The article should be viewed in the context of High Scalability. I think the author is more saying that you should test which works faster, rather than solely relying on algorithmic complexity.
Maybe I have a different idea about what's important for highly scalable architectures, because in my mind ignoring algorithmic complexity theory and adhering to mantra's like 'stop using linked lists' is one of the best ways to end up with a solution that doesn't scale at all.