Some work is valuable because of what it produces. Some is valuable because of what it does to the person doing it. The difficult part is telling them apart. Writing an uncomfortable first paragraph, wrestling with a concept, and making a decision with incomplete information can all be slow. They can also be how you develop judgment. Speed up every one of those moments and you may get a better-looking result while skipping the practice you needed. That is a design question, not an argument against AI. Which parts of the process are overhead, and which are the reason you are doing it? Choose your friction. Suppose you want to understand an idea. Having AI write a fluent explanation might help you get oriented. Having it ask a question and wait while you struggle with an answer asks something different of you. One delivers an artifact. The other gives you a chance to grow. You can choose both, at different moments. Use a fast explanation to find your bearings, then close it and reconstruct the idea yourself. Ask for criticism of your reasoning. Notice the difference between recognizing an answer and being able to produce one. A small experiment Pick one ability you want to strengthen. For the next AI conversation, ask the model to question your work before it offers its own. At the end, write down one thing you can now do without it. Hesym's point of view: the best collaboration leaves you with more than a finished task. It leaves you better equipped for the next one.