We discuss Cramer's contributions to analytic number theory in a modern context, in particular showing that if one believes Cramer's heuristic that there are infinitely many pairs of consecutive primes \( p, q\) for which \( q-p\geq \{ 1+o(1)\} (\log p)^2\), then one should believe that there are infinitely many pairs of consecutive primes \( p, q\) for which \( q-p\geq \{ 2e^{-\gamma}+o(1)\} (\log p)^2\).
Article