In Defense of the Indefensible: A Very Naive Approach to High-Dimensional Inference