I agree with you, that seems to be what the author is saying, and his conclusion doesn't follow. Also, I'm not sure if he really means mocking or any kind of test doubles.

I thought everyone agreed that writing code upfront before you need it is always a bad idea, especially specific implementations like a database.