Lesson 32 · 2 min
Why AI forgets in long chats
Long chats slowly get worse because the model has to juggle more and more text, and older details get crowded out.
You start a chat, set some rules, and for a while the answers are sharp. An hour later the chatbot repeats itself, ignores a rule, or mixes up two versions of your plan. You are not imagining it.
Every reply is written by rereading the whole conversation so far. As the chat grows, there is more to weigh, and details in the middle tend to get less attention than the start and the end.
When the chat gets longer than the model's context window, the app has to cut or compress the oldest messages. Whatever was in them can be gone for good.
Old drafts also cause trouble. If you changed your mind three times, all three versions are still in the chat, and the model may pick the wrong one.
Do it now
Reset a tired chat
- 1. Find a long chat where the answers have started to slip.
- 2. Ask for a clean summary with the prompt below.
- 3. Check the summary, fix anything wrong, then paste it into a brand new chat and carry on.
Write a short handover note for this chat: my goal, the final decisions (only the latest version of each), the rules I gave you, and what is left to do. Keep it under 150 words.
Quick check
Keep going
Next →How chatbots "reason" step by stepWords from this lesson
- Context window
- The maximum amount of text an AI can look at in one go. Anything beyond it is forgotten for that conversation.
- Memory
- A chatbot feature that saves facts about you between conversations so you do not have to repeat them.
- Token
- A chunk of text, often part of a word, that AI reads and writes in. Limits and prices are counted in tokens.