The Divorce
of Departments

If you want to understand computer science you need to understand it grew up in two departments, pure math and electrical engineering, and that they converged over time due to external pressure. You can see similar pressure in how mechanical engineering evolved into automotive as an industry-driven rather than purely theoretical differentiation. So this isn’t uncommon. But in computer science’s case you end up with a significant population who come from the pure math side, arguably because it’s easier, whose talent therefore lies in configuring a machine they do not themselves understand.1

The convergence was never a courtship; it was logistics. The war needed firing tables, the tables needed machines, and the machines needed people who could hold both a soldering iron and a proof. For roughly one generation those were the same people. Von Neumann could walk from the blackboard to the machine room and be the most dangerous person in each. The institutions noticed the grant money before they noticed the discipline, which is the usual order.

Then the boundary formed, the way boundaries do: not where the subject divides naturally but where the teaching divides conveniently. The instruction set became a treaty line. Above it, mathematics could pretend the machine was ideal; below it, engineering could pretend the program was someone else’s problem. Both pretenses were productive. Treaties usually are, until the territory moves.

The tubes became transistors and the transistors became integrated circuits, and the programmer class rode the treaty line through all of it, largely unaffected and largely unaware. Three substrate revolutions in one professional lifetime, and the view from above the ISA barely flickered.

Because of this, the substrate can shift without the wider programmer class necessarily noticing or being affected. That is the point of the barb, not the insult of it: a population whose talent lies in configuring a machine it does not understand is precisely a population whose machine can be exchanged beneath it.

Which makes the interesting question the one nobody staffs a committee for: whether, like mechanical engineering making the early computers and then losing them, electrical engineering ends up divorced once quantum comes around. Mechanical built the first machines and got the divorce anyway. It would be strange if the pattern stopped at exactly the department that trusts patterns least.

The honest answer is that the divorce would not feel like one from above the treaty line. Classical silicon keeps running the world; the quantum machines split off sideways and take the machine-room talent with them; the programmer class notices nothing, because their machine never changed. A field can lose a limb without losing its balance. Whether it can lose a limb without losing something else is the question worth keeping open.

1. 1. The claim is about population talent distributions, not individuals. The counterexamples are real and are not the point. ↩
2. 2. Show HN corpus, n = 1,590, deterministic CSS/DOM checks, 5–10% false-positive rate by manual QA. Krebs, 2026. Every number is a grep. ↩