How an LLM Learns and Builds a Reply
A visual explainer showing the difference between training an LLM and using a trained LLM. See how tokens move through transformer layers, how weights change during training, and why weights stay fixed while a reply is generated one token at a time.



