This is probably related to an earlier post a while back about long integers,
exact(1e14)=1
Quote: This is probably related to an earlier post a while back about long integers, exact(1e14)=1 Really?
I am getting similar results for larger exponents (in CAS mode and non-CAS mode). The cut-off appears to be 1E14, though.
Yes, anything greater than 1e14 exhibits this behavior with exact().
approx() seems to be OK, though (regardless of the magnitude of the argument).