Nor is the LLM output tightly coupled to the training data. They'll "eagerly"[1] fill in the blanks wherever it sounds good. [1] here I don't mean to imply age

Nor is the LLM output tightly coupled to the training data. They'll "eagerly"[1] fill in the blanks wherever it sounds good. [1] here I don't mean to imply agency, just vigor.