Apollo Stored Its Flight Program in Wires Threaded Through Magnetic Rings
MIT needed compact storage that could survive a mission, but its programmers had to hand finished code to the factory months before launch.

A software release could be a bundle of wires you could hold. Updating it involved a rather more demanding operation than pressing Deploy: somebody had to change the wiring inside a sealed module.
Restorer Ken Shirriff describes the Apollo Guidance Computer’s fixed memory as storing “192 bits per core.” Ordinary erasable core memory used a ring for each bit. Rope memory squeezed many fixed bits around the same magnetic element by giving each sense wire a physical route.
The ring still changed its magnetic state during a read. That change induced a voltage in wires passing through it, and amplifiers turned those small signals into computer-readable bits. A bypassing wire missed the pulse. The program lived in which wires went where, so flipping a ring did not rewrite the program.
Shirriff’s restoration account distinguishes 2,048 words of erasable memory from 36,864 words of rope ROM. Similar-looking rings were doing very different jobs. One held changing work; the other determined what instructions the computer could fetch.
MIT had already selected core rope for a Mars probe computer design, according to NASA’s hardware history, and carried the approach into Apollo. More information fit in less space, a persuasive feature when your computer has to leave the planet. NASA lists the Block II computer at 70.1 pounds, with an identical machine in the lunar module.
Compactness sent a bill to the software team. NASA’s history says programs had to reach manufacturing “months before a scheduled mission.” Each sealed module needed rewiring to change its bits. Raytheon tested the manufactured ropes against the program’s delivery tape, which could catch a wiring mistake; correctly weaving the wrong instruction would still give you the wrong instruction.
The path into those wires was surprisingly elaborate. NASA’s Computers Take Flight describes source code on 80-column punched cards, an assembler producing binary on magnetic tape, and simulators using that tape. Tested software produced separate weaving and checking tapes on Mylar. One drove manufacture; the other checked what manufacture produced. The installed computer then ran its own memory check.
That is a release pipeline with an actual factory in the middle. Forget the reassuring green button for a moment: the final check had to establish that the object in the spacecraft contained the same bits developers had tested.
The restoration exposes a useful escape hatch. Ground-test simulator boxes let an external computer supply program words through the rope connections, so developers could try code without manufacturing another rope. Even this famously physical software had a way to rehearse a change.
The simulator offers a practical model for today’s hardware makers. If the finished artifact is expensive to change, make the development version cheap to change and preserve the interface between them. Apollo’s software could escape the weaving machine during testing. Which parts of our supposedly easier release processes still cannot?