Depends what your goal is. Your approach is to optimize for developer productivity. Mike's is to optimize for machine productivity.

The question is whether it is possible to do both at once, hence Mike's point about the ideal language needing to find the right high-level abstractions (machine-efficient and developer-intuitive). What we currently have is somewhere in-between, but more towards the machine-efficient side. It's quite easy to produce inefficient code, without realising it, in today's high-level languages.