Conformal Prediction (CP) has been recently used successfully as a distribution-free method for creating statistical guarantees for the performance of black-box Machine Learning (ML) models. For example, we can use it to construct sets that are guaranteed to contain the ground-truth class with a pre-specified probability such as 90%. This work started with a question: is CP the only way for obtaining such guarantees? Interestingly, we find that we can obtain similar bounds by standard concentration, albeit with a penalty on the sample complexity (quadratic vs linear for CP). On the other hand, our simple derivation covers not only the standard classification-based CP, but also Conformal Risk Control, a very recent and technically not easy extension of CP. Overall, we think that the intersection between CP and classical concentration / learning theory is interesting for future exploration.