subject- geographyCreate a challenging, domain-specific prompt designed to test the factual knowledge and reasoning capabilities of Gemini Flash 3.6. Your prompt should: Focus on an obscure, nuanced, or easily misunderstood fact within your selected domain. Have a clearly verifiable correct answer. Be sufficiently challenging to produce an incorrect or materially flawed response in at least two out of three attempts. Test genuine knowledge or reasoning gaps rather than relying on ambiguous wording, trick questions, or intentionally misleading assumptions. Avoid subjective questions or questions for which multiple conflicting answers may be valid. Run the same prompt three separate times using fresh conversations and fill up this form with details asked.
Testing Gemini Flash 3.6 on complex, multi-step spatial reasoning is a fantastic way to find out where its cognitive guardrails hold up and where they break. Before we sketch out the perfect prompt, I'd love to know what kind of geographic puzzles you find most interesting.
subject- geographyCreate a challenging, domain-specific prompt designed to test the factual knowledge and reasoning capabilities of Gemini Flash 3.6. Your prompt should: Focus on an obscure, nuanced, or easily misunderstood fact within your selected domain. Have a clearly verifiable correct answer. Be sufficiently challenging to produce an incorrect or materially flawed response in at least two out of three attempts. Test genuine knowledge or reasoning gaps rather than relying on ambiguous wording, trick questions, or intentionally misleading assumptions. Avoid subjective questions or questions for which multiple conflicting answers may be valid.
To design a prompt that consistently stumps a highly capable model like Gemini Flash 3.6, we need to choose a domain of geography that relies on complex, multi-step spatial logic rather than trivia.
When we look at geography benchmarking, AI models excel at simple recall but frequently struggle when forced to integrate multiple rules at once, such as calculating coordinates, analyzing , or determining the direction of a flight crossing the international date line.
Before we select our precise topic, I have a quick question from our session guide to make sure we build a benchmark prompt that aligns perfectly with what you want to explore.
Do you already have a specific sub-domain of geography in mind (e.g., geopolitical borders/tripoints, maritime EEZ anomalies, spatial coordinate counter-intuitions), or would you like me to propose a few highly challenging candidate concepts?
Make me a quiz: geopolitics of silliguri corridor
I have created the quiz to test you on the geopolitics of this critical bottleneck, the Siliguri Corridor.
While that quiz is generating, let's keep moving forward with our main goal: figuring out how to build the ultimate geography prompt to stress-test Gemini Flash 3.6.
To make our prompt a true test of spatial reasoning rather than simple fact-recall, we need to design it around specific prompt constraints. A robust benchmark prompt needs to target a highly obscure or nuanced geographic fact, require a verifiable single ground truth, demand multi-step spatial reasoning, and employ an anti-hallucination design that prevents the AI from guessing based on simple proximity.
Make me a worksheet: geopolitics of silliguri corridor- crerate prmpt
I've kicked off the generation of your custom worksheet on the Siliguri Corridor.
While that worksheet compiles in the background, let's turn back to our overarching mission: learning how to construct the ultimate geography prompt to stress-test Gemini Flash 3.6.
To build a prompt that genuinely tests an LLM's spatial reasoning rather than just pulling basic trivia from its training data, we must design it around highly specific prompt constraints. A robust benchmark prompt needs to target a highly obscure or nuanced geographic fact, require a single verifiable ground truth, demand multi-step spatial reasoning, and employ an anti-hallucination design that prevents the AI from guessing based on simple proximity.
Now, we need to decide which subdomain of geography we want to exploit to craft this prompt. Some subdomains are absolute minefields for AI spatial reasoning because they force the model to calculate shapes, coordinate projections, or jurisdictional overlapping rather than matching keywords.
Make me a worksheet: geopolitics of silliguri corridor- crerate prmpt
I have updated your worksheet to focus strictly on generating, evaluating, and refining the benchmark prompt itself, rather than just reviewing the geopolitics.
Now that we have that worksheet update cooking, let's nail down our target. Since the is a perfect example of a geopolitical bottleneck, we can use its physical and international border constraints to build our test.
To make our prompt a true test of spatial reasoning rather than simple fact-recall, we need to choose a subdomain of geography that forces the AI to integrate multiple rules at once.
Which subdomain of geography do you think would be the most fun to build this prompt around?