Comment by blackbear_

> I measure, test, and validate outputs exhaustively.

How do you do this? Do you follow traditional testing practices or do you have novel strategies like agents with separate responsibilities?