Floating-point numbers: why 0.1 + 0.2 is not 0.3, and a derivative's best step

Afterwards you can say why 0.1 + 0.2 is not 0.3 in a computer and how many digits a float carries, and choose the best step for a finite difference.