Comment by lloydatkinson

> they don't control for breaking the application's tests.

Essentially it's an absolutely worthless study then. This AI fatigue got boring a long time ago, this is just painful now.