What we emphasize and reward on assessments signals to students what matters to us. Accordingly, a great deal of scholarship in chemistry education has focused on defining the sorts of performances worth assessing. Here, we unpack observations we made while analyzing what "success" meant across three large-enrollment general chemistry environments. We observed that students enrolled in two of the three environments could succeed without ever connecting atomic/molecular behavior to how and why phenomena happen. These environments, we argue, were not really "chemistry classes" but rather opportunities for students to gain proficiency with a jumble of skills and factual recall. However, one of the three environments dedicated 14-57% of points on exams to items with the potential to engage students in using core ideas (e.g., energy, bonding interactions) to predict, explain, or model observable events. This course, we argue, is more aligned with the intellectual work of the chemical sciences than the other two. If our courses assess solely (or largely) decontextualized skills and factual recall we risk (1) gating access to STEM careers on the basis of facility with skills most students will never use outside the classroom and (2) never allowing students to experience the tremendous predictive and explanatory power of atomic/molecular models. We implore the community to reflect on whether "what counts" in the courses we teach aligns with the performances we actually value.