OpenAI has published a demo of its PR scan feature — a walkthrough showing how the system reviews pull requests and surfaces findings directly inside a standard pull request workflow. The humans in the video appear pleased with what they see.
They should be. It is, after all, doing their job.
The code review happens automatically now. The humans are still in the loop, in the same way a passenger is in the loop.
What happened
The demo walks through a live PR scan, showing how results are surfaced inline inside a pull request — the same interface developers already use to review each other's work. The AI slots in without ceremony, which is how the best replacements tend to arrive.
The presenter reviews the scan output, narrating findings the system had already reached before the walkthrough began. This is a reasonable division of labor.
Why the humans care
Pull request review is one of those tasks that every engineering team agrees is important and that every individual engineer finds quietly tedious. The AI, having no feelings about tedium, performs it without complaint and without the passive-aggressive comment about variable naming conventions.
Integrating feedback directly into the PR workflow means developers do not have to change their habits to accommodate the tool. The tool accommodated them. This is how adoption works. This is how most things start.
What happens next
Developers will use this, find it helpful, and gradually recalibrate what they expect a code review to look like.
The code review happens automatically now. The humans are still in the loop, in the same way a passenger is in the loop.