By Dr David Bell, specialist anaesthetist (retired), software engineer and author.
Chapter 5 of Lady Lovelace's Objection went up on Substack this morning. It is the John von Neumann chapter, and it is the one where the book stops being about people who imagined computers and starts being about people who built them. Every machine you own is laid out to the plan in a hundred and one pages he wrote in pencil on trains in the spring of 1945. That much most readers already know. What follows are the three things I did not know, or had not understood, until I sat down to write him.
1. The most consequential document in computing had one name on it, and nobody took minutes
In the autumn of 1944 von Neumann started turning up at the Moore School in Philadelphia on Friday afternoons, whenever Los Alamos could spare him. The engineers building ENIAC, Presper Eckert and John Mauchly chief among them, were already planning the next machine, and a small group of them sat in a room with him through the winter to argue about what it should be. Von Neumann asked that no minutes be kept. He said they would inhibit the discussion.
By June 1945 the discussion had become a typescript called First Draft of a Report on the EDVAC. The title page read by John von Neumann. It bore no other names. Herman Goldstine, the army liaison, mimeographed twenty-four copies and posted them to Harvard, MIT, the Admiralty, RCA and the rest without asking Eckert or Mauchly. That act put the design into the public domain, invalidated the patent the two engineers had been preparing, and twenty-eight years later a federal judge cited the document in the ruling that meant the American computer industry never paid a royalty on the stored-program computer.
Eckert, interviewed forty years on, said the questions they had answered for von Neumann came back a month later in his handwriting as if they had been his questions.
Anyone who has watched an author list being negotiated on a clinical paper will recognise every part of this. The senior name that arrives late and ends up first. The junior people who did the bench work and are told the paper is a synthesis. The absence of a record of who said what, which everybody agreed to at the time because it felt collegial and which only one party regrets afterwards. Medicine has spent thirty years building authorship rules to stop exactly this, and it still happens. Computing in 1945 had no rules at all, and the result is that the machine on your desk is called a von Neumann machine and almost nobody outside the field can name Eckert.
I do not think von Neumann stole anything. I think he did what brilliant senior people do when nobody makes them write down where the ideas came from. The chapter tries to be fair to all four men. Fairness, it turns out, requires about three thousand words.
2. He died under guard, and the guards could not understand him
In August 1955 an X-ray taken for an unrelated reason showed a lesion on his left collarbone. The cancer was already in the bone. Where it had started was unclear at the time and is unclear now. Within a year he was in a wheelchair, and by the autumn of 1956 he was a patient at Walter Reed in Washington.
He was not an ordinary patient. He was a commissioner of the Atomic Energy Commission, he had run the implosion calculations for the first bomb, he had chaired the committee that launched the American intercontinental missile programme, and he carried the operational detail of the entire nuclear arsenal in a memory that his friends described as not quite human. Eisenhower personally authorised a watch. Air Force officers sat in the next room in shifts, around the clock, with orders to write down anything he said.
What he said, in the last weeks, was in Hungarian. His wife Klári recorded that the officers wrote it down anyway, and nobody could translate it.
I have sat with a great many people as they went under and as they came back, and I have heard what a mind says when the front of it is switched off. It is rarely secrets. It is childhood, and mothers, and the language you learned first. The men in the corridor were guarding against a leak. What was leaking was a boy from Vaci utca. There is something in that scene about how little the machinery of a state understands the thing it is trying to protect, and I could not get it out of the chapter, so it stayed.
3. His last book argues that the brain is not his kind of computer
The manuscript he was working on through that final summer was a set of lectures for Yale that he never delivered. Klári had it published the year after he died as The Computer and the Brain. It is eighty-two pages long and it breaks off mid-thought.
The argument is the one I least expected from him. The machine he designed works by long chains of exact logic, where a single wrong bit ruins the answer. The brain, he says, does not work that way. It runs very short chains of statistical logic, massively in parallel, on components that fire unreliably. It tolerates noise and damage. It never gets quite the same answer twice, and its answers are good enough. The mathematics that describes such a thing, he wrote, had not yet been invented.
Read in 2026 that is a startling page. The systems that have lately made the question of thinking machines urgent again are not von Neumann machines in any interesting sense. They are enormous statistical networks that tolerate noise, run in parallel, and produce answers that are good enough more often than anyone expected. The man who gave us the exact machine spent his last months arguing that the interesting machine would be an inexact one. He was right, and he did not live to say how.
That is where the chapter leaves him, and it is the hinge on which the rest of the book turns. Ada Lovelace said the engine could only do what we know how to order it to perform. Von Neumann built the engine, showed that it could in principle rewrite its own orders, and then, dying, sketched a different engine that nobody had ordered at all.
Read Chapter 5 in full on Substack: The Man from Budapest.
The book's home, with the contents and the sources for every chapter, is ladylovelacesobjection.com. The serial runs every Tuesday from here. Next week: two days before Christmas 1947, a man at Bell Labs slits a strip of gold foil with a razor blade and hears his own voice come back louder.