Despite the challenges that primary historical sources present for students and instructors alike, engaging students with such sources as part of their mathematical learning experiences is an increasingly common instructional practice. Yet research on teaching and learning with primary sources remains a niche activity. It is thus reasonable to ask whether the use of primary sources merits broader attention from the mathematics education community. This article explores this question through the lens of Sfard's commognitive theory of learning, which views mathematics as a discourse and mathematical learning as the process of becoming a participant in that discourse. Importantly, commognitive tenets apply to interpretations of mathematical development at both the historical and individual levels. I use these tenets to analyze a particular type of student project that adopts a guided reading approach to primary sources. Specifically, I examine examples of such projects that have been developed and used in the US to teach core mathematical topics at the university level since 2003, using commognitive theory to characterize the potential learning opportunities they afford. As with all instructional materials, these opportunities can be missed in classroom implementation. Nevertheless, the discourses represented in different instructional materials render some learning opportunities easier to offer students, while constraining the possibility of others. I thus also briefly examine some of the differences between these primary source projects and the standard textbooks that they are intended to replace.