Library

When evaluating possible design solutions, people tend to overlook options that subtract something from the existing design rather than add something to it, even if the subtractive options would be better (and would make the overall design or system simpler, which has its own benefits). One obvious follow-up question to think about is how this can be addressed in designing systems and organizations and their incentive structures. Musk is famously adamant that you should delete more, but outside of a leader who drills this into everything, I’m not sure how best to institutionalize it.

Alchian criticizes rational profit-maximization as something a firm can’t really do in practical reality, and turns to evolution instead. Only firms that happen to take actions that generate profits survive, the rest “dies”. The outcome is that it looks as if existing firms were actively maximizing their profits. In this framework, it doesn’t really matter if they calculated their optimal behavior, bumbled around and got lucky, or decided to copy what was working for someone else without even knowing why it worked for them. What can I say, I’m a sucker for evolutionary framings. I also really like the role attributed to imitation, spend five minutes on LinkedIn or in a lecture about corporate innovation and it feels on the dot.

An essay about Duncan’s approach to hiring. I like it for the underlying approach moreso than any of the specific advice he gives: The skill you should be cultivating is seeing yourself and other people clearly, everything else follows from that. Selecting a candidate, then, is less about testing them, and more about figuring them out so you can see where they fit, and whether that’s the position you are trying to fill. I’ve had very little experience sitting on the employer’s side of job interviews, but as a candidate I’ve rarely felt like the person on the other side of the table was actually curious about me. And besides how the interaction feels, I agree that this is how you actually get signal from talking to people.

How do you succesfully achieve large-scale ventures? According to Flyvbjerg, you make them modular (instead of turning each part into a bespoke subproject) and you go fast. Modularity enables learning curves (each module you roll out makes you better at rolling out the next), allows you to experiment cheaply on some of the modules (which in turn feeds back as learning), and delivers value incrementally. Speed allows you to avoid variance: Every year your project goes on is further from the initial conditions under which you planned it and deemed it worthwhile, which increases risks. The empirical argument is about as thin as you’d expect of a business paper, but I like the framework and think there’s a lot to learn from it.

One of the best texts on org structure and management, written in the context of feminist activist groups in the 1970s, but applicable a lot more broadly. A lot of orgs pride themselves on not being hierarchical or overly formal, thinking what happens in absence of such structure is more just and meritocratic. Freeman argues that there is no such thing as a group without structure, and what actually arises in such cases are opaque, unaccountable, and hard to navigate informal structures of influence and power. Structurelessness, then, is the cop-out of people who don’t want to deal with creating structures, or bring a naive kind of skepticism of any and all authority. This doesn’t mean you should put in place a dictator and count on them staying benevolent, or make everything uber-rigid — but you should make structures legible, so they can be navigated on equal footing by everyone while they are working, and deliberately criticized and changed when they no longer are.

A large part of what underpins my worldview and politics is summed up in this essay series. The achievements of modernity are mind-bogglingly amazing, but we don’t realize it because their comforts and the infrastructure that supports them are to us as water is to fish. Things could, of course, be a lot better, and we shouldn’t stop trying to fix the bad parts, and trying to get to a point where more of us get to enjoy the good parts. That being said, it annoys me to no end when people act like we used to live in Eden until their favorite bogeyman came along and messed it up. The fix, as so often, is curiosity about how things actually work.

Three-part series on the question “what are mental illnesses, really”. Casual debates around this abound — “what even is normal”, etc., but then people still go out and try getting a black-and-white diagnosis of ADHD. This series of posts does a great job exploring the topic. The first part covers taxometrics, a set of methods you use to figure out whether some condition is more likely to be caused by something underlying you have or don’t have, or is just the tail end of trait variation. The second part looks at psychiatric conditions through the lens of dynamical systems. The third part asks whether conditions are more likely to be a failure of evolution, or one side of some tradeoff.

Good exploration of a reality missed by the naive programmer: Information has properties that are socially and/or legally relevant that are not directly present in the information itself. For example, provenance of source code matters — if I copy a small function verbatim in violation of a license, that’s meaningfully different from if I happen to write it myself the same way someone else did before me. Probably an interesting re-read now in light of training of and content generation using AI.

I did not “get” biology in school either — I did well enough, but it felt like memorization and I never really cared about it. I discovered it for real in my 20s, after falling down a bit of a rabbit hole via a hacker type’s youtube video on a DIY genetic engineering project, and this incredible essay really captured my feelings about the subject once things started clicking into place and I grokked its significance.