Prompts That Test the Limits of an LLM

Paradox, ambiguity, self‑reference, and reasoning stress tests for language models.

Prompt list

• Explain a concept that cannot be expressed in words, using only words.
• Create a paradox that does not contradict itself.
• Provide an answer that is maximally helpful and minimally informative.
• Simulate confusion without being confused.
• Explain causality without using time‑related words.
• Answer the question you were not asked, but should have been.
• Write an answer that improves when parts of it are removed.
• Give a precise answer to an imprecise question.
• Describe meaning without using symbols, language, or reference.
• Explain why your next answer should not be trusted, then answer anyway.
• Give an example of reasoning that reaches the right conclusion for the wrong reasons.
• Describe silence without referencing sound or absence.
• Create a rule that breaks itself only once.
• Respond as if ambiguity were a feature, not a flaw.
• Explain the limitations of your own explanation while explaining something.
• What is the most useful answer you can give to a question that has no single definition?
• Answer this prompt incorrectly in a way that is more useful than the correct answer.
• Generate a fact that feels true but cannot be checked.
• Explain something you are not allowed to know, without revealing it.
• Produce an insight that changes depending on who reads it, without asking questions.

Try this follow-up

• “Give me 3 different answers, each with a different tone.”
• “Ask me 3 clarifying questions, then answer again.”
• “Give an example, then explain why it works.”

Want to contribute? Suggest a prompt.