In an X post linking to his essay "The Reverse Information Paradox", Satya Nadella argued that AI flips economist Kenneth Arrow's old information problem on its head: instead of sellers risking too much disclosure to make information valuable, buyers may reveal their own proprietary know-how just to get useful results from AI. In the essay attached to the post, he says prompts, corrections, evals, workflow traces and agent use can become a form of institutional knowledge that model providers may learn from over time.