Why Your AI Chatbot Forgets: Context Windows Decoded

A context window is the AI's short-term memory measured in tokens, not words — once full, the oldest text silently drops off.

Models read best at the start and end of a long chat; details buried in the middle get ignored — the 'lost in the middle' effect.

Indian-language text eats far more tokens per word than English, so Hindi or Tamil chats hit the limit faster.

'Memory' features are a separate fact-store, not a bigger window — they only save a few facts you've flagged.

Start a fresh chat for new topics and paste only the lines that matter to keep answers sharp.