My most constructive advice is that CS education should "go concrete" often enough to de-mythologize the predictive powers of big-O notation, and have the cours

My most constructive advice is that CS education should "go concrete" often enough to de-mythologize the predictive powers of big-O notation, and have the course contain at least one exercise which shows how misleading it can be in practice. Maybe have them download and sort a 32GB data set on a 64GB SD card on a RPi3 board ? The smart ones will realize that downloading can be their first pass over the data, and all of them will be befuddled by the interesting performance anomalies when photo-grade SD-cards go on sabaticals, if you annoy their simplistic flash-adaptation layers. If they characterize "the trouble", they may find that obsolete sequential algorithms for tapes run faster than random access algorithms in that environment. ...which is precisely the kind of thing companies expect a CS-graduate to be good at.