Every new tool arrives promising to think with you, but the medieval logician Jean Buridan — famous for the donkey paralyzed between two equal bales of hay — identified the real hazard centuries before the first algorithm: when a system generates equally plausible options with no friction between them, the human agent doesn't sharpen; they dissolve. AI assistants are extraordinarily good at removing the resistance that forces you to commit. Ask for five email drafts, and you get five. Ask for a project plan, and three appear instantly. The friction of producing options used to be the mechanism by which you discovered what you actually preferred. Buridan's insight, taken seriously, suggests that the problem with offloading generative work to AI isn't laziness — it's that you're also offloading the low-grade struggle through which preferences become legible to yourself. The practical move isn't to use AI less; it's to always articulate your preference *before* you see the output, even briefly, even clumsily. One sentence, typed or spoken, before you hit generate. That prior commitment gives you a reference point to judge against rather than a blank slate to be filled.

What is the last decision you made by choosing among AI-generated options rather than testing AI output against something you'd already formed — and how would the result have differed?

Drawing from Scholastic Philosophy / Medieval Logic — Jean Buridan (Quaestiones in decem libros Ethicorum Aristotelis, c. 1330s)