Prompt Engineering, Plainly

What actually improves the output of a language model - clear, specific, well-structured instructions - and what's just folklore, explained without the hype.

1. A Prompt Is an Instruction, Not a Spell A language model predicts the most likely continuation of your text, so clear, specific, well-structured instructions get better outputs and vague ones get vague outputs.
2. The Techniques That Actually Help The handful of prompting techniques that genuinely improve output: give context and a role, be specific about format and length, show examples, break hard tasks into steps, and say what to do rather than only what not to.
3. Plain Limits What prompting can't do: it can't add knowledge the model lacks, can't guarantee correctness, and 'magic phrases' are mostly folklore - the real work is iteration, testing on real cases, and watching for prompt injection.