Cool read, and agree that closing the computation > experimental loop is key!