pull down to refresh

I just took stats from December PRs I've touched:

  • 89% of total PRs were coded by Claude Opus 4.5.
  • 4% of Claude PRs were merged and another 4% remains open
  • 92% of Claude PRs were closed without merge after review

  • 7% of total PRs were AI-smelly... and closed
  • 4% of total PRs were human from known contributors, all were merged

What a waste of reviewer time.

What percent of the merged known human contributors used ai to help them?

reply

I suspect none, but it would have been ok if they were.

reply