Which also meant you could do a lot of optimisations - like moving the stack pointer to point to the display memory area and using 'push' to blit a chunk of image data to screen which was quicker than using an 'ld' loop.

Also having polymorphic functions where you could write over a couple of bytes to make the function behave differently at runtime - saving code memory space - and learning the hex for most of the common opcodes at same time.