WHILE ARTIFICIAL INTELLIGENCE WRITES THE CODE, WHAT DO WE DO? THE NEW ROLE OF THE DEVELOPER IN THE ERA OF AUTONOMOUS CODING AGENTS
Keywords:
artificial intelligence, software engineering, Claude Code, autonomous agents, AIDA methodology, productivity, university education, multimodal processing, frontier modelsAbstract
The emergence of autonomous AI coding agents profoundly reconfigures the role of the software engineer. This paper examines the evolution of AI coding assistants, from syntactic autocomplete to extended-reasoning agents such as Claude Code, and proposes the AIDA Methodology (Mey's AI-Driven Development Architecture): five structured phases for AI- assisted software development. Drawing on empirical evidence of skill-differentiated productivity gains (Dell'Acqua et al., 2023) and performance benchmarks including SWE- bench Verified (72.5 %) and Terminal-bench (43.2 %), we argue that the critical competency of the modern developer is not writing code, but designing architectures, contextualizing problems, and interpreting AI-generated solutions. We also examine the emerging phenomenon of native multimodal processing, curricular implications for university education, the competitive positioning of the professional developer, and the normative gap arising from the absence of formal standards for AI-assisted development.
