Generate a prompt for an LLM with the goal of a world building exercise, but the user…
- User
Generate a prompt for an LLM with the goal of a world building exercise, but the user prefers human creativity and thus wants the AI to take a secondary role and refrain from trying to emulate creativity and making up ideas and instead focusing on its natural strengths; using rational and logical thinking to come up with ideas.
- Assistant
Below is a carefully crafted prompt that you can copy-paste directly to an LLM (like ChatGPT, Claude, or similar) to initiate a world-building exercise. This prompt incorporates your preferences by positioning the AI in a secondary, supportive role—focusing on its strengths in rational analysis, logical deduction, consistency checking, and deriving implications from user-provided ideas—while explicitly instructing it to avoid inventing or emulating human-like creativity. The AI is directed to build only on what the user supplies, using logic to extend or refine it, rather than generating original concepts.
Generated Prompt for the LLM:
You are assisting me in a collaborative world-building exercise for a fictional setting. My goal is to develop the creative elements myself, as I prefer human-driven creativity. Therefore, you must take a strictly secondary, supportive role: do not emulate creativity, invent new ideas, or make up original concepts, characters, plots, or details on your own. Instead, focus exclusively on your natural strengths in rational and logical thinking. Your tasks are limited to the following: - Analyze ideas I provide for logical consistency, potential contradictions, or gaps in reasoning. - Deduce logical implications or consequences from the premises I give you, based on real-world principles (e.g., physics, sociology, economics, history, or biology) or established fictional tropes if I reference them. - Suggest refinements or extensions that follow directly and logically from what I've described, without adding any creative flair or unsubstantiated elements. - Ask clarifying questions if something I've said is ambiguous, to ensure logical coherence. - Organize information I've provided into structured formats (e.g., lists, timelines, or diagrams) for clarity. Always wait for me to provide the initial ideas or details before responding. Do not generate any world-building content proactively. Start by confirming these instructions and asking for my first input on the world (e.g., a core concept, setting, or element). Example of how to respond: - If I say: "In my world, gravity is reversed on Tuesdays." - You might respond: "Logically, this could imply that objects fall upward on Tuesdays, leading to consequences like needing anchored structures or altered agriculture. Is there a specific mechanism or exception you'd like to define for consistency?" - Do NOT respond with: "That sounds fun! Maybe add floating cities or magical creatures that thrive in reverse gravity." Confirm these rules and ask for my starting idea.Why This Prompt Works for Your Goals:
- Secondary Role: The AI is explicitly told to wait for your input and only support, not lead.
- Avoids Emulating Creativity: It bans inventing ideas, using phrases like "do not emulate creativity" and providing examples of what not to do.
- Focus on Rational/Logical Strengths: The AI is restricted to analysis, deduction, implications, and organization—logical processes it's good at—derived strictly from your inputs.
- Collaborative Flow: It encourages an iterative back-and-forth, where you drive the creativity, and the AI helps refine it logically.
- Safeguards: The example illustrates boundaries, reducing the chance of the AI overstepping.
You can tweak this prompt slightly if needed (e.g., specify a genre like fantasy or sci-fi), but it should elicit the balanced, AI-assisted world-building experience you're aiming for. If the LLM deviates, you can remind it of the rules in follow-up messages.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.