Today's peer review process for scientic articles is unnecessarily opaque and oers few incentives to referees. Likewise, the publishing process is unnecessarily inecient and its results are only rarely made freely available to the public. In this article we outline a comparatively simple extension of arXiv.org, an online preprint archive widely used in the mathematical and physical sciences, that addresses both of these problems. Under the proposal, editors invite referees to write public and signed reviews to be attached to the posted preprints, and then elevate selected articles to published status.