I would like to describe why manual software test works at the beginning but fails eventually.