A recent study by METR found that AI-powered coding assistants could slow down experienced developers rather than increase productivity.
Developers in the study predicted a 24% increase in speed when using AI but actually took 19% longer to complete tasks with AI tools.
The slowdown was attributed to developers' overestimation of AI capabilities, complex codebases, inconsistency in AI suggestions, and difficulty in grasping implicit context within large repositories.
Researchers caution that these results may not apply universally and suggest that AI tools could still benefit less experienced developers or those working on smaller projects.