The Short Memory

Imagine you are sitting in a classroom. There is a large white board. It is clean. It is waiting for you to write. You start by writing your name. Then you write a question. Then you write an answer. The board holds everything you have written so far. It can hold a lot. But it cannot hold forever.

This board is the AI’s memory. In technical terms, we call this space the context window. Think of it as the size of the white board. It is not a permanent storage shelf. It is a temporary working surface.

When you ask the AI a question, it looks at the board. It reads what is already there. It sees your name. It sees your question. It sees your answer. It uses all of that to understand what you want next. If the board is full, it cannot add more. It must erase something. Usually, it erases the oldest notes. The newest notes stay. The oldest notes vanish. This is not a bug. It is a rule of the board. The AI does not have a long-term memory like your brain. It does not remember what you said last week. It only remembers what is on the board right now. If you start a new conversation, the board is wiped clean. The AI starts fresh. It does not know your name. It does not know what you asked five minutes ago. This can feel strange. You might think the AI is forgetting you on purpose. It is not. It is simply following the rules of the board. The board has a fixed size. You cannot make it bigger. You can only write on it. When the space runs out, the eraser comes out. The AI does not choose what to erase. It follows a simple pattern. It keeps the most recent words. It drops the oldest ones. This is how the board stays clean. It is a necessity. Think of it like a conversation in a small room. If the room is crowded, people leave. The newest arrivals stay. The earliest guests go home. The room has no memory of who left. It only knows who is standing there right now. This is the same idea as the whiteboard. The AI does not have a long-term memory bank. It does not store your past chats in a filing cabinet. It has no shelf for old conversations. It only has the board. The board is wiped clean when you start a new chat. This is why you must sometimes remind the AI of what you said earlier. If the board is full, the AI forgets the oldest notes. It forgets the beginning of the story. It forgets the first question you asked. It forgets the rules you set at the start. This can feel confusing. You might think the AI is ignoring you. It is not. It simply ran out of space. The board is full. The oldest notes are erased to make room for the newest ones. This is the core rule of the whiteboard. It is called the context window. In plain words, this is the size of the AI’s short-term memory. It is the amount of text the AI can look at all at once. Think of it as the size of the whiteboard. A small board holds a few sentences. A large board holds many pages. The bigger the board, the more the AI can see at once. This matters because the AI does not have a permanent memory. It does not remember you from yesterday. It does not remember what you said in a previous chat. It only remembers what is on the board right now. If you close the window, the board is wiped clean. The next time you open it, the AI starts fresh. It knows nothing about you unless you tell it again. This is why you must sometimes repeat yourself. You are writing on the board again. The AI reads your new words. It does not look back at the old ones. It only sees what is currently visible. This can feel frustrating. You might think the AI is ignoring you. It is not. It simply cannot see them. The board is full. The oldest notes have been erased to make room for the newest ones. This process is called the context window. Think of it as the size of the whiteboard. A small window means a small board. The size is measured in tokens. A token is a small piece of text. It might be one word. It might be a punctuation mark. Think of tokens as the individual letters on the board. The AI counts them. It does not count whole sentences. It counts these small pieces. This is how it knows when the board is full.

Imagine you have a small notebook. You write one sentence. Then another. The pages fill up. You cannot write more until you erase something. The AI works the same way. It has a limit. This limit is called the memory limit. It is the maximum number of tokens it can hold.

If you remember only one thing

The context window is how much the AI can hold in its head at one time. Think of 128,000 words as four novels stacked on a desk. When the board is full, the oldest notes are wiped to make room for new ones.