Prompt
A prompt is the input text given to a language model to guide what it generates.
What Is a Prompt?
A prompt is simply the text a model receives before it generates a response. It's the model's only source of instruction — everything about what to do, how to do it, and in what format has to come through the prompt, since the model has no other channel for receiving direction.
Key Idea
In a real application, "the prompt" is often assembled from several pieces — a system prompt, conversation history, retrieved content, and the user's latest message — combined into one sequence the model actually sees.
What Shapes Prompt Quality
- Clarity — a specific, unambiguous instruction produces more reliable output than a vague one.
- Examples — showing the model what a good answer looks like, as in few-shot prompting, often works better than describing it in the abstract.
- Structure — organizing a prompt with clear sections (instructions, context, examples, the actual request) helps the model parse what's being asked.
- Relevant context — including exactly the information the model needs, and no more, avoids both under-informing and diluting the prompt with noise.
Common Mistakes
Assuming "the prompt" is only what the user typed
In most real applications, the prompt the model actually sees is assembled from a system prompt, history, and other context — not just the latest user message.
Writing a vague instruction and expecting a specific result
A model can only respond to what's actually in the prompt — ambiguity in the prompt tends to produce ambiguity in the response.
Padding a prompt with irrelevant context "just in case"
Unnecessary content consumes context window and token budget, and can distract the model from what actually matters.
Interview Question
What is a prompt, and what actually determines whether a prompt produces a good result?
A prompt is the text a model receives before generating a response — its only source of instruction, since there's no other channel for telling it what to do. In a real application, it's usually assembled from several pieces: a system prompt, conversation history, retrieved content, and the user's latest message, combined into one sequence. What determines quality is clarity of instruction, relevant examples where they help, clear structure, and including exactly the context the model needs — not more, since irrelevant content consumes token budget and can dilute what actually matters.
What an interviewer may ask next
- Is a prompt always just what the user typed?
- Why might adding more context to a prompt actually make results worse?
- How do examples in a prompt compare to a purely descriptive instruction?
Explain It in 30 Seconds
A prompt is the text a model receives before generating a response — its only channel for instruction. In a real application it's usually assembled from a system prompt, conversation history, retrieved content, and the user's message combined into one sequence, not just what the user typed. Quality comes from clear instructions, relevant examples, clear structure, and including exactly the context needed, since irrelevant content wastes token budget and can dilute what matters.