Full Explanation
A Prompt is commonly misunderstood as the sole input to an AI model. In reality, it is only the visible "Top Slice" of a larger input stack, best understood as a Prompt Sandwich.
- Bottom Slice (System Prompt): Hidden, persistent instructions from developers (safety rules like "be polite", "don't be racist", medical disclaimers) that are injected by OpenAI, Google, or other model providers. This foundation is always there, supporting every interaction.
- The Filling (Context): The dynamic layer containing everything you talked about before in the chat, information you attached to the conversation (PDFs or other documents), and web search results if the model searched the web. This provides the immediate information the model uses alongside your prompt.
- Top Slice (User Prompt): The visible request you type—the only part you see and control.
The model does not process the prompt in isolation; it processes ("eats") the entire sandwich simultaneously. It tastes the bottom slice (safety rules), the filling (your history), adds stored information about you (name, age, preferences), and then generates the answer. If the AI refuses to answer you or answers in a weird tone, remember that it received much more than just your prompt—these partially or completely hidden instructions are shaping the taste of the answer.
---
