Experienced software developers assumed AI would save them a chunk of time (fortune.com)

A recent study by Model Evaluation and Threat Research reveals that experienced software developers may not achieve the productivity gains anticipated from AI tools. In an experiment involving 16 developers who undertook 246 tasks, those utilizing AI tools like Cursor Pro and Claude 3.5/3.7 took 19% longer to complete tasks compared to when they worked without AI. This challenges the prevailing belief that AI will significantly enhance workplace efficiency, highlighting that the integration of AI can sometimes lead to increased task completion times. The findings underscore the importance of context in software development; developers often had to adapt AI outputs to fit their specific projects, which involved extensive debugging and time spent crafting prompts. The research suggests that while AI can provide useful outputs, it may not always align with the needs of experienced professionals who possess valuable contextual knowledge. This study serves as a reminder that despite the promise of AI's impact on productivity, its implementation requires careful consideration and further research to fully understand its effects in real-world scenarios.