Longreads + Open Thread

Longreads

• Paul Kedrosky and Eric Norlin speculate about the economic impact of AI on programming in a novel way: programmer compensation is a Baumol Effect where higher productivity in hardware forces software wages to rise, but specifically because they're a less productive yet essential complement. Reframing tech economics as a persistent hardware glut coupled with a software shortage is a powerful model. (And don't miss the LLM example where ChatGPT solves a programming puzzle with a nice hack rather than a plodding and literal approach.)
• And on the consumer end of LLMs, Steve Yegge writes a quick primer on how LLMs improve programming, with a focus on how limited context windows mean that LLMs can only ingest part of the context they need. (One question I have, which is well beyond my ability to answer: is there a way to reduce the token count needed to understand a codebase by structuring it nicely. If a given function takes two integers as inputs and returns a boolean, and that function has a descriptive name, maybe the code doesn't need to be part of the context window since that information is enough to deduce what it does. This, of course, puts a very high premium on descriptive code and on non-hallucinatory models, but if it increases the scope of the code an LLM can comprehend and add to, that could be a worthwhile tradeoff.)
• Andrew Walker interviews John Maxfield on banks. It's a bad idea to go through a volatile period in the market with no frame of reference, but it's nearly as bad to go through it with only the most recent analogue as a frame of reference—because a lot of what's changed will be in reference to what went wrong last time. Maxfield has been studying banks for a decade plus and can rewind through many past crises. One key argument: "Everything adjusts. The other thing to keep in mind is that we've been through nine major banking crises in the US history. We've been through many more minor crises or probably, I don't know, as many as two dozen crises total, when you can put both minor and major ones together. You know what always happens? It always goes back to just the way it was. It always does." And don't miss the point about going really in-depth with CEO interviews.
• Virginia Heffernan in Wired: I Saw the Face of God in a Semiconductor Factory. A beautiful look at the spiritual dimension of technology, both in the literal sense (who knew that so many people at TSMC are so devout?) and in the sense that any human accomplishment of sufficient scale will put people into a religious frame of mind. It's sometimes easy to lose track of how fast numbers compound: "Every six months, just one of TSMC’s 13 foundries—the redoubtable Fab 18 in Tainan—carves and etches a quintillion transistors for Apple. In the form of these miniature masterpieces, which sit atop microchips, the semiconductor industry churns out more objects in a year than have ever been produced in all the other factories in all the other industries in the history of the world."
• This Guardian piece on 3D printing is a good reminder that many bubbles are wrong just because they're early. 3D printing is growing, not as a consumer-facing product but as a way to quickly build customized, lightweight prototypes and parts: "Almost all – 99% plus – custom hearing aids are now 3D printed in acrylic resin, and have been for years. Additive manufacturing is widely used in dentistry: teeth aligners, which are increasingly taking the place of traditional wire braces, would be almost impossible without 3D printing. Adidas and Nike use the technology in their shoes. There are 3D-printed parts on all new aircraft and in a growing number of cars." It's also a way to shorten supply chains in terms of both distance and time, and since every supply chain operates at the pace of its slowest or least predictable component, that has many downstream effects.
• This week's Capital Gains covers bidder density, a surprisingly powerful concept. If you've ever wondered why you sometimes hear stories about companies scaling up with incredibly cheap ads in the early days of Facebook and Google, or wondered why investors pay so much for users when they're evaluating early-stage ad-supported companies, this piece should help. To get Capital Gains weekly, subscribe here. (It's free.)
• I went on the Moment of Zen podcast to talk about SVB, fractional banking, crypto, and social collapse with Antonio García Martínez, Dan Romero, and Erik Torenberg.

Books

• The Rise and Fall of the Conglomerate Kings is a series of quick profiles of 1960s conglomerateurs. Some of the conglomerate boom was just sloppy or aggressive accounting—companies reporting high earnings per share by buying businesses trading at a lower multiple and paying mostly in stock. But some of them were remarkably forward-looking: Gulf + Western built an overnight delivery network for auto parts and later diversified into, among other things, movies. (Apparently this was unpopular with people in the actual movie business; in Mel Brooks’ Silent Movie, the villain runs a conglomerate called Engulf and Devour.) The conglomerate story is partly one about mean reversion: Textron underperformed other conglomerates because it was mostly built around buying cheap companies in a tax-efficient way, but many of these bounced back, whereas other conglomerates that bought high-growth electronics companies in the 60s found themselves underperforming when defense priorities changed.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Riffing on this thread: how does AI change company structures? Can managers handle more direct reports if it's easier to automatically summarize communication? Or does rising efficiency per worker mean we actually need more layers of management to effectively interpret data and make decisions? Or does "middle management" stop being a meaningful distinction entirely when labor and capital are more directly fungible?

Reader Feedback

On last week's open thread, Calvin McCarter has thoughts on moral hazard:

It's interesting to look at situations outside the realm of finance where people did bad things in the past that need to be deterred in the future, but punishing those people would have too much collateral damage, and it's not even clear where to draw the line about who and what to punish because "Moloch does it." One example is how China addressed the Cultural Revolution after it was over. To enable economic growth, it needed to be clear that another Cultural Revolution would never be allowed to happen again. But sending this message by litigating responsibility would cause too much turmoil that would also detract from rebuilding the economy, and so it was necessary to basically forgive those responsible and memory-hole the whole episode. The solution in China's case was a pseudo regime change: bringing a Party victim of the Cultural Revolution, Deng Xiaoping, into power and allowing him to wholly restructure economic policy.

Analogously, one might argue that the Fed should do what is necessary to ensure financial stability now -- even if it sets bad precedent. But then, in the aftermath, instead of adding new limitations to Fed powers that are both of doubtful credibility (and thus still pose moral hazard) and also possibly overly restrictive during an actual emergency, one should instead create a new institution NotTheFed. Because NotTheFed is not the Fed, the bad precedents of the Fed no longer pose moral hazard, and NotTheFed can credibly pinky-promise to not follow Fed-set precedents. Meanwhile, because NotTheFed is a clean-slate institution, it will not be hamstrung by pesky 13(3) rules during the next crisis. In other words, this pseudo regime change in the regulatory system reduces moral hazard in the boom part of the cycle, while giving NotTheFed more flexibility during the next financial bust.

Of course, the problem with this scheme is that once NotTheFed sets bad precedent, you'll need to dismantle it and repeat the cycle with NotNotTheFed, which (logicians tell us) equals The Fed.

A Word From Our Sponsors

This newsletter is brought to you by Tegus, your modern research launchpad.

Investors waste too much time looking for actionable information across disconnected sources. Tegus streamlines the information you need to move quickly with conviction and make better decisions. With access to the largest transcript library, expert calls with no markups, driveable financial models, and searchable SEC filings, you’ll be the smartest expert in the room.

Right now, readers of The Diff can trial the Tegus platform for free.

Diff Jobs

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

• A VC firm using data science and ML to source and evaluate opportunities is looking for a software engineer to lead their data engineering efforts. (Menlo Park, CA or NYC)
• A well funded seed stage startup founded by former SpaceX engineers is building software tools for hardware engineering. They're looking for a full-stack engineer who enjoys working with customers to design and build software. (Los Angeles)
• A company building zero-knowledge proof-based tools to enable novel financial arrangements is looking for a senior engineer with a research bent. Ideal experience includes demonstrations of extraordinary coding and/or math ability. (NYC or San Diego preferred, remote also a possibility.)
• A new health startup that gives customers affordable access to preventative care and lifestyle interventions seeks a founding engineer. 7+ years of JavaScript experience preferred (TypeScript is ideal), and payments experience is a plus. A great opportunity for anyone excited to make healthcare better by treating problems cost-effectively before they're catastrophic. (US, remote; Austin preferred)
• A profitable AI startup is looking for ML engineers to help build new services to help small companies accelerate their growth. (SF)

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.