Longreads

• Jeff Dean, Sanjay Ghemawat on writing faster code. It would massively undersell this to call it a list of good C++ tips (I personally haven’t tried to learn C++, because the impression I got is that it’s very easy to learn enough to be a net counterproductive C++ programmer and hard-but-worthwhile to get good). Instead, it’s closer to Jiro Dreams of Sushi: compressed conclusions from people who’ve thought very deeply about something complicated, and have gotten in a vast number of reps. While the implementation details are language-level, the general attitude is not—I suspect that reading this is good for your performance writing Python or playing Factorio.
• Arpit Gupta and Alex Imas on training a transformer-based model to predict policy responses to changes in the economy. This piece is nice because it’s a case of better machine learning techniques pushing out the efficient frontier of economic research: in general, if you use a well-specified model, there are more ways you can manipulate it algebraically to say interesting things (the fact that EMH can be restated as "here’s an exhaustive list of the reasons prices wouldn’t necessarily reflect fundamental value" is a nice intuition pump, for example). And if you use real-world data, you’re looking at a blurry and variably laggy snapshot of what’s going on in the real economy, what economic actors think is going on, what they think other economic actors think is going on, what policymakers think is going on (and what they think everyone else thinks), etc. So, there’s a tradeoff between being fully descriptive and having any concrete view of the underlying mechanisms. This tradeoff will always exist, but what can change is the shape of the curve, and where it intercepts with the X and Y axes—we may have slightly more rigorous theories if we constrain them to the real world, or our beautiful models may correspond less loosely to what actually happens.
• In the New Yorker, Shayla Love asks whether cognitive dissonance really exists. There’s a famous study that tracked members of an apocalyptic cult founded in suburban Chicago in 1954. They predicted a UFO-driven apocalypse/rapture by the end of that year, and, when it didn’t happen, they actually doubled down and started proselytizing! Like basically any mid-century study of human behavior that leaves you, the reader, feeling superior, these results were basically fake: the researchers were regular attendees at the group’s meetings ("at some Seekers meetings, half of those present may have been infiltrators,") and they regularly interacted with and egged on their test subjects. So, you might think this blows up the concept entirely. But not so fast! There are other, better-documented cases of people holding on to ideas in the face of evidence against them, such as the grad students who worked under the author of that UFO cult case study, one of whom is still alive and gave an interview on the topic, arguing that even if apocalyptic cults don’t actually proselytize after their doomsday date has passed, this wouldn’t invalidate the theory. Which is true; it wouldn’t. But if it were true, it would also mean that we’d have a more nuanced sense of what cognitive dissonance is: we update slowly, but for the most part we don’t just blindly continue to believe something after there’s significant evidence against it.
• The Terminalist has a fun history of the financial data business. Financial messages have such a high ratio of potential real-world impact to total number of bits ("Sell 1,000 shares XYZ" and "Sell 100,000,000 shares XYZ" have only a small gap in bit count, but mean very different things). So they tend to push the limits of communications technology: whenever there’s a new way to deliver information with record-breaking speed or accuracy, there’s a very good chance that the specific piece of information that set that record was the price of S&P futures or a major FX cross. This story has played out with multiple communications technologies, and has made fortunes every time.
• Henry Oliver on why we still love Jane Austen. Austen is weirdly good at telling stories that are very much of a particular place and time, but that also have characters and scenes that are basically immortal—Emma is at least partly a warning to the verbal, bookish people who read Austen that they should use their verbal abilities for good and not just to humiliate poor Miss Bates. Oliver concludes that Austen’s real project is moral education; they’re good stories, but didactic ones meant to shape good people. So Austen, at least, lived up to her moral obligation her skills entailed.
• In Capital Gains, we look at alpha from choosing the right people to trade against. This works in both directions; if you have skill in betting on small-cap stocks, there’s a good chance this skill translates to larger and more liquid ones, at least if you can get a seat at the right fund. And staying small has its own risks; obscure asset classes are obscure for a reason.

You’re on the free list for The Diff! Light publication schedule this week and last, though don’t miss how AI’s effect on mental health roughly inverts that of social media ($). Some of this year’s biggest hits include routers, Apps, AGI, what would the afternmath of an AI bust look like?, what happened to working your way up from the mailroom?, Gonzo startups, and when prices peak before fundamentals. On the paid side: the loneliness economy has higher productivity growth ($), general-purpose technologies revalue natural resources ($), why pod shop portfolio managers are unusually vulnerable to being replaced with AI ($), why volatility fattens traders’ fingers ($), political assassinations will happen for weirder and weirder reasons ($), why lax enforcement for white-collar crime will make 2028 an unusually crazy and well-funded election ($), the case for running (some) AI businesses with (temporarily) negative gross margins ($), and what agents do to marketplace economics ($). Upgrade today for full access to the archives!

Upgrade Today

Books

Permanent Revolution: The Reformation and the Illiberal Roots of Liberalism: One of the easiest ways to fall into the trap of presentism is to extrapolate backwards from some big social shift and assume that it was part of a longer directional process. Sometimes, it’s a cycle: the 50s-to-70s transition towards later marriage, less stark gender roles, etc. certainly felt revolutionary to people at the time, but they were partly overestimating just how weird the 1950s were in that respect—those boring, conformist suburbanites were the children of the flappers and swells of a generation earlier.

Similarly, you can imagine drawing a trend line that starts with medieval attitudes towards religion—close integration between church and state, judicially-mandated witch-burnings and torture of heretics, a tightly controlled information environment—which slowly gave way to a pluralist Enlightenment view that separated church and state and let people practice their beliefs as they chose. As it turns out, at least in England, what actually happened is that those medieval stereotypes best describe the country during the early days of the Reformation, when Calvin was the best-selling writer in the English language and there was torture and heretic-burning of Catholics; the Enlightenment was, in a sense, a reactionary political movement trying to restore the relatively sane late medieval status quo.

This setup makes the book feel like it’s going to be a political history, but all of that frames the actual content, which is that it’s a literary history looking at Shakespeare, Milton, Bunyan, and Spenser as people who are trying to make sense of these changes both as a personal spiritual exercise and as a Straussian attempt to calm people down.

In Shakespeare’s case, the revised reading is that he was early to worrying about the risk of totalitarian theocracy, and wanted to warn people about it. Milton, Bunyan, and Spenser fall into a different category: in this reading, they’re all quite devout and are coping with 1) a radically stricter standard for outward and inward religious adherence, and 2) the absence of the sacrament of reconciliation as a means of getting some kind of moral closure. Instead, they produce incredibly neurotic works of poetry and prose (on Paradise Lost, the author notes just how odd it is that, "a poem written by a member of the inner core of a failed, anti-monarchical revolutionary junta should represent the core of a failed, anti-monarchical revolutionary junta as being led by Satan").

The book describes its historical model as "the Zhou Enlai school,” i.e. it takes about 150 years to figure out the implications of a revolution. Which sounds about right: you only see the permanent impact in the generation that grows up under the new status quo and doesn’t view it mostly in the context of what happened just before. That sounds like a good benchmark; Zhou Enlai may have been misquoted (he was asked about whether he thought the French Revolution was a good thing, but thought he was being asked about French protests a few years before, which is why he answered "Too early to say"), but he accidentally nailed it. There isn’t enough context to judge a big social change until so much time has passed that you’re surprised to find out what actually changed and when.

Open Thread

• Drop in any comments or links of interest to Diff readers.
• Now’s a great time for 2026 predictions. What’s going to happen?

Diff Jobs

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

• A software PE firm with an exceptional long-term track record is looking for operations executives with experience in procurement and vendor management. (Remote, US).
• YC-backed startup automating procurement and sales processes for the chemicals industry, which currently relies on a manual blend of email, spreadsheets, legacy ERPs, etc. to find, price, buy, and sell over 20M+ discrete chemicals, is hiring full-stack engineers (React, TypeScript, etc.). Folks with exposure to both startups and bigtech, but also an interest in helping real-world America with AI preferred. (SF)
• A first-of-its-kind PE firm applying blockchain technology to transform legacy businesses is seeking an Investment Partner to lead deal execution and portfolio management. Ideal candidates have 12+ years of transaction and portfolio experience at leading tech-focused buyout funds (Silver Lake, Vista, Thoma Bravo, Francisco Partners, TA Associates, HG, Thrive Capital, or similar). No crypto experience required—they’re looking for elite PE operators who are curious about frontier technology. (NYC)
• A leading AI transformation & PE investment firm (think private equity meets Palantir) that’s been focused on investing in and transforming businesses with AI long before ChatGPT (100+ successful portfolio company AI transformations since 2019) is hiring Associates, VPs, and Principals to lead AI transformations at portfolio companies starting from investment underwriting through AI deployment. If you’re a generalist with deal/client-facing experience in top-tier consulting, product management, PE, IB, etc. and a technical degree (e.g., CS/EE/Engineering/Math) or comparable experience this is for you. (Remote)
• 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)

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.