Agreed. I'm not seeing any use of a automatic profiler to find the hot spots of one's code. If you don't know that, you're aiming blind on what to optimize and how much. The only profiling mentioned is something manual you have to add to each scope you want to time.
In so, so much code, basic big-O analysis, concurrency analysis (minimize mutable state / coordination), and even a little thought about CPU caches (including TLB) yields huge dividends.
