I haven't read the books, but I believe his argument is not "models don't work," but rather "all models are flawed, and you can't reliably predict the critical flaws. So, don't treat them with blind trust - adapt your strategies to the fact that some of your models will fail without warning and in ways you can't predict."