Cognitive performance is both heritable and sensitive to environmental inputs and sustained practice over time. However, it is currently unclear how genetic effects on cognitive performance change over the course of learning. We examine how polygenic scores (PGS) created from genome-wide association studies of educational attainment and cognitive performance are related to improvements in performance across nine cognitive tests (measuring perceptual speed, working memory, and episodic memory) administered to 131 adults (N = 51, ages = 20–31, and N = 80, ages = 65–80 years) repeatedly across 100 days. We observe that PGS associations with performance on a given task can change over the course of learning, with the specific pattern of change in associations differing across tasks. PGS correlations with pre-test to post-test scores may mask variability in how soon learning occurs over the course of practice. The associations between PGS and learning do not appear to simply reconstitute patterns of association between baseline performance and subsequent learning. Associations involving PGSs, however, were small with large confidence intervals. Intensive longitudinal research such as that described here may be of substantial value for clarifying the genetics of learning when implemented as far larger scale.