Many people think of a language model’s parameters as “long-term memory” and user prompts as “short-term memory”—but that analogy falls short. This post breaks down the real roles of parameters, context, activations, and output in Large Language Models, explaining why it’s more accurate to think of the model as a tool and your input as the material it works on.