Having gone through about a hundred interviews on the employer side and a dozen on the employee side myself, I think I can drill down the essence of the problem

Having gone through about a hundred interviews on the employer side and a dozen on the employee side myself, I think I can drill down the essence of the problem to: "Everything that can be standardized is a bad predictor of hiring success." In the end, I now have a continuously A/B-tested scheme that I'm pretty happy with, but still most of it is intuition. Which is very applicable to the job itself: Intuition is nothing but unconcious and highly multidimensional optimization across the vast range of hard/soft skills required for the job by means of pattern recognition and heuristics. Everything that has simple measures (automated coding tests, phone screens by nontechnical people,...) can be gamed and produce lots of false positives AND negatives (see the recent Google screening for the GWAN guy). In the end, I can read much more real-world success from a quick coding challenge that shows me an applicants' thinking, craft, naming, library choice and so on. I'd consider myself a top 5% interviewer by now, but I have no clue how to automate, scale or replicate this.