The Diff - Longreads + Open Thread

This issue of The Diff is brought to you by our sponsor, Intercom.

Longreads

• I'd heard allusions to a study on this years ago, but hadn't tracked it down until now: a survey on how consumers think about inflation showed that they attribute their nominal wage growth to their own efforts, even if that wage growth is mostly driven by inflation. This is a behavioral claim that bolsters the case for expansionary policies even if they lead to higher inflation: people like wage increases, even if they're nominal, and they hate pay cuts, even if they end up better-off in real terms. Sometimes the most cost-effective way to rebalance wages between different workers is to raise price levels and leave some people behind in real terms. Via Paul Krugman's Substack.
• Rohit Krishnan wrote the case against AI hitting a wall a week ago, which turned out to be very well-timed given that OpenAI has a model that massively blows past the ARC eval, with performance close to a human engineer albeit at a cost of ~$3k per task. One point he makes on data shortages is that there's a lot of data we just don't bother to collect, but that we could theoretically start using—conversations, images from security cameras, sensor data from vehicles and machinery, all of this is information about the real world that models could start building inferences on. Text tokens were easy because they were abundantly available, initially for effectively zero cost, but they're not the last step.
• S. E. Smith on what link-rot does to culture. One of the paradoxes that the millennial generation experiences is that a lot of the weird media we remember from our childhoods is accessible on-demand via YouTube or BookFinder. And then a bunch of our formative early-adulthood media experiences are apparently gone forever; there are weird once-viral videos that are either hard to track down or have vanished completely, and old blogs that exist only as spotty excerpts in Archive.org. The piece complains about AI (with the usual hazy scope 1 estimates that aren't offset by scope 3 estimates of how much energy we save from using these same tools), but AI does provide a more permanent means of storage. Even if your middle school Tumblr is eventually not worth the cost of storing and hosting it, the tokens will end up in a model somewhere, and the spirit of the content will live on forever.
• The Terminalist has a great writeup of FactSet, but really a writeup of how data economics replicates other business' U-shaped value capture curve. As the piece notes, PC assembly had lower margins than either making CPUs or making the software that ran on computers, and similarly a company that ingests and filters data has worse economics than a business that collects the data through a network effect-driven product like an exchange, or that converts it into useful conclusions. This is one of the better pieces I've read on the economics of a single business—I learned a lot about an industry that I worked in full-time for half a decade.
• Why academics leave Harvard for jobs in industry, from Harvard Magazine. The money is nice, but many of the complaints aren't so much about money as the time burden required to get it. Sometimes it's astonishing to think that every year, we spend multiples of the equivalent brainpower of golden age Athens or renaissance Florence on the task of applying for grants.
• In this week's episode of The Riff, we cover pod shops, whether or not this is the top of the market (it feels that way most of the way up), how to price better AI, and more. Listen with Spotify/Apple/YouTube.
• In Capital Gains: it's hard to give and receive life advice because people vary so much in how much effort, in which tasks, counts as hard work. Just in time for New Years Resolution pre-planning!

Books

The Nvidia Way: Jensen Huang and the Making of a Tech Giant: it feels like these kinds of books used to be a lot more common, but it's frustratingly hard to get a book-length narrative telling the history of a major company. But these are valuable, because growing companies get to continuously make a better first impression on a bigger group of people until they peak. That makes it a useful exercise in perspective to look back and see Nvidia as a chaotic company whose first launch flopped and which came close to failure many times. Reviewed in more detail here ($).

Open Thread

• Drop in any links or comments of interest to Diff readers.
• The most important story of 2024 was the same one as 2023, with a more demanding set of benchmarks and bigger capex numbers. But are there other trends that inflected this year? Looking at stock charts, there’s a lot of enthusiasm for quantum computing and a long tail of equally not-yet-commercial products. What else? 

A Word From Our Sponsors

In 2025, every company is going to recognize the need for a customer service AI agent. Why? Because AI agents provide customers with better answers and faster resolutions, all at a lower cost to you. Fin by Intercom is the world’s leading AI agent for customer service. Fin can resolve up to 80% of questions in an instant—which customers love. And you can lower your cost to serve by 96% and finally get off the service rep hiring treadmill. The best part is that Fin is now available on every customer service helpdesk. That means you get all of Fin’s benefits without
having to change platforms. See why more than 3,700 businesses have hired Fin for AI-first customer service.

Diff Jobs

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

• A premier proprietary trading firm is looking for smart generalists to join their investor relations team, working with external investors, rating agencies, and the internal finance team. Investment banking and/or investor relations experience preferred. Quantitative background and technical aptitude a plus. (NYC)
• A hyper-growth startup that’s turning customers’ sales and marketing data into revenue is looking for a product engineer with a track record of building, shipping, and owning customer delivery at high velocity. If you like to build, this role is for you. (NYC)
• A Google Ventures backed startup founded by SpaceX engineers that’s building the data infrastructure and tooling for hardware companies is looking for a staff product manager with 5+ years experience, ideally with AI and data intensive products. (LA, Hybrid)
• Ex-Ramp founder and team are hiring a high energy, full-stack engineer to help build the automation layer for the US healthcare payor-provider eco-system. (NYC)
• A company building the new pension of the 21st century and enabling universal basic capital is looking for a head of capital markets to help scale capital markets (debt and equity) efforts. Background in banking, hedge funds, or fintech, including embedded lending, preferred. (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.