An Ars Technica article that takes a detailed look at the workings of the AGC, and how Eyles and the team at MIT programmed around the faulty abort status bit.
I know several people here have been following the
AGC restoration effort documented by Curious Marc, this is a nice companion piece to it.
As limited as the AGC hardware was (36k words of read-only rope core to stuff all the code in, and a paltry 2k words of RAM), the software was was incredibly sophisticated, with some jobs being very akin to virtual machines, allowing the programmers to execute instructions beyond the AGC's core instruction set.
Good stuff.
