Apollo 11 launched on 16 July 1969 with a 70-pound guidance computer and a flight program that was already physically committed to copper wire.

The computer used integrated circuits for its logic, but its flight program was not stored on a modern-style chip. It lived in core rope memory at a Raytheon factory in Waltham, Massachusetts, where wires were threaded through or around tiny magnetic cores to encode the flight software. Once a rope module was built, changing the code was no small edit. A serious mistake could mean rewiring or rebuilding part of the memory itself.

The computer they actually flew

The Apollo Guidance Computer was designed at the MIT Instrumentation Laboratory under Eldon Hall and built by Raytheon. By Apollo 11, the Block II version took up less than a cubic foot of cabin space, ran at a clock speed of about 2 MHz, and carried 4 kilobytes of erasable core memory for working data and 72 kilobytes of read-only core rope memory for the flight program.

The figures look small now. They were unprecedented at the time. Many computers of the era were still large machines built for laboratories, companies, or government facilities. Fitting a real-time, fault-tolerant guidance computer into a spacecraft required miniaturisation that drove the early integrated-circuit industry, and NASA was an early and large buyer of the silicon chips that would later go into everything else.

What we think modern comparisons tend to miss is that the AGC was not simply a slow computer. It was a real-time, multitasking machine with priority scheduling and error recovery built in. During the descent of the Lunar Module, it raised a series of 1201 and 1202 program alarms while approaching the surface. It did not crash. It shed lower-priority tasks, kept the critical guidance loop running, and let the astronauts land. Some modern USB-C chargers have more raw processing power than the AGC. None of them would have survived that particular afternoon.

How software became hardware

Core rope memory was unusual even in the 1960s. The principle was simple. A grid of tiny doughnut-shaped ferrite cores was arranged on a frame. To represent a one, a sense wire was threaded through the centre of a core. To represent a zero, the wire passed around it. When current flowed through the wire, cores with wires through them registered a magnetic change, and cores with wires around them did not.

What this meant in practice was that the software was literally hardware. The program was not data sitting on top of a memory chip. The program was the wiring pattern of the memory.

Engineer Ken Shirriff, who has documented the AGC in technical detail at righto.com, estimates that the wiring of a single rope module took roughly eight weeks of work and cost about fifteen thousand dollars in 1960s terms. The Block II AGC carried multiple rope modules. Code freezes had to happen well in advance of launch, with each version verified, manufactured, and inspected before flight.

Who actually did the work

The weaving was done by women employees at the Raytheon Waltham plant. Many of them were recruited from the local textile industry and from the Waltham Watch Company, the same company that helped produce the high-precision gyroscopes used in the Apollo guidance system. The skills mattered. Threading the wires correctly across thousands of cores in a fixed sequence required steadiness, attention, and an error rate close to zero.

Apollo engineers referred to the process as the “LOL method”, for “Little Old Ladies”, a nickname Margaret Hamilton’s team at MIT used affectionately for the weavers. Hamilton, who led the software team that wrote the code being woven, was known internally as the “Rope Mother”. The Smithsonian National Air and Space Museum has collected her papers, including listings from the program photographed beside her in the now-famous image where the printed stack is taller than she is.

Whether the “LOL” nickname was widely used at the time is contested. Shirriff has questioned whether the term was current during the program or whether it became attached to the work later. What is not contested is that the women who wove the ropes were skilled technical employees, with their work inspected repeatedly by both Raytheon and federal government inspectors. Mary Lou Rogers, one of the Apollo weavers, has recalled that each component had to be checked by three or four people before being signed off, with a federal inspection team coming through on a regular basis.

The story keeps getting told the wrong way

For a long time the standard version of this story has carried a small condescension inside it. The workers are described as little old ladies. The labour is described as craft. Some 1960s Raytheon press materials and engineer accounts described the work as “tender loving care”, language that flattered the women in a way that also quietly removed them from the engineering side of the project.

That framing has been challenged. The Making Core Memory project, led by Samantha Shorey, Daniela Rosner, Brock Craft and Helen Remick, has spent several years trying to undo the “no thinking, no skill” reading of the weaving work. Their workshops, in which participants attempt to weave functioning core memory patches by hand, demonstrate the level of focus, technical literacy, and precision the work required.

One version of the story that has drifted into popular history is that the weavers “had no idea” what they were working on. That is wrong. The work was on a federally inspected, high-profile NASA contract. The women knew which program they were assembling. Some, including Rogers, have spoken about it directly.

Reading it now

The useful thing about Apollo 11’s rope memory, in our reading, is as a counter-image to how most people now think about software. The AGC’s flight program could not be patched after launch. It could not be patched in the final months before launch in any straightforward way. By the time the rocket left the pad, the code had been physically committed to copper wire by people whose hands had worked on it for weeks. The cost of getting it wrong was measurable in lives and in months of lost work.

Modern software has not had that physical constraint for a long time. Code now ships in builds that can be updated overnight. Bugs are patched mid-flight. The result is software that does more, faster, and that also drifts from a discipline that once treated error as something to be designed out rather than fixed later.

Apollo 11’s guidance code was written by Hamilton’s team at MIT, woven by women at Raytheon, and inspected by federal officials before it ever flew. By the time Neil Armstrong stepped onto the surface, every line of it was already, literally, copper wire.