This classroom experience report examines early identification of at-risk students through behavioral monitoring during in-class programming exercises. In two Computer Science courses (n=57), we implemented automated code snapshot collection at 20-second intervals during programming activities that immediately followed instructor demonstrations. Analysis revealed three distinct engagement patterns: “Early Birds” (44.4%) who accessed exercises promptly and achieved strong academic outcomes, “Active Strugglers” (33.3%) who participated consistently but struggled with problem-solving, and “Delayed/Disengaged” (22.2%) who exhibited 10+ minutes access delays during class. These behavioral profiles enabled identification of at-risk students weeks before midterm examinations. We developed targeted interventions based on observed patterns: struggling students received practice materials tailored to their specific difficulties. Among 11 students receiving pre-midterm interventions, 81.8% achieved high performer status (both exams +85%), substantially exceeding baseline expectations. This practical approach demonstrates how routine classroom activities can yield actionable insights for early intervention. While these results are promising, we acknowledge limitations including small sample size and single-institution context. This experience suggests that simple behavioral monitoring during existing instructional activities offers a scalable approach for early student support.