Take microbenchmarks seriously as true statements that such & such measurements were made of particular programs.

Please don't generalize that into language X is beating language Y.