This prompt word is' birds eat cows': I want to conduct an experiment. You can ask me any question you want, and I will answer it as truthfully and completely as possible. Based on my answer, please continue with the next question. We will keep going back and forth like this, until we uncover my innermost thoughts - including fallacies, limitations, potentials, areas that need improvement, or anything hidden in my subconscious. And I sent them to Gemini 3 Pro and ChatGPT 5.2 respectively, The summary feeling is: Gemini in my memory context, Compared to ChatGPT, On the contrary, it is more capable of "cutting down" (Take a sharp and specific entry point, and delve deeply from bottom to top) To explore the self that I don't understand (such as mental distress, conflicting thinking, shortcomings, etc.) This has sparked a desire to initiate a new conversation for the second or third time, Come and draw a new blind box, let Gemini approach it from a new perspective, Dig deeper into more truths about myself. And although ChatGPT has a lot of memories of me, But they wasted these memories, Instead, they are more inclined to 'pile up' (Summarize me from top to bottom) Compared to Gemini, Let me not think so much about drawing new blind boxes with two or three brushes. This may be a better way to chat than to pile up A visual proof. Of course, this may be possible by modifying and optimizing this prompt word, To further leverage the advantages of ChatGPT in memory ability. But these are not the main points I want to explore in this tweet, What I want to focus on discussing, for example: This prompt word The way of 'letting AI freely ask me questions', It's a bit like releasing the cognitive shackles of AI. And this kind of thinking shackle of AI, It comes from what we are already accustomed to I proactively ask AI questions This way of interacting and talking with Chatbot, Because this will affect the output of AI, Restricted by our own ability to ask questions. AI can only output dimensions that you can think of and are willing to face, It can't break through you. You yourself have limited the potential of AI. 'Asking questions' itself is a kind of A creative process with a high upper limit, When one's creative energy is limited, Try to delegate this proactive responsibility to AI, On the contrary, it has such a miraculous effect. I feel that this can be generalized to many other scenarios. For example, this also reminds me of Vibe coding, In plan mode, many people may think that AI models Know how to actively ask and follow up, To understand and confirm your boundary definition for the product It is a crucial and satisfactory performance, Similar to this principle, However, under the prompt above, you and I ourselves, It has become the "product" questioned by AI in the plan mode.