As someone with CS background but no experience in academia, this is true but only up to a point. Optimizations like cache efficiency and branch predictions are

As someone with CS background but no experience in academia, this is true but only up to a point. Optimizations like cache efficiency and branch predictions are important and often trump big O considerations, but are mostly engine-level issues. If you are normalizing for the average programmer, the #1 reason why their program is slow is that they're using the wrong algorithm (often implicitly with e.g. library functions or database queries). If anything I'd like more people to be aware of how to reason in big O terms rather than less.