David Mindell gives more details of the AGC rope core memory in his book Digital Apollo. The software was stored in the wiring pattern, not in the cores themselves. The MIT programmers had to deliver final code three to four months before each mission. That turned out to be an advantage because "risky last minute changes of the program just before the flight were physically prevented," according to program director David Hoag.
IIRC, and this is secondhand from folks at MIT Instrumentation Lab, it was a simplified version of core memory. Ones were represented by cores and zeroes by the absence of cores, so the memory woudnt need the write cycle after the read to restore the state of the memory and there was no need to boot the memory.
