A recent study conducted by the nonprofit research organization METR has found that the use of artificial intelligence (AI) coding tools can slow down experienced software developers by 19%. The study tracked 16 veteran developers as they addressed 246 real GitHub issues within million-line codebases they were already familiar with. Half of the tasks were completed with AI assistance, using tools such as Cursor and Claude. Contrary to popular belief that AI accelerates coding productivity, the findings suggest that seasoned developers may become less efficient due to the need to provide context, verify, and correct AI-generated code. Experts, including Ramine Tinati from Accenture’s APAC Center for Advanced AI, have noted that while AI tools may enable faster task completion, this does not necessarily translate into greater overall productivity. The study highlights the continued importance of developer expertise and understanding of code functionality, even in the age of generative AI.