Capturing the changing landscape of student code in an era defined by AI

This paper presents a longitudinal study on the transformation of programming practices among undergraduate students enrolled in the Computer Networks course at the Faculty of Computer Science, within “Alexandru Ioan Cuza” University of Iași. Spanning thirteen academic years (2012–2025) and analyzing over 4,300 individual project submissions, the study offers a rare, data-driven perspective on how software development behaviors evolve in response to technological shifts and the rise of AI-powered code generation tools. Through a hybrid methodology combining static code analysis, cyclomatic complexity evaluation, and LLM-based assessments of code quality, we uncover a profound inflection point following the public release of GPT-based tools. The findings reveal a sudden expansion in code volume accompanied by a counterintuitive simplification of structure, reflecting a transition from organically developed solutions to AI-accelerated, pattern-driven development. Skill and structure scores increased post-2022, but the stability of median values exposes a stratified adoption, raising questions about students’ conceptual mastery. These results mirror a wider global phenomenon represented by a cognitive realignment in how younger generations solve problems, reason, and create, shaped by ubiquitous access to intelligent systems. Rather than confirming a gradual shift, our data points to a disruptive pedagogical rupture, one that challenges foundational assumptions about learning, autonomy, and authorship in technical education. This study calls for a redefinition of learning outcomes in the era of generative AI, advocating for curricula that not only integrate such tools, but also cultivate reflective, critical engagement with them. The future of computing education depends not on resisting AI, but on teaching students to think beyond it.