Longreads + Open Thread

Longreads

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

• Human Invariant interviews a successful YouTube scriptwriter. Lots of great thoughts on the structure of the media industry, and on what equilibrium YouTube creators are aiming for. For example, on thumbnail photos: videos get views based on the thumbnail, but their ranking ultimately depends on watch time. So it's possible to be too good at thumbnails, relative to the content. So "In a sense, the thumbnail has to be "unsatisfying" but with the promise that you will get satisfied if you click.'" The piece also talks about how fragmented the YouTube labor market is: when everyone's aiming to get the attention of specific kinds of viewers, it makes the skills involved less fungible. (Via Trevor McKendrick's newsletter.)
• Alex Kesin asks: would the plot of Breaking Bad work with today's cancer treatments. Answer: no, Walter White would probably pay about $14k-$20k out-of-pocket, and his five-year survival probability would be 50-60% instead of 10-15%. Which, as the piece notes, is not something the scriptwriters would have had trouble dealing with; they could just give him a different kind of cancer instead. But it's a fun way to do a cohort analysis on healthcare: we pay a lot for it, and we pay pretty sizable markups to pharmaceutical companies. But, in the end, they're at least directionally keeping their side of the bargain.
• Scott Alexander analyzes the vibecession: the median American is making more money than ever in real terms, but consumer confidence is weak. He points to housing as the category that has seen the biggest cost increases, particularly for people who live in cities with a big media industry and whose views have more weight in the discourse. But one thing to note here is that even with more optimal housing policy, we'd still expect housing to take a growing share of people's budgets. New York rent is the price you pay for access to New York's restaurants, museums, stores, parks, etc., not to mention the labor market and dating pool. The better other stuff gets, the more it makes sense to pay a premium to access it. And the more efficiently housing is allocated, the more bidders there will be for a chance to live in the most desirable places. So this is just one more instance of the intra-Ivy League class war dynamic that characterizes so much discourse: the people writing op-eds are in the bottom 10% by income of the class of '0-whatever, and this colors their and thus everybody's opinions about tech founders, private equity executives, and the like.
• Max Chafkin and Dina Bass on the incredible, ubiquitous Excel. It has half a billion users, making it the biggest single software product by paying user count. Excel is sticky partly because users seamlessly transition from using it as a calculator to using it as a programming language. And it's an interesting language paradigm—basically an IDE whose default interface is a debugger showing you the inputs to your functions and their outputs, while hiding the functions themselves. That might be the gentlest way to introduce someone to programming: by having a coding interface that's focused on showing you potentially buggy outputs rather than the logic that produced them.
• Anthropic has a paper estimating the impact of current models on productivity, by looking at how long users took to complete various tasks with Claude, and comparing it to how long those tasks usually take. It's an example of recursive self-improvement, albeit with a human in the loop: there wasn't a clean dataset for how long it takes to prepare an invoice, fix a printer, build a financial model, etc., so—they asked Claude to guess! Which means that if you're an AI-skeptic, you can dismiss this piece because it's an AI model telling you how great it is to use AI models. The paper is full of caveats and cautions, but in at least one case they're too cautious: they note that the aggregate productivity gains they expect are a function of AI adoption, but if LLMs also make it easier to make a case for using more LLMs, the pace of that adoption is partly tied to capabilities-squared: when they get better in general, they get better at telling you so.
• In this week's Capital Gains, we ask if markets can be too efficient. Getting instantaneous feedback seems like it would make long-term planning harder, but that feedback is about the long-term impact of whatever caused prices to change. We also consider the meta-efficiency question: the best way for prices to eventually approximate intrinsic value is to have a system complex enough that from time to time they very much don't.
• On ReadHaus, a user shared this chat asking about the Buffett Indicator, of market cap to GDP. I'm actually not a big fan of it: it makes some sense if countries are pretty similar to each other and over time in how big their equity market is, but if there are more IPOs, and if a country wins more exchange listings for multinationals, it'll tend to show misleading results. You can stress-test it in other ways: if there were a giant surge of PE deals that took half of the US's market cap private, and the remaining stocks ran up 50% in response, the Buffett Indicator would say that market cap to GDP had declined and that stocks were cheaper, but they'd actually be more expensive.

A Word From Our Sponsors

Imagine being a fly on the wall while two industry leading portfolio managers candidly debate your largest position. That's the Trata experience.

We're building the world's largest repository of buy-side knowledge. By connecting 125+ funds with over $150bn in combined AUM and counting, Trata captures the nuance that expert call transcripts and sell-side reports miss. Our platform hosts and transcribes anonymous 1:1 debates between the sharpest bulls and skeptics, giving you a front-row seat to a real-time evolution of the institutional thought process around a name.

Featured by Matt Levine and backed by YC and others ($3M+ raised), Trata is rapidly becoming the standard for checking your blind spots. Stop relying on commoditizing expert networks and the same podcasts everyone else listens to, start listening to the raw, unfiltered dialogue of the market’s highest-signal participants.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Buffett used market cap to GDP. Alan Greenspan called economic cycles by looking at scrap metal prices. Li Keqiang didn't trust the CCP's statistics and made his own GDP indicator based on electricity use, rail freight, and bank loans. What are some other fun alternative indicators for the state of the economy? (Bonus points if they help calibrate whether or not we're in a vibecession.)

Diff Jobs

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

• Ex-Bridgewater, Worldcoin founders using LLMs to generate investment signals, systematize fundamental analysis, and power the superintelligence for investing are looking for machine learning and full-stack software engineers (Typescript/React + Python) who want to build highly-scalable infrastructure that enables previously impossible machine learning results. Experience with large scale data pipelines, applied machine learning, etc. preferred. If you’re a sharp generalist with strong technical skills, please reach out. (SF, NYC)
• A Google Ventures-backed startup founded by SpaceX engineers that’s building data infrastructure and tooling to accelerate product development for hardware companies is looking for a deployment strategist to ensure that the platform creates maximum value for customers with sophisticated engineering organizations across aerospace, transportation, renewable energy, and more. (LA, Hybrid)
• Series-A defense tech company that’s redefining logistics superiority with AI is looking for a MLE to build and deploy models that eliminate weeks of Excel work for the Special Forces. If you want to turn complex logistics systems into parametric models, fit them using Bayesian inference, and optimize logistics decision-making with gradient descent, this is for you. Python, PyTorch/TensorFlow, MLOps (Kubernetes, MLflow), and cloud infrastructure experience preferred. (Salt Lake City or NYC)
• A hyper-growth startup that’s turning the fastest growing unicorns’ sales and marketing data into revenue (driven $XXXM incremental customer revenue the last year alone) is looking for a senior/staff-level software engineer with a track record of building large, performant distributed systems and owning customer delivery at high velocity. Experience with AI agents, orchestration frameworks, and contributing to open source AI a plus. (NYC)
• Well funded, Ex-Stripe founders are building the agentic back-office automation platform that turns business processes into self-directed, self-improving workflows which know when to ask humans for input. They are initially focused on making ERP workflows (invoice management, accounting, financial close, etc.) in the enterprise more accurate/complete and are looking for FDEs and Platform Engineers. If you enjoy working with the C-suite at some of the largest enterprises to drive operational efficiency with AI and have 3+ YOE as a SWE, this is for you. (Remote)

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.