Harjot Gill, co-founder and CEO of CodeRabbit, told OpenAI he pen tests his own platform with cyber models that break into live systems without saying if access is local or remote.
His answer is to fight fire with fire.
Gill said he pushed 2,500 to 3,000 pull requests in the last 30 days and called code review the first bottleneck every company hits after adopting coding agents.
He said OpenAI has partnered since inception and each reasoning model release brought large gains in finding issues at scale.
OpenAI credits CodeRabbit with shipping 20 times more software with the same team using Codex to govern AI-generated code, a production claim beyond the Holiday Extras Codex rollout and its two-week rebuild.
What false-positive rate will developers accept before they start ignoring an AI reviewer that blocks every pull request?
