>That's not a fair reading. I don't see anywhere that the article makes a any suggestion about when to optimize, it only suggests techniques that you could presumably use when you actually need to.

Then you don't have a fair reading of my comment. I didn't say to never use it. There may be time-sensitive areas that perform a lot of work, very often. In those cases, go nuts, squeeze out every ms per frame you can. In the vast majority of cases, don't bother. Focus on maintainability and architecture and choose the right data structures (that's where your bottlenecks are going to be). Don't bother with low level details.