GitHub Copilot and Commit Activity Research Findings

According to a longitudinal study published on arXiv, tracking developer activity before and after adopting GitHub Copilot found no statistically significant change in commit-based activity.

On the other hand, survey responses consistently showed that developers subjectively felt their productivity had improved when using the tool. Two possible explanations are offered for this gap between the data and perceived experience.

• Limitations of commit frequency: Commit frequency may be a flawed metric for measuring productivity. If developers produce more output through fewer, larger commits, the data may make activity appear stagnant.
• Qualitative improvement in work experience: The fluidity of work may have improved—tasks feel faster and require less effort—even though the actual volume of output remains the same.

This carries important implications for companies that decide on subscriptions while expecting measurable increases in output when adopting AI tools. It suggests that the value of AI tools may lie not in quantitative figures (commit counts) but in qualitative experience (work efficiency).