100s of world-class researchers may be good at coming up with new ideas and hypotheses towards AGI, but are they good enough programmers to test all of them in reasonable time, with relevant data sets?