What If the Transistor Was Invented in 1921?

It could, though it took her most of the decade to prove it at the scale the calculating-machine company wanted. By 1939, the company's flagship model used eleven hundred of her valves instead of mechanical relays, and could complete a payroll calculation in minutes that had taken a trained clerk with a mechanical machine most of a day.

When war came, that machine's descendants were requisitioned almost immediately, not for payroll but for the much harder arithmetic of artillery tables and, within two more years, the first serious attempts at codebreaking machinery. Eleanor herself, now thirty-nine and listed on seven patents, was recruited to a research unit outside Washington that no one would later admit existed under that name, working alongside engineers who had spent their careers on vacuum tube computing and were visibly uncomfortable taking instruction from a woman with no university degree at all.

She did not care, by then, whether they were comfortable. The valves she had built in her father's hardware store were, by 1943, running calculations that shortened the processing time for intercepted naval traffic from days to hours — not because anyone had planned a revolution in computing, but because a government desperate for speed had finally stopped asking who held the patent and started asking only whether the device worked.