Prompt

What is it?

A Prompt is the input signal provided to a Large Language Model to steer its output. Because LLMs are probabilistic engines completing a sequence, the prompt sets the initial trajectory for that completion.

It is not just a “query”; it is programming in natural language. It defines the constraints, style, persona, format, and task parameters for the model.

Why is it Important?

• Control Mechanism: It is the primary way we control the behavior of non-deterministic models.
• Task Definition: A good prompt turns a generic text predictor into a specific tool (e.g., “Act as a Python expert” vs. “Act as a poet”).
• Performance Optimization: Techniques like Chain-of-Thought (asking the model to “think step by step”) dramatically improve reasoning accuracy without changing the underlying model.

Technical View

From a system view, the prompt is concatenated with system instructions and user history before being tokenized.