Explorando o papel dos modelos de linguagem de grande escala (LLMs) na transformação do desenvolvimento de software

Abstract

In this study, we analyze the impact of Large Language Models (LLMs) on software development, focusing on how artificial intelligence technologies—such as ChatGPT, GitHub Copilot, Claude, and Gemini—are transforming the way developers write, test, and maintain code. The main goal of this research is to understand how these tools automate programming tasks, enhance productivity, and influence technical practices in the industry. The methodology consists of a survey conducted with Brazilian software developers, aiming to identify their perceptions regarding the benefits and limitations of using LLMs in daily workflows. The results highlight that these tools are widely used, especially in coding and debugging, offering speed and efficiency gains. However, concerns about excessive dependency, security risks, and the opacity of model suggestions are also recurrent. This work contributes to the academic and professional debate by presenting empirical evidence of the transformative role of LLMs in software engineering and by proposing responsible guidelines for their adoption. In this sense, we consider that the use of LLMs in development processes requires a balance between automation and human supervision to ensure sustainable innovation.