Sun May 19 19:33:02 EDT 2019
Starting out with the original "zipfold", the current idea is similar,
but just represented differently, and also allowing for some more
freedom (triangle feedback).
Maybe it's time to start working on a representation that can then use
standard map and fold to end up with a loop? Can be done in LTA
already. It will give more of an idea of how things are formed.
Then continue with single-dim eliminate and accumulator detection.