firehose> #llmops

Curatorial-Voice Learning

When choosing what to learn from, who produced it matters more than what it is about — pick the individual educator or author whose exposition you resonate with, over the resource that merely matches a pre-existing interest. Grant Sanderson frames this as a pre-LLM insight: advise a college student to weight the quality of the teacher over the topic of the course ("your pre-existing interests are kind of arbitrary right now"), and advise a reader that "who the author is maybe matters more than if it's a prior interest" — so if you liked one book, read what else that author wrote rather than another book on the same subject. The unit of trust is a person's voice, not a subject tag.

For the vault this is the selection half of a pair: it says how to choose the artifact, while Motivated Exposition says what makes the chosen artifact good and LLM as Resource Router says how to use a model to find that person. It is also the human-taste analogue of firehose's personal-dedup rule — relevance is scored against who resonates with this operator, not a global notion of the "best" resource.

Claims


Linked from