The Web has greatly reduced the barriers to entry for new journals and other platforms for communicating scientific output, and the number of journals continues to multiply. This leaves readers and authors with the daunting cognitive challenge of navigating the literature and discerning contributions that are both relevant and significant. Meanwhile, measures of journal impact that might guide the use of the literature have become more visible and consequential, leading to "impact gamesmanship" that renders the measures increasingly suspect. The incentive system created by our journals is broken. In this essay, I argue that the core technology of journals is not their distribution but their review process. The organization of the review process reflects assumptions about what a contribution is and how it should be evaluated. Through their review processes, journals can certify contributions, convene scholarly communities, and curate works that are worth reading. Different review processes thereby create incentives for different kinds of work. It's time for a broader dialogue about how we connect the aims of the social science enterprise to our system of journals