A prompt is the input you give a model. Sometimes that input is just a question, but in practice it is usually a combination of instructions, context, examples, and constraints.
Prompts matter because the model is always trying to continue the pattern it sees. A better prompt gives the model a clearer pattern to continue.
More than a single sentence
In real systems, a prompt may include:
- a system message or policy layer
- user instructions
- retrieved context
- prior chat history
- examples of the desired format
- tool definitions or schemas
That is why prompt design is not just about clever phrasing. It is about shaping the full input so the model has the right context and the right boundaries.
The next chapters look at the most important prompt concepts to get right early.