LLMs Are Still Just Predicting the Next Word. So Why Are They So Good Now?

by tudorhn | View on Hacker News