The growing body of empirical studies on discourse in mathematics education have drawn on, broadly, sociocultural or situated views of learning but do not make issues of power central and/or explicit. For this reason, then, we situate the present analysis within the intersection of two bodies of literature: professional development focused on mathematics classroom discourse and Foucauldian analyses of power relations. In our poststructural analysis of a mathematics classroom, we interrogate the use of surveillance and its relation to meaningful mathematical discourse. Using Foucault's "techniques of surveillance" and "modalities of power," we analyze videos from one teacher-researcher's (TR) classroom at two snapshots in time to answer the questions: "What happens to a mathematics TR's use of teacher discourse moves (TDMs) and surveillance practices across three years of doing action research on their classroom discourse?" and "How might we know whether these TDMs are doing 'good' things in the classrooms?" Our findings show the complex ways that a TR's work on classroom discourse is intertwined with modes of power and techniques of surveillance. First, we found that shushing and silencing, normalized as a part of teaching, decreases when a teacher positively values the use of (mathematical) language by students. This decrease in shushing and silencing details what we understand to be 'the good' and constitutes the primary finding of this study as we seek to expand what is understood by "enhancing student language": an overall increase in language use by students and an overall increase in their use of mathematical language.