Agent Memory
An AI model does not remember anything by itself. An agent needs memory to finish long tasks and to know you next week. This lesson shows how short-term and long-term memory work, and what to save and what to forget.
New words
- Short-term memory
- The chat so far, kept inside the context window for this task.
- Long-term memory
- Facts saved outside the model, so they last after the chat ends.
- Summary
- A short text that keeps the main points of a long chat.
- Memory store
- The place where notes are saved, such as a file or a database.
- Retrieval
- Finding the right saved note and adding it to the prompt.
- Stale memory
- A saved fact that was true once but is not true now.
What you will learn
- The tutor that forgot your exam date
- A model starts fresh every time
- Short-term memory: the chat in the window
- Long-term memory: notes outside the model
- Try it: a tutor bot with a small window
- What to save, and what not to save
- Trade-offs
- In a real system
- Interview tip
- Key takeaways
10 quiz questions at the end.
The full lesson, with interactive demos and the quiz, opens after you sign in with Google. It is free.