The Effect of Vibe Coding on the Programming Skills of Beginner and Advanced Students at Telkom University Purwokerto

The growing use of generative AI has introduced vibe coding as a new programming approach in higher education, but its varying impacts on students’ experience levels have not yet been extensively explored in the Indonesian context. This study investigates whether vibe coding affects programming competencies differently between early-stage and advanced-stage students in the Software Engineering Program at Telkom University Purwokerto. A comparative design with a mixed-methods approach was used, involving 30 respondents: 15 early-stage students (Semester 4) and 15 advanced-stage students (Semester 6). Data were collected via a five-point Likert-scale questionnaire measuring the intensity of vibe coding (Section B) and perceptions of programming competence (Section C), supplemented by a prompt quality rubric validated by faculty members based on a case study of a circular single-linked list. Quantitative analysis employed Spearman’s rank correlation and the Mann-Whitney U test, while qualitative findings from the rubric were triangulated with questionnaire results. The results indicate that advanced-level students have a significantly higher coding vibe intensity (U = 57.5; p = 0.023), while beginner-level students report a significantly higher perceived ability (U = 39.0; p = 0.002). The Spearman test yielded a non-significant negative correlation in both groups, with the final-year group showing a moderate trend (rs = −0.50). No significant differences were found in prompt output quality between groups (U = 92.0; p = 0.400), and triangulation revealed inconsistencies between perceived ability and actual output quality. These findings indicate that vibe coding without structured instructional guidance has the potential to widen the gap between perceived and actual programming competence, underscoring the need for the explicit integration of prompting literacy into the programming curriculum.