched to loose fiscal policy partly offset with tight monetary policy. This approach tends to redistribute wealth to lower earners. They're more likely to be the beneficiary of government programs. On the other side of the equation, when the inflationary impact of high spending is offset by higher interest rates, which tends to depreciate assets. So the net effect of all this maneuvering is that consumption shifts to lower earners, and that usually leads to growth, at least temporarily.

Unbundling and Rebundling

Low-priced airlines like Spirit and Frontier have a model where they break all the components of a flight into separate products with their own price tag. This has the obvious advantage of letting customers pay for what they're actually going to consume, and the more cynical advantage that if you sort a list of flights by price, the unbundled fare will usually look cheaper. That model has struggled as bigger airlines did exactly the same thing, and now those low-cost airlines are converging on their competitors' pricing, too, by creating bundled ticket offers that include all the perks by default ($, WSJ). There are times when it's hard for companies to match their competitors' pricing model, but it's easier to replicate that than to copy other parts of the operating model, especially if there are multiple channels for selling the same product.

Attrition

According to data from Carta, the failure rate for startups started rising in 2022 and has been growing steadily since then, despite subsequent improvements in the funding environment ($, FT). There are a few factors here:

  1. It's one data provider, which might not be perfectly representative.
  2. There are two waves of losses to think about: first, companies that had been planning to raise money right when the funding drop started, and who promptly ran out. Second, there are the companies that were well-capitalized coming into 2022, cut costs to get to a more sustainable spending level, and then found that the growth they were getting with this lower spending didn't justify later venture rounds. There are some models that just don't work without a certain level of funding, and while that doesn't mean the business is invalid, it does mean that its existence is dependent on the financial cycle.
  3. "Startups" are a heterogeneous category. A lot of the growth in funding has gone to AI-related companies, and while the rest of the startup world isn't in the same funding deep freeze it was in 2022, aggregate numbers overstate the recovery.

Diff Jobs

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

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.

Find a Role

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.

Find Talent with Diff Jobs