In 1944, Howard Aiken and IBM engineers completed Harvard's Mark I, a 51-foot-long calculator that handled wartime calculations for the US Navy

Picture the noisiest "smart" gadget you own. Maybe it's a laptop fan on its last legs. Now imagine a machine the length of a city bus, making the kind of racket a freight train could appreciate, all to do simple arithmetic. This machine was real; it was housed at Harvard, and according to the National Inventors Hall of Fame, it was the brainchild of engineer Howard Aiken and a crew of IBM designers, Clair Lake, Frank Hamilton, and Benjamin Durfee among them, who built the device in 1944 to solve equations the U.S. Navy needed solved yesterday.

They called it the Automatic Sequence Controlled Calculator. Everyone else knew it as the Harvard Mark I. It's easy to dismiss the Mark I as a quaint artifact. But the device that outweighed a pickup truck is, in fact, the awkward, oversized ancestor to the phone in your pocket, and its story has more than a few uncomfortable echoes for anyone keeping up with the modern gold rush around AI.

The machine that outweighed a pickup truck


The Mark I isn't a compact piece of hardware. According to the Harvard Gazette, the machine stretched 51 feet long and 8 feet high and weighed 10,000 pounds. At the heart of its glass-paneled casing was a 50-foot-long rotating shaft driven by an electric motor. The machine's core contained thousands of switches, dials, and relays that clicked to life to shuttle information around the machine. The Mark I could handle a handful of additions per second; multiplications required six seconds per calculation, while divisions took up to 15. Your microwave's digital clock has a faster processor. But the Mark I was, in 1944, the fastest calculating machine in the U.S.

A Navy contract

There's a comforting mythology around technology in America: lone geniuses, scrappy start-ups, the occasional dorm-room invention. The Mark I isn't a product of that school of thought. It was imagined by a Harvard professor, built by IBM at a research facility in Endicott, New York, and turned to real-world use, as the Computer History Museum notes, in calculating things like ballistics tables and equations related to the Manhattan Project.

In other words, America's first computer was a defense contractor with a user interface. That pattern continues right up to the present day. Today's AI boom makes its living with cloud contracts with the Pentagon, subsidized silicon chip development from the federal budget, and research labs that trace their bloodlines to the World War II era. "Disruptive startups" tend to have a habit of leaving out the government contracts that pay their bills.

The woman history almost erased

The Mark I's most consequential hire wasn't someone with an engineering degree and a nickname like "Turing." It was Grace Hopper, a Vassar math professor who signed up for the Navy in 1943 and was assigned to work with Aiken, becoming one of the Mark I's first programmers. Hopper stayed on at Harvard after the war and moved with the project to its successor, the Mark II. It was there, on September 9, 1947, three years after Mark I went into service, not during her original wartime assignment, that her team found a moth jammed in one of the machine's relays and taped it into the logbook with the note "first actual case of bug being found."

The term "bug" for a technical fault actually predates that incident by decades (Edison used it in an 1878 letter), and the Smithsonian notes the logbook entry probably wasn't even in Hopper's own handwriting; so the fairest claim is that she helped popularize the moth story and the word "debugging" through her later lectures, not that she coined either term. Separately, Hopper went on to build the first compiler and is widely credited as the driving force behind COBOL, developed years later at Eckert-Mauchly, a contribution long under-recognized relative to Aiken's, rather than one that was ever mistakenly credited to him.

Why a 1944 calculator should matter to you

The Mark I was taken apart, piece by piece, fifteen years after it was built and shipped off to places like IBM, Harvard, and the Smithsonian. It never got smaller, cheaper, or more adorable; it just got replaced. The real lesson from the Mark I isn't one of retrocomputing nostalgia; it's that every single "revolutionary" device we build today is the latest in a long line of government-funded, underpaid, uncelebrated machines that happen to solve a problem we didn't know we had.

Next time you see an article about how some new AI algorithm is going to fundamentally change the way we approach a class of problems overnight, remember there used to be a 10,000-pound machine in a basement that actually did it, and how easily the people and details behind it get flattened or forgotten, even when the machine's own name survives.