What are common technical mistakes to avoid when crafting prompts?
David McCarthy
•
Last updated on May 15, 2025 at 6:00 PM
- Over-specifying—too many constraints can confuse the model or produce contradictions.
- Using formatting the model can't parse (unsupported HTML tags, complex tables).
- Ignoring temperature / max-tokens—sometimes a parameter tweak beats another paragraph of instructions.
- Exceeding context length—outputs get cut off mid-sentence when you break token limits.
- Failing to test variations—a prompt that works once may fail under slightly different wording.