Longreads + Open Thread

Longreads

Programming note: The Diff is off Monday, back Tuesday.

• A Letter a Day has a fascinating piece from both participants in the deal to sell a Cryptopunk NFT for ~$7.5m. As the author, KG, puts it, "For those of you who are pro-crypto, these pieces will explain why crypto != tulips and why NFTs = art. For those of you who are anti-crypto, these pieces will explain exactly why the crypto bubble happened."
• Fascinating Chinatalk interview with Tyler Cowen on AI. One of many interesting points: for people who grow up knowing how to use AI, the world will be very different. But for people who grow up using AI without understanding it, the world will be much more like the pre-Internet world, where some kinds of information and some forms of creative expression are simply unavailable, not because they don't exist, but because it's not obvious how to access them.
• This is an older piece, from 2008, profiling the rise and fall of David Berry, a trader who worked for Canadian bank Scotiabank. It's a good meta-story on trading careers: he traded preferred shares, which were something of a backwater, but eventually controlled 62% of the market, and cut a deal with his employer giving him a share of trading profits. His downfall seems mostly to have been from that: 20% of trading profits turned out to be a lot more than the CEO of the bank was making at the time, so the deal got negotiated down and he was eventually fired on a vague compliance pretext. A decent career path in finance is to choose something that's gotten less popular in the last decade, but probably won't go away entirely during your lifetime; being the hardest-working person in a given field is mostly a matter of choosing a less competitive field. But hiring and firing decisions aren't purely monetary, and sometimes a manager is more willing to accept lower total profits than to accept an underling who can afford a nicer house and car.
• Chris Moore, a producer with a long roster of fairly successful movies, writes about how the shift from box office and DVD sales to subscriptions has made his job more difficult. (There's nothing quite like a massive one-time windfall to reset expectations about the status quo—when the music industry was worried about piracy, it wasn't just because of the convenience of Napster and the like, but because they'd earned such massive profits from the CD.) It's a good piece, and a depressing one: from a business perspective, subscriptions are better than one-time purchases because the buyer has to decide to stop rather than start buying. But from a creative perspective, it means there's a premium on volume production of good-enough rather than on excellence. And the more successful a subscription platform is—or, at least, the bigger it gets—the more churn rather than new subscribers dominates the calculation, and thus the stronger this pressure gets.
• Alex at Continuous Variation has an insightful and intermittently beautiful piece about large language models. Are they a tool, or an art form? (And is a CPU a kind of sculpture? And what is the point of poetry?) Part of the unmeasured social value of new technologies is that they give us a new frame of reference for understanding older things that we've gotten used to but haven't had to strictly define until there's something new to compare them to.
• And elsewhere in the Diff universe: our spinoff newsletter, Finance FAQ, breaks down one big idea each week from the worlds of finance, tech, and corporate strategy. This week's issue: how and why to think about the world in terms of supply chains. And you can sign up to read more here.

Books

• Master of the Senate: The third volume in Robert Caro's LBJ biography, this book follows LBJ from being a first-term senator to being the majority leader, with a substantial portion of the text dedicated to the passage of the 1957 Civil Rights Bill. It's a very good look at what political maneuvering looks like day-to-day: threats, favor-trading (lots of this!), and careful media manipulation. One important takeaway is that the final voting totals for a bill are not a good indicator of how contentious it is: that civil rights bill 1) barely got through, and only thanks to extensive negotiations and parliamentary maneuverings, and 2) ended up getting 72 votes in the senate, with 18 opposed. Looking at the raw vote totals, it looks like a popular choice, but looking at the details it was a dicey one. That's a good thing to keep in mind with other kinds of successes: if something remains a 50/50 proposition for a long period, it's going to converge on 100% or 0%, and either way, lots of people will switch sides at the last minute in order to be aligned with a winner. And part of politics, at least as LBJ practiced it, is creating the perception of an overwhelming majority so people will join it.
• Investing Amid Low Expected Returns: This is a good follow-up to Expected Returns, an essential part of the finance canon. People who select individual stocks or bonds are trying to get paid based on some kind of information/analysis advantage, but this book takes a step back and looks at the question of which assets produce what kind of risk-adjusted return, and under what circumstances. It's important to know what environment you're operating in even if the actual operations you do are focused on a narrow slice of that environment. In this case, the message of the book is somewhat dreary: after a generation-long decline in interest rates, future returns will be lower than the recent past indicates, both because historical numbers include windfall gains and because starting yields are lower. So the great capital accumulators of the next decades will include a larger than usual share of people who made their money relatively young and then lived a long time while investing cautiously.

Reader Feedback

From the comments on last week's post looking at how AI's economics might be closer to those of the steel industry than the software industry, here's another good analogy from Calvin McCarter:

Re: AI and steel, another parallel I think about is electric vehicles, Tesla, and the auto industry. It's quite possible that, because EVs are so simple, the overall profits of the auto manufacturing sector will decline. OEMs were complex system integrators for an insane number of parts suppliers, but in the future many components will no longer need to be made, and integration of the smaller number of remaining components will be capital-intensive yet commoditized. Similarly, a lot of software development is basically making various bespoke software-glue components, and AI could automate this away. And making the AI that does this will be capital-intensive yet commoditized. To the extent that Tesla's still-high valuation is justified, it's because it makes its own batteries, cuts out the car dealership middlemen, and maintains a charging network. Similarly, the money in AI could be derived upstream (making AI chips and running datacenters), developing a brand & loyal relationships with customers (designing sticky interfaces that people love and trust), and offering ongoing services to those customers (various products built around AI).

A Word From Our Sponsors

Protecting your money during a downturn takes more than a robo-advisor and a prayer

As many have recently learned the hard way, the chart doesn’t only go up – it can also go down. Picking stocks is difficult. And after a decade+ bull run, the market is changing. Is your portfolio?

Composer offers an array of professionally-created investment strategies that trade based on logic and data. Take advantage of advanced algorithmic trading — no PhD required, just point and click.

With Composer, your portfolio reacts to the market, rather than emotions and sensationalized tweets. It can move your portfolio into its best performers when the market is doing well, and hedge risk during volatility.

Create a free account in 90 seconds

Open Thread

• Drop in any links or comments of interest to Diff readers.
• An upcoming Diff piece will look at the increased abstraction in servers (from a physical server in a closet to something you own in a datacenter to a statistical approximation of a single computer rented to you by a hyperscaler to Lambda, Cloudflare Workers, etc.). Any readers who have experience with or strong opinions on this are encouraged to reach out. Just leave a comment or reply.

Diff Jobs

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 fintech startup that lets investors trade any theme as if there were an ETF for it is looking for a senior backend engineer. (NYC)
• A crypto proprietary trading firm is actively seeking systematic-oriented traders with crypto experience—ideally someone with experience across a variety of exchanges and tokens. (Remote)
• A company using Web3 to decentralize customer loyalty programs is looking for a founding senior engineer with Solidity experience and an interest in brands and the arts. (Brooklyn)
• A high-growth provider of market data to both retail investors and institutions is looking for a staff-level frontend developer. Experience scaling a product/platform is ideal. (US, remote)
• We have multiple roles for investment researchers with alternative data experience: an alt data consulting firm is looking for analysts with experience working with data and a fundamental research bent (Remote), and a systematic hedge fund seeks alternative data researchers with more of a programming/data science background (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.