전문가는 LLM이 아닙니다.
옆에 둔 마크다운입니다.
Karpathy의 LLM Wiki, brain의 thoughts/ 트리, Nate Herk의 advertising_masterclass.md — 이름만 다를 뿐 같은 아키텍처입니다. 큐레이팅된 마크다운이 옆에 있어야 모델이 비로소 도메인 전문가로 동작합니다.
LLM은 무엇이든 압니다. 그래서 아무것도 모릅니다. 도메인은 모델이 아니라 마크다운이 결정합니다.Also in EnglishThe expert isn't the LLM. It's the markdown sitting next to it.
Karpathy's LLM Wiki, this brain's thoughts/ tree, Nate Herk's advertising_masterclass.md — three names for one architecture. A curated markdown sitting next to the model is what turns it into a domain specialist.
An LLM knows everything. Which is why it knows nothing in particular. The domain is decided by the markdown, not the model.