AI Coding Tools Slow Down Experienced Developers by 19 Percent, Study Reveals

Does AI Speed Up Coding? New Study Finds Surprising Results

A recent study challenges the common belief that artificial intelligence (AI) tools accelerate coding tasks. According to research by the nonprofit group METR, experienced software developers actually took 19% longer to complete coding assignments when using AI-powered assistants.

The study involved 16 developers with an average of five years of experience. They worked on complex software projects both with and without AI tools, primarily relying on Cursor Pro and the Claude 3.5/3.7 Sonnet models during the AI-assisted sessions.

Expectations vs. Reality

Developers initially predicted that AI assistance would cut their coding time by about 24%. After completing the tasks, they estimated a 20% time reduction. However, the data told a different story: the AI tools actually slowed the developers down by nearly a fifth.

This unexpected outcome suggests that current AI coding assistants may introduce friction rather than streamline the workflow for experienced programmers.

Interpreting the Findings

The authors caution against broad conclusions from the study. It focused exclusively on developers with substantial experience, leaving open the possibility that less experienced coders might benefit more from AI support.

While experimental factors might have influenced the results, the consistency of the slowdown across the analysis indicates the effect is likely real rather than a quirk of the study design.

Industry Context

Despite these findings, major tech companies continue to invest heavily in AI for software development. For example, Google's CEO Sundar Pichai revealed that over 25% of new code at Google is now generated by AI systems and then reviewed by human engineers.

“This helps our engineers do more and move faster,” Pichai said during Google's Q3 2024 earnings call, highlighting the complementary role AI plays in coding rather than replacing human input.

What This Means for Researchers and Developers

• AI tools may not yet reliably speed up coding for experienced developers.
• Newer or less experienced programmers might see different effects, warranting further study.
• Human oversight remains critical when integrating AI-generated code.
• Continued research and development are needed to improve AI coding assistants' efficiency.

For professionals interested in the evolving intersection of AI and software engineering, exploring targeted AI courses can provide valuable insights. Resources such as Complete AI Training's latest AI courses offer practical knowledge on leveraging AI tools effectively.