From Typist to Collaborator: A Systematic Review of AI’s Impact on Programming Efficiency and the Shift in Developer Roles

Abstract

With continuous advancements in AI and large language models, AI in programming has evolved from simple code completion to intelligent agents with contextual understanding. This change has not only had a significant impact on the field of programming but also on programmers' employment. This paper aims to explore new working models for future programmers and analyze the efficiency and potential risks of AI programming. This paper systematically reviews key empirical studies from relevant papers published between 2020 and 2025 and finds that AI exhibits significant efficiency issues in programming. On the one hand, in LeetCode, AI performs well on low-to-medium difficulty problems, and AI programming has already surpassed 85% of human programmers in algorithm competitions. On the other hand, the actual development efficiency of AI when handling large open-source projects actually decreases by approximately 19%, and a series of safety and efficiency issues are found in AI programming, such as code redundancy and over-coupling when writing large projects. This article explores the transformation of programmers from code writers to AI collaborators, examining the advantages and disadvantages of AI in programming. It also elaborates on the specific requirements for future programmers, such as a deeper understanding of the nature of tasks and the need to maintain a dynamic balance between programmers and AI to achieve greater efficiency in project completion.