Foundations
Prompt
The instructions and context you send to a model — in production, usually assembled by code rather than typed by a person.
In a chat window the prompt is what you type. In a production system it is constructed programmatically: a system prompt describing the model's role and constraints, plus whatever data the task needs, plus the user's actual request.
Prompt quality genuinely matters, but it is often over-emphasised. In real systems the bigger wins usually come from giving the model better information (retrieval) and better tools (tool calling) rather than from more elaborate wording.
Treat prompts as code: version them, review changes, and measure whether an edit actually improved anything. See eval.