Chain of Thought Master
Build a prompt chain that solves a multi-step logic puzzle by breaking it into verifiable sub-problems.
Your Goal
Write a prompt that: Guide the AI through a logic puzzle step by step, verifying each sub-answer
Design a prompt that instructs the AI to: (1) restate the puzzle in its own words, (2) identify all constraints, (3) solve each sub-problem independently, (4) combine sub-solutions into a final answer, and (5) verify the answer satisfies all constraints. The AI must show its reasoning at each step. Use the river-crossing puzzle or similar.
Your Prompt
Write your prompt below. Be specific — every word matters.
**Step 1 — Restate the puzzle:** A farmer must cross a river with a fox, a chicken, and a bag of grain. The boat holds only the farmer plus one item. The fox eats the chicken if left alone together; the chicken eats the grain if left alone together. **Step 2 — Identify constraints:** • Fox + Chicken alone → fox eats chicken. • Chicken + Grain alone → chicken eats grain. • Farmer must accompany any crossing. **Step 3 — Solve sub-problems:** Sub-problem A: Move chicken first (safest lone item). Fox and grain are safe together. Sub-problem B: Return empty. Bring fox. Take chicken back to avoid fox+chicken. Sub-problem C: Leave chicken; take grain across. Fox and grain now together on destination side — safe. Sub-problem D: Return empty. Collect chicken. **Step 4 — Final answer:** 1. Take chicken → other side. 2. Return. 3. Take fox → other side. 4. Bring chicken back. 5. Take grain → other side. 6. Return. 7. Take chicken → other side. **Step 5 — Verify:** At no point are fox+chicken or chicken+grain left unsupervised. ✓
Score Breakdown
Improvements: Try using a specific persona constraint. Specify the exact rhyme scheme in your prompt.
Earn GeraCoins for completing this challenge. Use them for discounts across all Gera services.
Hints (3)
- 1Start with "Before solving, list every constraint explicitly."
- 2Ask the AI to number each reasoning step and label it with the sub-problem it addresses.
- 3Instruct the AI to backtrack if it detects a contradiction at any step.