A study by METR reveals that software developers overestimate the productivity gains from AI, as they took 19% longer to complete tasks when using AI tools, despite anticipating a 24% time savings. The findings suggest that while AI may not hinder productivity, developers' trust in AI models and the complexity of mature codebases can lead to misconceptions about efficiency.