Artificial intelligence coding tools are supposed to speed up software development, but a recent study found the opposite. Researchers from Model Evaluation & Threat Research (METR) tested these tools in a controlled trial and discovered that they made developers slower. The study involved 16 experienced developers who worked on large open-source projects.
The developers predicted that using AI tools would reduce their work time by 24%, but after completing the tasks, they estimated it actually took them 19% longer to finish. The study attributed this slowdown to five factors: over-optimism about AI usefulness, high developer familiarity with repositories, large and complex repositories, low AI reliability, and implicit repository context.
The researchers note that their findings are specific to experimental tools and conditions, but caution that they should be reviewed in a narrow context. They also emphasize that current AI systems can still improve developers’ productivity in certain settings, but future models may do better. The study’s results highlight the need for developers to carefully consider how AI tools impact their work and find ways to optimize their use.
Source: https://www.theregister.com/2025/07/11/ai_code_tools_slow_down