1 Million PRs, 5 Months: The Real DevOps Bottleneck Isn’t Review (Plus RubyKaigi’s 2026 Ruby Survey)
New data this week complicates the “AI broke code review” story: teams were already shipping in batches of 2–50 before AI showed up, which means deployment validation, not review, was always the ceiling. Meanwhile RubyKaigi’s fresh 2026 attendee survey shows Ruby 4.0 caught up to 3.4 in adoption within six months — and Claude Code, not GitHub Copilot, is now the default coding assistant in the Ruby world. Step through both below.
Where Does a 2026 AI-Generated PR Actually Get Stuck?
GitHub Copilot generated over 1 million pull requests in five months. If “human review” were the bottleneck, throughput should have collapsed. It didn’t — because the real constraint moved downstream. Click through the lifecycle of one PR to see where it actually waits.
Reported distribution of how teams actually ship changes in 2026 — batching, not per-PR review, sets the pace. Click a bar.
State of Ruby, per RubyKaigi’s 2026 Attendee Survey
~400 Japan-centric RubyKaigi attendees, surveyed April 2026. Pick a category on the left.


