Byrne Hobart
Today's issue of The Diff brought to you by our sponsor, Warp.
Longreads
- Ben Ryder Howe writes a half-praising, half-concerned piece on Costco. Plenty of fun details (they account for half of global cashew sales, supplied by more than a million farmers). In rankings of which companies people trust, Costco comes out on top. They've managed to build a model that's aligned with the revealed preferences of shoppers—i.e. big spending on bulk items—but that's less aligned with what shoppers want to want.
- David Brooks on Tom Wolfe. Wolfe was a dangerously good observer of social norms and coiner or popularizer of phrases ("the right stuff," "good ol' boy," "master of the universe," "push the envelope"). But he was so good at this that many modern social observations seem peculiarly tied to 1980s New York, 90s Atlanta, etc. What's especially appealing and frustrating about Wolfe is that, even if you can't mimic the style, you can capture the substance: his technique was to spend a lot of time with his subjects and take copious notes. He was unusually observant, but his work was also the output of a lot of observing.
- Tyler Cowen interviews Nate Silvers. Fun throughout. It has a good entry for the "history is full of random contingencies" files: Silver got into politics as a consequence of being a professional online poker player and suddenly needing to find a new job when online poker was de facto banned in 2006. (Perhaps this says something about the social utility of gambling. On the other hand, maybe bad poker players subsidized his statistical education, which was then put to higher leverage uses.) One great thought experiment from the piece is how the typical competitive poker player today would compare to the best in the world fifty years ago, with Silver arguing that advances in poker theory would probably make them best in the world. There's also a good bit on incentives, particularly companies' incentives around vanity assets. FiveThirtyEight's parent company is Disney, and when Disney was doing well, they were indifferent to the site losing a bit of money each year. As a result, FiveThirtyEight hadn't exercised it’s commercial muscles much, so when Disney started focusing more on the bottom line, they ended up destroying the value of that asset—at this point, the biggest contributor to FiveThirtyEight's brand is that not everyone knows Silver is no longer involved.
- T. Greer on the Silicon Valley canon, and how tech people differ from political people in their reading habits. (There will be a future Diff piece on the most politically impactful works of mass-market nonfiction. As a preview, one reason the US's pandemic preparedness wasn't even worse was that George W. Bush once read a book about the Spanish Flu pandemic while on vacation and insisted that there be some kind of contingency plan.) This piece makes a strong case that tech is more bookish than politics (importantly, this does not mean that tech people are necessarily more intellectual) and has some interesting arguments on the different kinds of depth and breadth that exist in both domains.
- In Speedwell Research, a meditation on the purpose of analysts publishing price targets for stocks, and the incentives that drive them. There is a lot more repeat business in mostly having an incremental view instead of a wildly off-target one, especially if you are occasionally willing to take a big swing. And it's also intellectually honest—if every price target is extreme, it implies that either the world has gone crazy or the analyst has. But it's also true that there's a financial incentive to hug close to consensus and to be biased positively. As long as people understand those incentives, the work analysts do remains useful, but if they don't, the economic function of analysts is to transfer money from people who aren't in on the joke to the ones who are.
- This week in Capital Gains, we're looking at distribution: which companies care most about who owns the customer relationship, and which industries are defined by the struggle to figure this out?
- In The Riff this week, we're talking about why tech antitrust is hard, how some companies offer workers the safety net they'd vote for if they could, the importance of intangibles, and elites. Listen with Twitter/Spotify/Apple/YouTube.
Books
Why Machines Learn: The Elegant Math Behind Modern AI: Before reading this book, I "knew" that the way many tools in AI work is by converting a real-world question into a linear algebra problem and then solving it, but beyond reciting that line I was not in a great position to describe exactly what that entailed or how it worked. Why Machines Learn is a very good guide to what, practically, is going on when I ask ChatGPT a question and get back a well-formed answer.
The book borrows its structure from Sophie's World: chapters tend to start with some narrative, introducing a few personalities and sharing some fun stories about them before moving on to the more academic material. There are lots of details I hadn't known about—the first known mention of linear algebra comes from a two thousand year old Chinese text; the Intel 4004 was designed by one of the early pioneers of neural networks; Geoffrey Hinton's academic journey was longer and weirder than I'd realized, including a brief period studying philosophy followed by a stint as a carpenter.
The book does a good job of walking through the actual math behind AI. It's hard to know what the right balance is, here—if you don't have any previous exposure to linear algebra or calculus, you'll probably be lost, but it doesn't have any more math than necessary, and relegates some of it to appendices.
Overall, this is a great book for converting some of the intuitions about how AI models work into more concrete math. It's also a great work of intellectual history: many different threads from many researchers over multiple generations got us to where we are today, and in some cases the work that has great practical importance today was a just a theoretical construct hardware was more expensive.
Open Thread
- Drop in any links or comments of interest to Diff readers.
- The Nate Silver interview asks whether Silver would have been the best poker player in the world if he went back in time half a century but knew how modern poker is played. What are some other domains where that would or wouldn't hold true?
A Word From Our Sponsors
Still running payroll the old way?
Warp is built for founders and startups who have less time on their hands. Warp bundles payroll, compliance, and benefits into a single dashboard, automating annoying registrations that come with hiring across states or offshore.
Unlike bloated platforms pushing unnecessary features like applicant tracking systems and performance reviews for your employees, Warp is laser-focused on the essentials. Get in, get it done, get out.
Say goodbye to endless payroll reminders and cryptic notices from tax departments. Reclaim your time and focus on what truly matters – growing your startup.
Get started now at joinwarp.com/thediff.
Diff Jobs
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- Growing pod at a multi-manager firm is looking for new quantitative researchers—strong STEM academic/work background, but no finance experience required, 250k+ (NYC)
- A new data startup that's seeing early traction bringing transparency to a $1.5tr asset class is looking for an experienced founding engineer. This is a full-stack role; Python required, React experience a plus (West Palm Beach, FL)
- A startup building a new financial market within a multi-trillion dollar asset class is looking for a data operations specialist (NYC).
- A fast growing start-up enabling SMB focused software businesses to offer accounting functionality to customers with an API call is hiring a founding full-stack engineer. (SF)
- An AI startup building unique tools for compliance-focused companies is looking for a director of information security and compliance with 5+ years of infosec related experience. 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.