How do you know when the LLM prompt of your app is really improved when you change it? How do you know what LLM actually performs best with this prompt? What I'm building can half-automates the former and fully automates the latter, if you so desire. It provides a structural way of working so insecurity about the LLM component of your app will be history.