How an answer types itself out word by word: streaming over HTTP
An AI answer appears gradually not because of a visual effect but because of streaming: an HTTP response body is a stream, so the server can send it in chunks and the browser can show each one right away. On the front end you read it two ways - the dead-simple EventSource (SSE) or the flexible fetch with a stream - and it's the second one that the LLM libraries use under the hood.
