The post has a good core, but is wrapped in opinions like this which detract. I agree with you entirely: I produce my best code when I test everything, and then someone else tests and discovers the parts I excluded from 'everything'.