Engineering Research Associates (ERA)

Contribution to AI

Codebreaking paid for the machines before science could. The wartime Navy group that reassembled itself in a St Paul glider factory in 1946 brought with it a habit of thought — statistical comparison of enormous quantities of intercepted text — that demanded both speed and, more awkwardly, somewhere to keep the data. Its answer was the magnetic drum, engineered to capacities and reliabilities nobody else was reaching, and cheap enough that a machine could hold tens of thousands of words when electrostatic storage held a thousand. Drums became the working memory of computing's first commercial decade, in the classified machines, in Univac's scientific line, and in the drum-based business computers on which a generation first learned to programme. The survey the firm wrote for the Office of Naval Research and published in 1950 as High-Speed Computing Devices was, for several years, where engineers outside the secret programmes learned what the components of a digital computer actually were. Atlas, running that same year, demonstrated that one general-purpose stored-programme machine could replace a room of single-purpose comparators. Its successor was Seymour Cray's first design credit, and the model that followed shipped an interrupt facility, letting a computer be pulled off a batch job to take instrument readings in real time. The engineers who left in 1957 to start Control Data carried the whole method with them.

Drafted with AI and edited by hand (claude-opus-5, reviewed 2026-08).