Incremental theory compression

Implementation and explanation have different goals. An ideal implementation is non-reducible and efficient. It cannot be reduced to a simpler implementation, and it does its work without too much overhead. The implementation is the reduced, final state.

Theory is different. A great theory explains well. When I learn computerly things, I have stumbled upon a pattern of repeated copy/compress/expand cycles. I write out some code that does the thing. I copy it, compress the non-interesting parts, and expand the implementation to do new things. Done well, each cycle improves my theory of the thing I'm explaining.

Want more? Read Semantic Compression by Casey Muratori.