// Next-token prediction
How an LLM picks a word
Context tokens
Probability of the next token
Append · repeat · one token at a time