[bradhilton]

Yeah, the takeaway shouldn't be "our model is smarter," but that we were able to train weak models to as good or better than the best for this specific task. Depends on what you're doing, but sometimes that is enough.