Amplification and automation are two outcomes of the same AI, distinguished by what the human asks it to do. Automation asks AI to replace a human contribution: the output happens with less of you in it. Amplification asks AI to extend a human contribution: the output carries more of your judgment, reach, or craft than you could produce alone.
The distinction is behavioral, not technical. The third essay in the series shows the same tools producing opposite results in live performance: in one case the AI-assisted script quietly compressed the people on stage, optimizing out their real-time judgment, curiosity, and connection; in the other, the AI amplified exactly those qualities. Nothing about the model differed. What differed was what it was being asked to amplify.
The test to apply to your own AI use: after the AI's contribution, is there more of your judgment in the result, or less? Compression is rarely chosen on purpose. It accumulates through small delegations that each seem harmless, until the human contribution the work was supposed to showcase has been optimized away. Which is why this distinction belongs at the center of any designed AI×HI relationship.