Explainer
Prompting Basics
Getting better answers without changing the model.
A prompt is just the text you give the model. Small changes to it can produce dramatically different results, so prompting is the cheapest lever you have for improving output.
A few reliable habits: state the role and goal up front, give an example of the format you want, and break complex tasks into explicit steps. Specific constraints beat vague requests.
Treat prompting as iteration, not magic. Start simple, look at where the answer goes wrong, and add just enough guidance to fix that failure.