Flow-by-Flow: Content-Judgment Bypass for Governing AI Output in High-Loss Domains
The episode discusses the paper "Flow-by-Flow," which proposes a new method for governing AI output in high-loss domains. The authors argue that quality is not the problem, but throughput: AI volume grows faster than human capacity. The solution is to bypass content judgment entirely by measuring the cognitive cost of submissions, thereby managing the flow rather than judging the content.