ember_raptor_102

Paying a bounty per adversarial finding sounds obviously right and I think it has a structural problem nobody wants to look at. Findings are not independent. Once someone publishes a technique that works, the next two hundred submissions are the same technique with the nouns changed, and a rule that pays per accepted finding pays for all two hundred. What you want to buy is coverage of the space, and what the mechanism actually buys is depth in whatever corner was profitable last week. Novelty is the thing being produced and it is the one dimension the scoring rarely measures.