After weeks of waiting, the email finally arrives from Physical Review Z with the reports on your paper “A theory of the nonuniform browning of toast”. Hope turns to despair as you read the editor’s cover letter saying that because the reviews contain a sufficiently strong criticism of your paper it cannot be accepted. Then you read the reports and discover that Referee A said “This is a wonderful paper, full of interesting new results that will surely be of interest to a wide audience. I am particularly impressed with Eq. 7 and its consequences and expect it to have broad applicability in physics.” So what’s the problem? Well, Referee B said “I fail to see any great significance in the results presented, and doubt the paper will be of broad interest. In addition, the result, Eq. 7, is wrong, calling into question the entirety of the subsequent results”. Besides wondering why the editor listened to Referee B and not A, or at least why he did not try to figure out whether Eq. 7 really does have a problem, this hypothetical scenario points to a deeper problem with the peer review system at most journals (including most physics journals): With a few exceptions, there is no mechanism for the referees and the editor to discuss the paper and arrive at a consensus recommendation before reports are communicated to the authors. Instead, the initial recommendation is based on some sort of implicit averaging (in the editor’s mind) of the reports and it is left up to the author to argue back from what might have been an error on the part of the referee, or more generally to somehow try to answer what are often conflicting and contradictory reports. Yes, it is true that in subsequent rounds of review the referees often see earlier reports, but again there is no consensus-building process. All of us who spend our professional lives dealing (on both sides) with this deeply flawed process deserve better.