Atlassian Research: What Do Developers Think About Code Readability in the Age of LLMs?

This post presents research findings on code readability in the age of LLMs, specifically comparing human-written code with RovoDev-generated code. The research surveyed 118 practitioners and conducted an empirical study measuring code readability using metrics like Lines of Code, Comment Ratio, Cyclomatic Complexity, Maintainability Index, and Halstead Metrics. Key findings indicate that RovoDev-generated code is as readable as human-written code across most metrics, with negligible differences in Java, Kotlin, Go, and Scala, and only small differences in TypeScript and Python. The research concludes that LLM agents can produce trustworthy and maintainable code, supporting the widespread adoption of tools like RovoDev Coding Agent.