On the afternoon of September 9, 1947, a team of engineers at Harvard’s Computation Laboratory was hunting a fault in the Mark II Aiken Relay Calculator when they unscrewed Panel F, peered into Relay 70, and found a moth. Not a metaphorical glitch. An actual moth, roughly two inches across, wedged between the contacts of an electromechanical relay and stopping the current cold. Someone — the logbook does not say who, though the entry is widely associated with Grace Hopper’s team — pulled the insect out with tweezers, taped it onto a page of the operations log, and wrote beside it: “First actual case of bug being found.”
That logbook page now sits in the collection of the Smithsonian’s National Museum of American History, moth and tape still intact.
The machine that ate the moth
The Mark II was a beast. Built for the U.S. Navy at Harvard under Howard Aiken, it weighed roughly 25 tons, ran on around 13,000 high-speed electromechanical relays, and used punched paper tape for instructions. Each relay was a small mechanical switch. Two metal contacts that closed when an electromagnet pulled them together. Multiply that by thousands and you get a machine whose entire arithmetic depended on flat slivers of metal touching each other cleanly, thousands of times a second.
Anything between those contacts, whether dust, a fleck of solder, or a moth, would break the circuit. The Mark II had been under construction and testing at Harvard since 1945 and was not delivered to the U.S. Navy Proving Ground at Dahlgren, Virginia until the spring of 1948, six months after the moth was found. The lab building itself was spectacularly hospitable to insects drawn by the warmth of the relays and the lights of the room. Hopper later recalled that the windows had no screens, and that all the bugs in the world wandered in at night.
Grace Hopper, then a Navy Reserve lieutenant assigned to the Harvard team, was one of the original Mark I programmers and had moved with the project to its successor. She did not personally find the moth. She said so herself, repeatedly, in later interviews. But she told the story so often, and with such delight, that her name became inseparable from it.
The word “bug” was already old
This is the part that often gets lost. Engineers in 1947 did not invent the word “bug” to describe a technical fault. They were making a joke that already had teeth.
Thomas Edison was using “bug” to mean a flaw in a circuit by the late 1870s. In an 1878 letter to Theodore Puskas, Edison complained of “Bugs — as such little faults and difficulties are called” and noted that months of anxious watching, study, and labor were required to chase them out of his inventions. Telegraph operators used it. Aviation mechanics used it. By the time the Mark II logbook entry was made, the term had been kicking around electrical engineering for nearly seventy years.
Which is exactly what makes the September 9 note so funny, and why it survived. The annotation isn’t simply stating they found a bug. It is “first actual case of bug being found.” The humour depends on the word already being a dead metaphor. A team of engineers, exhausted from chasing a glitch, opens a panel and finds the cliché made flesh. They taped it in because it was a punchline.
Why the metaphor stuck
Words for abstract things tend to come from concrete things. Bodies, tools, animals, weather. Cognitive linguists describe this as mapping a familiar source domain onto a less structured target domain, the same mechanism that gives English phrases like “a warm welcome” or “a sharp argument.” Bodily and physical experiences quietly structure even our most technical vocabularies.
Software was a target domain desperate for source material. In 1947, almost nobody outside a handful of military and university labs had seen a computer, let alone reasoned about one. The machines were opaque, expensive, and behaved in ways their own builders could not always explain. Calling a fault a “bug” gave it a shape. Something small. Something hidden. Something that could, in principle, be found and removed.
The metaphor did real cognitive work. It told you what kind of problem you were looking at, local, discrete, fixable, rather than systemic or philosophical. Cognitive linguistics calls this kind of framing a conceptual metaphor, and the metaphors people reach for shape how they reason about the underlying problem, not just how they describe it.
Grace Hopper’s other inventions
Hopper was 40 years old in 1947, a former Vassar mathematics professor who had joined the Navy Reserve during the war and ended up at Harvard programming the Mark I by hand, using paper tape and patience. She would go on to design the A-0 compiler, early software that translated human-readable instructions into machine code, and then push for COBOL, the programming language that ran banking and payroll systems for half a century and still runs a startling amount of legacy infrastructure on which modern finance depends.
She retired from the Navy in 1986 at the rank of rear admiral, the oldest active-duty officer in the United States armed forces at the time. A guided-missile destroyer, the USS Hopper, was commissioned in her name in 1997.
But the moth story is the one that travels. Partly because Hopper told it so well. She carried a photograph of the logbook page to lectures for decades. And partly because it is, structurally, a perfect anecdote: a famous person, a famous machine, a tiny dead insect, and a one-line joke that turned out to be a creation myth.
What the logbook actually looks like
The page is graph paper, ruled in faint blue, with timestamps running down the left margin. The relevant entry is dated 9 September, marked at 15:45. A quarter to four in the afternoon. The moth is taped slightly off-centre, wings spread, its body still intact. The handwritten annotations around it record the time the relay failed, the test that exposed it, and the now-famous line about “first actual case.” Other entries on the page are routine: tests passed, tape loaded, machine restarted.
The Smithsonian holds the page in its computing collection today. The logbook itself had spent decades at the Naval Surface Warfare Center at Dahlgren, where the Mark II eventually went to work, before being transferred to the museum in the early 1990s. The moth has become one of the most famous insects in computing history.
A vocabulary that did not stay still
From “bug,” everything else followed. Debugging entered engineering vocabulary by the early 1950s. “Patch” came from the physical paper patches taped over holes in punched cards and tapes to correct errors. Another concrete thing dragged into the abstract. “Crash,” “hang,” “freeze,” “kernel,” “daemon,” “thread,” “fork”: software English is almost entirely built from borrowed body parts and farm equipment, because the alternative was inventing a private language nobody would learn.
Some of those borrowings were so successful they became invisible. Ask a developer in Bangalore or São Paulo what a bug is and they will answer in technical terms, not entomological ones. The dead-metaphor stage, when a word loses its original sensory meaning and becomes purely functional, is, paradoxically, when a metaphor is working hardest. The metaphors people stop noticing are usually the ones doing the most structural work in a field.
The accidents that name things
Computing history has a habit of being named by accidents. Silicon Canals has written about the first message ever sent over ARPANET, a 1969 attempt to transmit “LOGIN” that crashed after two letters, leaving “LO” as the inaugural utterance of the internet. The pattern repeats: a small mechanical failure produces a story so neat that it outlives the technology that generated it.
Almost none of the people in the Harvard lab on September 9, 1947 are still alive. The Mark II’s relays are dust. The paper tape is in archives. But the word the team used to laugh at themselves, a word Edison had been using since before Hopper was born, became, through that single logbook page, the irreducible unit of every conversation about software ever since.
So here is the uncomfortable question. If “bug” was already a dead metaphor in 1947, and worked harder precisely because nobody noticed it any more, what are the dead metaphors we are running on right now? “Cloud.” “Training.” “Hallucination.” “Neural.” Each one smuggles a frame into the room before the argument begins. Each one tells you the shape of the problem before you have looked at it.
The moth is still pinned to a sheet of graph paper at the Smithsonian. The real question is not what it meant in 1947. It is which of today’s casual words will turn out, eighty years from now, to have done the quiet work of deciding what we could and could not think about the machines we built.
