Post

This thread is a useful explainer on the difference between autocomplete and LLMs. Though I still think the analogy is helpful for dispelling the magical "it's thinking!" response. Read all the way through for an amazing demo of a heat-map tool showing LLM "attention" processes in action