The latest study of how developers actually use ChatGPT in the wild peels back the glamour and shows the rough edges that still matter when AI tries to help write code. It’s not a manifesto about AI replacing programmers; it’s a map of where a chatty large language model can accelerate software work—and where it can derail it. The researchers examined conversations that real developers shared on GitHub, not staged lab demos, to understand what makes an AI conversation useful for fixing bugs, adding features, or steering projects toward clearer paths. The teams behind the work hail from Drexel University and collaborators at Belmont University and Florida State University, led by Drexel’s Ramtin Ehsani. Their effort builds on a broader push to understand how conversational AI fits into professional software engineering, not just academy-grade benchmarks.
When ChatGPT Becomes a Software Engineer, What Breaks?
