Longreads + Open Thread

Longreads

• Kelsey Piper has a piece on Sonnet 4.7's eerie ability to identifyidenitfy the writer of a piece of text, even if the text wasn't posted online and is in a different genre from what they usually write. I've tested this out and can confirm that given a large enough online corpus, it can correctly identify me based on: recent emails, old emails, decades-old school assignments, and fiction. And can confirm that when it explains how it did this, it will fabricate details (Sonnet has some very flattering misconceptions about my educational and professional background.) This doesn't mean the end of pseudonymity, though; it means that everyone is entitled to pseudonymity, unless they get famous, in which case they're tied to that pseudonym forever. Of course, if inference keeps getting cheaper, it may reach the point where nosy people are just running random emails from their friends and colleagues through Sonnet to see if they have an online presence. In the meantime, some of this risk may be avoidable by running a text through AI before pseudonymously publishing it, which will protect whistleblowers but mean that pen names are either temporary or are the names under which people post pure AI slop, albeit with unusually long prompts.
• Joel Miller argues that adjusted for inflation, books are cheaper than they were in the 1960s. But they feel more expensive! One possibility is that, for people who grew up reading and sometimes buying books, the price of a new book was one of the first prices they anchored to—in the mid-90s, I was keenly aware that the next Star Wars book I wanted would be $5.99 plus tax. If you don't grow up reading a lot, you'll anchor to higher prices later in life. And if upwardly-mobile people shift more of their media consumption to books, they'll also experience this price increase at a time when their budget is growing faster (it's hard—I've tried—to go broke buying books).
• In The New Yorker, Julian Lucas profiles data recovery firm DriveSavers, and meditates on the question of just what it means to lose your data. Many of us have old hard drives, ancient smartphones, pay annuities to AWS in the form of old S3 buckets, etc., and in theory we all want that data. But in practice, some of that information is precious, and some of it is of sentimental value that will not survive first contact with actually retrieving it. Still, it's nice that if your electronic device contains your almost-complete dissertation, your novel, your journal, the last voicemail you got from a departed loved one, etc., there are data recovery nerds out there who absolutely love bringing it back.
• The Terminalist is thoughtful as always on competitive moats in the data business (though Pangram indicates that they had a little help coming up with a striking intro section). One of the interesting points, relevant to many parts of the software world beyond financial data, is that once there's an incumbent whose product is always full-screen on your customer's primary monitor, they're incredibly hard to displace. So challengers tend to go API-first—which means their first customers tend to be sophisticated, and to provide good feedback that makes the product appealing to the next cohort of prospective customers. Given how cheap it is to vibecode a user interface, serious products will be born headless and will either get a default interface or be distributed through someone else.
• There's a lot to say about this NYT interview in which various unpleasant people give their guarded-but-positive opinions on shoplifting, assassinations, etc.. Much has already been said about why this worldview is bad, but it's important to emphasize that it's not just stupid (it happens) but actively pro-stupidity. The problem with a moral system where, if you don't understand why someone would behave in a certain way, you get the right to punish them for it, is that this system accords the least power to people who are smart and well-informed, and maximizes rights for the dumb and disagreeable. So it splits the difference between saying the world should be run by the stupidest people and that it should be run by the most unpleasant ones. Either way, this imposes a pretty significant burden on the vast majority of well-behaved people, who do understand that there's a finite amount of medical care, and that at some point maximizing overall patient care requires telling someone that they aren't getting the care they'd like. It's an incredibly complicated topic, but it's also true that every healthcare system rations one way or another, and that if being an executive responsible for setting those rules mandates the death penalty, there are plenty of people running Medicare and the NHS who are similarly culpable. Interestingly, Hasan Piker states in the interview that he, personally, doesn't steal, because his parents memorably impressed on him that it's a bad thing to do. It would be nice if he were similarly paternalistic with his followers.
• A read.haus user asks about the overnight anomaly, where equities and other asset classes get most of their returns overnight, and specifically asks about the conspiratorial tone of one of the prominent researchers in this area. It's great conspiracy fodder! But the general rule with conspiracies is that they're hard to maintain in general, and that they're ridiculously hard to maintain if someone can defect by opening up an additional brokerage account and defecting against their co-conspirators. Market conspiracies do exist, but they have to operate so quickly that there isn't time to defect. So they're a poor explanation for the overnight anomaly, even if they do fit the data.
• In Capital Gains this week: hostile takeovers happen globally, but shareholder activism is a disproportionately American phenomenon. It's a historical contingency: we evolved pro-shareholder norms because the state was less involved in the economy than in other places, so corporate governance mattered more. And the US has sufficiently liquid capital markets that once an activist is involved in a company, it can lead to a preference cascade where their old shareholder base is quickly replaced with people who support the activist.

Books

Reentry: SpaceX, Elon Musk, and the Reusable Rockets that Launched a Second Space Age: A rough history of space travel is that in the 50s and 60s, it advanced at a pace few fields could match; after that, it slowed and stagnated, and suddenly, in the 2000s, progress picked up again. Reentry tells the story, through the form of vignettes about each launch. There's a particular kind of risk tolerance that entails committing to a deadline before knowing precisely what obstacles there are to meeting it. It's a nice forcing function, though probably safer to do with things like building a prototype of a mobile app than with launching a very expensive piece of hardware, full of dangerous fuel, into space.

There are many anecdotes about how hard people work when they're working under Musk. Musk himself has more variable output, at least from the perspective of someone at one of his companies, but this actually creates an interesting motivational uncertainty: if you haven't seen him at the office lately, does that mean he's stopped putting in hundred-hour weeks, or just that he's putting them in somewhere else? SpaceX employees tend to assume that the expectation is that they'll keep putting in the hours, working themselves—in one memorable story about Marty Anderson, literally—to the bone.

The other striking thing about the talent story is: one popular narrative is that SpaceX was aggressive at a time when NASA was getting lazy, but many of the people who worked under or with Musk had family connections to NASA, or had a formative experience that involved learning about NASA-run missions early in life. (Holly Ridings, NASA's flight director for SpaceX's Dragon launches, saw the Challenger explosion live in class—and decided that she ought to go work for NASA to make sure that didn't happen again. Some people are built differently.)

If you read through the book, you can see many places where, had history gone slightly differently, it would be a slimmer volume about some dot-com goofball who convinced himself that he could do space travel better than the organization that literally landed on the moon, and eventually got some people killed and lost all of his money. There were just a lot of close calls! But at some point, you have to see those outcomes as correlated: beating impossible-seeming odds is a skill, and it isn't evenly distributed.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• There are plenty of stories about big, high-profile engineering achievements, but what are some books about cases where the challenges were equally daunting but the results aren't as high-profile?

Diff Jobs

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

• Lightspeed-backed team building the engineering services firm of the future is looking for founding members of technical staff excited about working alongside civil engineers to translate their domain expertise into the operating system that powers the next era of great American infrastructure. If you’re an engineer with strong product intuition, who's energized by access to users, and excited by the prospect of transforming how we design and construct our built world with frontier AI, this is for you. (NYC, SF or Remote)
• AI Transformation firm with an ambition to build an economic world model to run swathes of the private, unstructured economy is looking for FDEs, Platform Engineers, and business generalists who understand how to solve problems.  
• Well-funded, frontier AI neolab working on video pretraining and computer action models as the path to general intelligence is looking for researchers who are excited about creating machines that learn from experience, not text. Ideally you have zero-to-one pre-training experience and/or are a high-slope generalist who’s frustrated that the big labs aren't doing this. (SF)
• High-growth startup building dev tools to help highly technical organizations autonomously test/debug complex codebases is looking for a senior design engineer to own their design system and build the visual abstractions customers rely on to simulate their software systems, find bugs, and quickly remediate them. A compelling portfolio, a rare blend of design and engineering chops, and a deep understanding of how the internet and browsers work required. (D.C.)
• Series A startup building multi-agent simulations to predict the behavior of hard to sample human populations is looking for researchers and engineers (ML, platform, infrastructure, etc.) to improve simulation fidelity and scale the platform to hundreds of millions of simulation requests. Problem-solving and genuine interest in simulation matter more than pedigree. Experience working with languages with an algebraic type system is a plus. (NYC)
• A Fortune 500 cybersecurity company with decades of proprietary security data is running an internal incubation with a pre-seed startup mentality and a mandate to build something new in AI. They are looking for a founding engineer who can ship fast, an engineer with a security background who’d be excited to contribute to OpenClaw’s security efforts, an AI researcher, and a generalist (ex-banking/consulting/PE background preferred) who wants to wear a bunch of different hats. Comp is FAANG+ and cash heavy. If you want to build something new in AI, but also need runway, this is for you. (SF/Peninsula) 

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.

And: we're now actively deploying capital into early-stage companies through Anomaly. Our focus is on defense, logistics, robotics, and energy. If you'd like to chat, please reach out.