Our insanely good fuzzy classification skills let us select and continually update a solving strategy. In computing terms: writes to long-term memory are slow,

Our insanely good fuzzy classification skills let us select and continually update a solving strategy. In computing terms: writes to long-term memory are slow, but reads are fast; short-term memory is extremely size-constrained and time-volatile, but fast for both writes and reads. Efficient algorithms for human brains should rely as much as possible on pattern matches, long-term memory reads, and limited short-term storage—exactly the opposite of what you want to do on a computer! It’s no wonder that the polynomial-time multiplication algorithms that work fine on paper aren’t going to perform very well in our heads.