빈 프롬프트에서 시작하지 않습니다 —
이미 쌓아둔 맥락으로 다음 맥락과 프롬프트를 만들어, 복리처럼 불립니다.
Don't start from a blank prompt — use the context you've already built to generate the next context and prompt, compounding it like interest.
ai-toolingcontext-engineeringcontext-beats-prompt
빈 프롬프트에서 시작하지 않습니다 —
이미 쌓아둔 맥락으로 다음 맥락과 프롬프트를 만들어, 복리처럼 불립니다.
새 리서치를 시작할 때 빈 화면에 프롬프트를 쓰지 않습니다. 이미 프로젝트 폴더와 CLAUDE.md에 맥락을 쌓아 뒀고, 에이전트가 그 폴더를 다 읽을 수 있으니, "너는 모먼트 리서치 디렉터다 — 우리가 쌓아온 자료를 토대로 NotebookLM에 넣을 리서치 질문 8개를 만들어줘"라고 시킵니다. 맥락을 만들 때조차 맥락을 씁니다. 그래서 아무것도 없는 백지에서 시작하는 일이 거의 없고, 한 번 쌓은 맥락이 다음 작업의 출발선을 끌어올리며 복리로 불어납니다.Also in EnglishDon't start from a blank prompt — use the context you've already built to generate the next context and prompt, compounding it like interest.
When you start new research, you don't type a prompt into a blank screen. You've already accumulated context in the project folder and CLAUDE.md, and the agent can read all of it — so you say: "You're Moment's research director; from the material we've built, write me eight research questions to feed NotebookLM." You spend context even to make context. That way you almost never start from a blank slate, and the context you built once raises the starting line for the next task — compounding like interest.