This paper proposes a simple yet highly effective method for continual learning that performs extremely well. It is well written, thorough, novel, timely, and interesting, with the potential for broad impact on the subfield of continual learning. However, there were concerns among multiple reviewers about how the paper represents prior work (specifically how FDR, GEM and A-GEM are represented by the authors), and how DER/DER++ are different from existing methods. Given the potential of the impact of this paper, the authors have a responsibility to correct these characterizations of prior work, and it is expected for the authors to do so before publication.