AI 네이티브 분업은 셋입니다.
두뇌는 기억합니다. 에이전트는 일합니다.
사람은 — 무엇을 시킬지 결정합니다.
세 역할은 서로 자리를 바꾸지 않습니다. 두뇌(/memory)는 누적된 컨텍스트의 거처입니다 — 어제 한 결정, 작년 운영했던 강의의 카피, 어제 추가한 카르파시 트윗까지 한 곳에 모입니다. 에이전트는 그 두뇌를 읽어 — 실제 작업을 수행합니다. 코드를 쓰고, 브라우저를 작동시키고, 영상의 썸네일을 만들고, 디스코드로 보고를 올립니다. 사람의 자리는 — 그 사이가 아닙니다. 무엇을 시킬지를 결정하는 자리입니다. 무엇을 만들지, 어떤 순서로 갈지, 어디서 멈출지를 정합니다.
이 분업이 자리잡으면 — 1인의 처리량이 팀의 처리량으로 바뀝니다. 한 사람이 동시에 다섯 개의 에이전트를 굴리고, 그 다섯 모두가 같은 두뇌를 참조합니다.Also in EnglishAI-native work splits three ways.
Brain remembers. Agents work.
The human decides what to set in motion.
The three roles don't swap. The brain (/memory) is where accumulated context lives — yesterday's decisions, last year's lecture copy, the Karpathy tweet filed an hour ago, all in one place. Agents read that brain and do the actual work: write code, drive the browser, generate thumbnails, post status reports to Discord. The human's seat is not in the middle of either — it's the seat of *what to set in motion*. What to build, in what order, where to stop.
Once this division holds, one person's throughput starts looking like a team's. One human runs five agents at once, and all five reference the same brain.