Longreads- Finance twitter fixture Quantian argues that prediction markets do not show probabilities, though the main effect he talks about is that the market structure makes it more rationally to bet in favor of long-shots than to bet against them, while what prompted it was that the roughly 50/50 election outcome is priced closer to 60/40 on Polymarket. So the specific argument doesn't apply to the case people are talking about, but it does raise an important point about prediction markets: it pays much better to buy a long-shot on which you have conviction than to take the other side of the trade even if you have equal conviction: "[P]eople are not going to take sub-tbill returns to buy a 98% “no” contract expiring in 6 months." This is a solvable problem: prediction market bettors aren't buying assets, just posting collateral for futures contracts, so it's reasonable for them to earn interest on that collateral. This can interfere with the market's own economics, but it also makes prices more accurate. In that model, buying a "NO" contract on a long-shot bet is basically picking up a bit of extra interest, along with some downside risk. And that means that one thing real-money prediction markets need is to be roughly as trustworthy as other financial institutions. If the most sophisticated market participants are applying an extra 10% hurdle rate to prediction market strategies to account for the risk that the market blows up, and the least sophisticated ones are sure RFK has a chance or that Harris will be swapped for Clinton at the last second, there will be plenty of obviously bad prices on the platform, which lowers the utility of the accurate ones.
- In The Atlantic, Marc Fisher writes about how hard it is to deal with shoplifters. There's a weird narrative that companies are faking problems with shoplifting as a cover for bad management decisions, which would work if the main effect of shoplifting showed up in store closures. But it also shows up in the form of more products being locked away until customers ask for them, and very few companies want to provide arbitary speedbumps for purchases, especially as e-commerce gets steadily more convenient. One of the most striking data points from the article, though, was how solvable this problem is: one store cut its daily theft loss from $200-300 to $75 by buying software that scanned security cameras for suspicious behavior. It has plenty of false positives, but catches many shoplifters in the act.
- Nabeel Qureshi reflects on Palantir. The best part of this piece is the discussion of forward-deployed engineers and product engineers. The basic Palantir loop is that customers articulate a problem, FDEs work with them on site and hack together some kind of solution that works, but not perfectly or not scalably, and then everyone compares notes and the product engineers figure out which parts of the FDE workload are getting duplicated and should be done right exactly once instead. One way to look at this is that the whole purpose of the company is to provide the optimal translation layer between maker's schedule and manager's schedule: if your customer wants their supply chain unkinked right now, you don't want to be a perfectionist about every detail of the code. But if you're writing the once-and-final version of your grand system for analyzing arbitrary supply chains, you don't want to ship a half-baked version because someone set a deadline more in reference to their own business needs than to how long the project really takes.
- Dario Amodei has a great essay on the potential upsides to AI. One of the important points this piece makes is that the pace at which AI can impact the rest of the world is mostly set by the rest of the world, especially at the limit, whether that takes the form of slow construction time for power plants and datacenters or because humans in the loop operate at human speed. This is both a reason to be more cautious about the upside of AI and less cautious about the downside: ending the world is an operationally-intensive task! The piece has some (speculative, but compelling) ideas for how AI could impact physical and mental health. Many of these fall into the same kind of extrapolate-the-line thinking that some AI skeptics hate, but calling for an inflecting trend to flatten or mean-revert is just as much of a judgment call as betting that it will continue.
- In Bloomberg, Samanth Subramanian looks at companies like Sportradar, that provide real-time odds for gambling. As with crypto, this increasingly-legal business gets to speedrun lots of things that took longer in their original incarnation: going from manual data collection to building data feeds to buying proprietary ones, for example. There's also a flavor of happy accidents, like when a bug in MySpace turned out to allow users to customize their profile pages. In this case: "During an Australian Open tennis match featuring Boris Becker, his staff forgot to close bets when play began, only to see activity surge tenfold." And there's even a recapitulation of the evolutionary path of big platform hedge funds: when Sportradar started, it was just scraping different bookies' websites to do simple arbitrage bets. But now it's gathering every possible bit of alternative data in order to be the first to correctly update a fundamental view.
- In Capital Gains this week, we covered the principal-agent problem, with thoughts on how the older model of corporations—a one-off way for the state to delegate certain matched privileges and responsibilities—had some things going for it.
BooksThe Money Trap: Lost Illusions Inside the Tech Bubble: There are business books about the person at the top of the org chart, because you can tell a story about a recognizable institution and still have a main character with the sorts of emotional highs and lows that make the story more than a zippier summary of the last few 10-Ks. And there are memoirs from line workers, usually doing some combination of celebrating how much fun the chaos was or feeling a bit bitter about what they have wrought. The Money Trap is a rarer one, a first-hand account of what working at Softbank was like from someone who was close to, but not at, the very top. Alok Sama was an investment banker who was tapped to be Softbank's CFO, putting him in a great position to talk about Softbank's two core competencies of making massive investments in high-growth companies and using these investments as feedstock for some high-wire acts of financial engineering. The book does have some dramatic tension: there was a strange whisper campaign, never quite resolved in the book, implying that Sama had engaged in double-dealing. This plotline adds a few interesting beats to the story—some corporate drama, betrayals for the greater good, meetings with ex-spies who now work in corporate intelligence. But it doesn't really get resolved. Worldly success can have a big impact on the particular sources of stress in your life, without changing much about the overall level. Sama followed the career track that lots of people say they're going to follow but can't quite follow through on: he spent some time in finance, made a big pile of money—then got an MFA so he could focus on more artistic pursuits. And the book definitely has the flavor of someone making his first big attempt to write (Sama did have a blog before, written in the same peppy style). If you've followed Masa Son's own career for any length of time, one of the recurring questions is how he hasn't managed to blow up just yet. And a great way to learn about that is to read a firsthand account from the person whose job was to make sure the Softbank check didn't bounce.
Polostan: Neal Stephenson writes three kinds of novels. Traditionally-futuristic science fiction, like the riotous anarcho-capitalism of Snow Crash; science-fiction-but-it's-happening-roughly-now (REAMDE, Seveneves); and historical fiction that would have been experienced by the characters as something out of science fiction (the historical parts of Cryptonomicon, all of The Baroque Cycle). Polostan is the latest entry in a new series on the latter. It's set in the 1930s, in the US and USSR, at a time when both places were recognizably modern but still had plenty of weirdness. One thing the book gets across is how existentially frightening the 1930s must have been to the American establishment. The USSR showed that it was possible for a country to undergo a communist revolution, and then to start invading its neighbors, industrializing, collectivizing farms, etc. (The single most harrowing part of the book is a description of what it might have been like to visit Ukraine during the Holodomor.) And at the same time, US-based revolutionaries were feeling feisty, too. What most people didn't know, but which a few characters in the book are in a good position to recognize, is that changes in weapons technology would make these kinds of issues even higher-stakes than they were before. What this book delivers is what Neal Stephenson's historical novels promise: a tour of a period in history that you may not be especially familiar with, alongside a very well-informed tour guide who has a gift for turns of phrase. What it doesn't deliver, so far, is a plot that makes very much sense. Both the US and USSR-set scenes have a fever dream quality to them. The book is part of a series, though, so treat it more as an enticing first entry than as a complete work. Open Thread- Drop in any links or comments of interest to Diff readers.
- Are there any other good examples of an industry rapidly running through the evolutionary cycle that its peers did more slowly?
Diff JobsCompanies in the Diff network are actively looking for talent. See a sampling of current open roles below: - A well-funded startup that’s building the universal electronic cash system by taking stablecoins from edge cases to the mainstream is looking for a senior full-stack engineer. Experience with information dense front-ends is a strong plus. (NYC, London, Singapore)
- A company building the composable open source middle-layer for the financial system is looking for a pre-sales solution engineer to help supercharge growth. (NYC, Paris)
- A well funded startup founded by two SpaceX engineers that’s building the software stack for hardware companies is looking for a staff product manager with 5+ years experience building and managing data-intensive products. (LA, Hybrid)
- A team of former SpaceX, Air Force, and MIT engineers are looking for a senior mechanical design engineer to help mass-produce rapidly deployable nuclear microreactors to enable energy superabundance.A team of former SpaceX, Air Force, and MIT engineers are looking for a senior mechanical design engineer to help mass-produce rapidly deployable nuclear microreactors to enable energy superabundance. (LA)
- An AI startup building tools to help automate compliance for companies in highly regulated industries is looking for a director of information security and compliance with 5+ years of info sec related experience at a software company. Experience with HIPAA, FedRAMP a plus. (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.
|
