It turns out that the maximum of the third derivative
is at the endpoint anyway (which you
could demonstrate by considering the sign of the derivative, for
example). Nathan's bound on f''' (1.5*exp(.5)) was just a little larger
than necessary (too "conservative"):