ReadAgent is a simple prompting system that leverages the advanced language capabilities of Large Language Models (LLMs) to determine which content to store in its memory. It compresses these memory sets into concise summaries called 'key point memories'. When ReadAgent needs to recall relevant details to complete a task, it takes action to consult the original text. By utilizing key point memories, ReadAgent can capture global context while attending to local details, effectively reasoning over very long contexts. This efficiency in handling large amounts of information is crucial for understanding.