# Why Does an LLM Answer One Piece at a Time?

You ask a question. The answer starts before it exists. Follow that one visible trick through web streams, prompt packing, learned numbers, GPU memory, and the fleet keeping the loop alive.