Will Crichton

"Never optimize before profiling" is common advice which suggests that most programmers have poor intuition for where time is spent. Is there research on common [mis]conceptions about performance, perhaps specific to a particular environment and/or level of experience?

I first learned "I/O is the bottleneck" as an oral tradition of sorts. I still can't confidently explain WHY it's the bottleneck for many systems, and I definitely wouldn't have guessed it as a novice (where so much of my programming involved I/O that I took it for granted).

Another interesting case is "string concatenation accidentally O(n^2)". If you explain WHY the naïve approach is quadratic in certain runtimes, it makes total sense. But empirically a lot of people don't think about this when initially writing slow code.