프롬프트 건틀릿: 어떤 프롬프트든 20가지 변형을 생성하고 순위를 매기세요.

Phase 1 (Intent): The user wants a structured, multi-phase creative exercise that transforms a single prompt into 20 divergent variations, then rigorously evaluates them against a defined intent. The audience is prompt engineers or AI enthusiasts seeking systematic ideation. Logical dependencies: intent definition → archetype-based divergence → expansion → scoring/ranking. Phase 2 (Divergence): 1. Hyper-Technical: "Parameterized Prompt Mutation Matrix" 2. Whimsical: "Twenty Mirrors of a Single Wish" 3. Pragmatic: "Idea Branching for Constrained Outputs" 4. Philosophical: "The Ontology of Prompting" 5. Playful: "Prompt Carnival: 20 Rides, One Ticket" 6. Minimalist: "Seed to Forest" 7. Futuristic: "Neural Divergence Protocol" 8. Nostalgic: "The Prompt Alchemist's Notebook" 9. Analytical: "Semantic Variance Under Fixed Intent" 10. Poetic: "A Question, Refracted" 11. Educational: "From One to Many: A Prompt Design Lab" 12. Chaotic: "Prompt Roulette: 20 Unstable States" 13. Strategic: "Optimizing Prompt Diversity for Robustness" 14. Mythic: "The Twentyfold Oracle" 15. Scientific: "Controlled Divergence in Language Model Inputs" 16. Artistic: "Palette of Prompts: 20 Hues" 17. Economic: "Cost-Benefit of Prompt Variation" 18. Zen: "One Prompt, Twenty Silences" 19. Cyberpunk: "Prompt Shards in the Data Stream" 20. Absurdist: "A Prompt Walks into a Bar... 20 Times" Phase 3 (Expansion): 1. "Given a base prompt, generate 20 distinct mutations by altering parameters (temperature, max tokens, system role) while preserving core intent. Output as a matrix with scores." 2. "Imagine a single wish spoken to 20 different genies. Write each genie's version of the wish, from literal to metaphorical, and rank which best fulfills the original desire." 3. "Take a user request and create 20 practical variations for different deployment contexts (chatbot, API, batch processing). Ensure each is self-contained and executable." 4. "Define the essence of a prompt as a philosophical proposition. Then derive 20 logical consequences, each phrased as a new prompt, and evaluate their fidelity to the essence." 5. "Design a carnival where each booth offers a different spin on the same prompt. Describe 20 booth experiences, from silly to serious, and judge which booth delivers the truest outcome." 6. "Start with a seed prompt. Grow 20 distinct trees, each with unique branches (style, tone, length). Rank the trees by how well they preserve the seed's DNA." 7. "Simulate a neural network that splits a prompt into 20 parallel universes. For each universe, write the prompt as it would appear under different training data biases." 8. "Compile a wizard's notebook with 20 spells, each a rephrasing of the same incantation. Test each spell's power to evoke the intended response." 9. "Analyze a prompt's semantic core. Produce 20 variations that shift surface features (vocabulary, syntax) but not deep meaning. Score each on semantic distance." 10. "Write a poem where each stanza is a different version of the same question. Then extract each stanza as a standalone prompt and rank their poetic and functional merit." 11. "Create a lab exercise: given a prompt, students must produce 20 variations for different audiences (child, expert, skeptic). Grade each on clarity and intent retention." 12. "Spin a roulette wheel 20 times. Each spin yields a random constraint (e.g., 'no verbs', 'max 5 words'). Apply each constraint to the original prompt and evaluate the results." 13. "For a critical system, generate 20 prompt variants to test robustness against adversarial inputs. Rank by resilience while maintaining original intent." 14. "Treat the prompt as an oracle's prophecy. Write 20 interpretations, from literal to symbolic, and determine which interpretation best guides action." 15. "In a controlled experiment, vary one independent variable (e.g., politeness, specificity) across 20 conditions. Report which condition best achieves the prompt's goal." 16. "Mix 20 colors on a palette, each representing a different emotional tone (joy, urgency, calm). Apply each tone to the prompt and judge the aesthetic outcome." 17. "Calculate the cost (time, tokens) of 20 prompt variations. Rank by efficiency and effectiveness, assuming a fixed budget." 18. "Meditate on a prompt. Produce 20 silent variations (e.g., removing all adjectives, using only nouns). Score each on minimalism and essence." 19. "In a cyberpunk world, a prompt is a data shard. Fragment it into 20 corrupted versions, each with a different glitch (missing punctuation, swapped words). Rank by recoverability." 20. "Tell a joke where the setup is the prompt, repeated 20 times with different punchlines. Each punchline is a new prompt. Rank by humor and utility." Phase 4 (The Gauntlet): Scores (0-100): 1. 92, 2. 88, 3. 85, 4. 80, 5. 78, 6. 75, 7. 90, 8. 82, 9. 87, 10. 70, 11. 84, 12. 65, 13. 89, 14. 72, 15. 86, 16. 68, 17. 60, 18. 74, 19. 79, 20. 55. Top 5: 1. **Parameterized Prompt Mutation Matrix** (92) - It directly operationalizes divergence with measurable variables, aligning perfectly with the intent's demand for systematic evaluation. Its technical framing ensures each variation is testable and reproducible, beating others that rely on metaphor. 2. **Neural Divergence Protocol** (90) - It bridges the whimsical and technical, offering a clear mechanism (parallel universes) that maps to the 20-variation requirement while preserving intent through bias control. Its sci-fi angle engages without sacrificing rigor. 3. **Controlled Divergence in Language Model Inputs** (89) - This variation excels in scientific precision, treating the prompt as an experimental variable. It outperforms poetic or absurdist versions by providing a falsifiable framework for scoring. 4. **Semantic Variance Under Fixed Intent** (87) - It isolates the core challenge (semantic fidelity) and offers a clean metric (distance), making it more actionable than narrative-based expansions. Its analytical clarity beats playful or nostalgic takes. 5. **Optimizing Prompt Diversity for Robustness** (86) - It adds a practical layer (adversarial testing) that the intent implies but doesn't state, showing strategic depth. This beats educational or economic versions by focusing on outcome quality over process.