{"success":true,"count":4,"items":[{"videoId":"ib74sLgjIBM","chunkIndex":0,"totalChunks":4,"title":"I Built Karpathy's AI Knowledge Base in Claude: Try it! — Part 1 of 4","thumbnail":"https://i.ytimg.com/vi/ib74sLgjIBM/maxresdefault.jpg","duration":2175,"uploader":"Systems Made Better","youtubeUrl":"https://www.youtube.com/watch?v=ib74sLgjIBM","keywords":["ai","knowledge-management","second-brain","productivity","personal-knowledge-base","workflow-automation","claude","note-taking","self-improvement"],"normalizedKeywords":["프로덕트","비즈니스·전략","커리어·성장"],"targetAudience":[{"who":"지식노동자","why":"정보를 쌓고 다시 활용하는 개인 지식관리 시스템을 바로 만들 수 있음"},{"who":"창업자","why":"팀의 판단과 맥락을 축적하는 내부 지식 자산 설계에 도움이 됨"},{"who":"프로덕트 매니저","why":"문서·메모·피드백을 구조화해 의사결정 품질을 높이는 방법을 배울 수 있음"},{"who":"학생","why":"공부한 내용을 정리·연결·복습하는 AI 학습 시스템 아이디어를 얻을 수 있음"}],"normalizedAudience":["지식노동자 일반","창업자·스타트업","프로덕트 매니저·기획자"],"summary":"이 영상은 Claude만으로 '자기개선형 개인 지식베이스'를 만드는 방법을 소개한다. 핵심 아이디어는 raw, wiki, outputs라는 세 폴더와 루트의 Claude.md 파일만으로도, 흩어진 자료를 쌓고 AI가 위키로 정리하며, 질문 답변을 다시 저장해 지식이 누적되는 순환 구조를 만든다는 것이다. 화자는 앤드루 카파시의 개인 지식베이스 사례를 참고해, 복잡한 벡터DB나 Obsidian 없이도 충분히 강력하게 작동한다고 주장한다.\n\n이후 실제로 Claude에 지시를 넣어 폴더 구조를 만들고, 상위 knowledge base와 하위 주제별 knowledge base를 분리하는 설계, 건강 점검과 중복·모순 탐지 같은 유지보수 규칙까지 붙인다. 영상의 메시지는 단순하다: 지식을 모으는 것보다 중요한 것은 AI가 정리·질문·개선을 반복하게 만드는 구조이며, 시간이 지날수록 그 시스템은 개인의 판단과 관점을 담는 자산이 된다는 점이다.","insights":["지식베이스의 가치는 저장보다 재사용 구조에 있다.","raw-wiki-outputs 분리는 단순하지만 강력한 설계다.","AI가 위키를 쓰고 질문 답변을 다시 넣어야 축적된다.","복잡한 도구보다 유지 가능한 폴더 구조가 더 중요하다.","건강 점검과 모순 탐지가 지식베이스를 살아 있게 한다."],"keyClips":[{"clipId":"ib74sLgjIBM:c0:1-13","startSegmentIndex":1,"endSegmentIndex":13,"startTime":2.55,"endTime":94.479,"durationSeconds":91.9,"preview":"왜 이 구조가 강한가","mustSee":true},{"clipId":"ib74sLgjIBM:c0:18-30","startSegmentIndex":18,"endSegmentIndex":30,"startTime":122.31,"endTime":193.12,"durationSeconds":70.8,"preview":"세 폴더 한 파일","mustSee":true},{"clipId":"ib74sLgjIBM:c0:31-37","startSegmentIndex":31,"endSegmentIndex":37,"startTime":193.12,"endTime":242.319,"durationSeconds":49.2,"preview":"카파시 사례의 힘","mustSee":false},{"clipId":"ib74sLgjIBM:c0:44-57","startSegmentIndex":44,"endSegmentIndex":57,"startTime":269.6,"endTime":343.68,"durationSeconds":74.1,"preview":"실제 구축 과정","mustSee":false},{"clipId":"ib74sLgjIBM:c0:60-88","startSegmentIndex":60,"endSegmentIndex":88,"startTime":357.12,"endTime":600.8,"durationSeconds":243.7,"preview":"지식베이스의 운영원리","mustSee":true}],"curatedSegments":[{"segmentIndex":14,"text":"Now, I found that day one of running it, your knowledge base is pretty basic, but day 100, it's a company asset that nobody else has.","startTime":94.479,"endTime":101.92,"durationSeconds":7,"level":"B2","overallScore":7.6,"rationale":"시간에 따른 자산화 통찰이 큼."},{"segmentIndex":77,"text":"Flag contradictions between articles, find topics mentioned, list claims not backed by a source, etc. Please write your proposed top level second brain MD and then knowledgebased MD for the productivity knowledge base example we're building.","startTime":484.879,"endTime":500.639,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"검증 기준과 작업 지시가 매우 구체적."},{"segmentIndex":86,"text":"I'm going to change that. This knowledge base is focused on the ethos of doing less but better, finding a balanced approach to a greater contribution to the world, deeper thinking, and stronger output whilst managing health, happiness, and balance in your life.","startTime":556.48,"endTime":575.92,"durationSeconds":19,"level":"C1","overallScore":8,"rationale":"생산성 철학이 잘 드러난 통찰 문장."},{"segmentIndex":16,"text":"This is the simplest way, but this is the point about a second brain as well. We find something brilliant, we save it, and then we lose it.","startTime":710.48,"endTime":716,"durationSeconds":6,"level":"B1","overallScore":7.6,"rationale":"문제 핵심을 선명하게 짚음."},{"segmentIndex":35,"text":"The point is it's a folder for capture. The organization is the AI's job.","startTime":820.88,"endTime":825.92,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"역할 분담 원칙을 명확히 설명."},{"segmentIndex":60,"text":"And then you basically walk away and let it do the job. What you come back to is information organized topic pages with summaries, connections between ideas you didn't know existed, an index that makes everything searchable in second.","startTime":985.519,"endTime":999.199,"durationSeconds":14,"level":"B2","overallScore":7.8,"rationale":"결과 가치와 효용을 잘 요약함."},{"segmentIndex":64,"text":"What I think Kapathy has figured out with this approach using LLMs is that the AI becomes the librarian.","startTime":1016.079,"endTime":1022.8,"durationSeconds":7,"level":"B2","overallScore":8,"rationale":"핵심 관점 전환을 담은 문장."},{"segmentIndex":66,"text":"And by the end, it's learning and improving on its own, helping you actually apply the knowledge to output.","startTime":1028,"endTime":1035.36,"durationSeconds":7,"level":"B2","overallScore":7.8,"rationale":"성과와 활용 방향을 제시하는 문장."},{"segmentIndex":8,"text":"So now we need to ask it questions and get things out of it. And it's this process that actually changes everything because every time you ask the agent a question you like the answer to, you can then save that back into raw or into the wiki and the system gets smarter the more you use it.","startTime":1242.95,"endTime":1257.2,"durationSeconds":14,"level":"B2","overallScore":8,"rationale":"학습 원리 설명이 선명하고 표현도 풍부."},{"segmentIndex":15,"text":"You can't win both simultaneously. Trying to is what produces burnout.","startTime":1300.559,"endTime":1304.799,"durationSeconds":4,"level":"B2","overallScore":7.6,"rationale":"번아웃 원인을 압축해 말한 통찰 문장."},{"segmentIndex":41,"text":"You'll save it back and the next answer quietly builds on a mistake.","startTime":1452.4,"endTime":1457.84,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"오류 누적 메커니즘을 잘 보여주는 문장."},{"segmentIndex":50,"text":"Every meeting transcript that mattered, every answer you've saved back into the system becomes a carefully cured, cross-referenced, linked, and summarized set of information that you can query with the librarian themselves.","startTime":2110.64,"endTime":2124.56,"durationSeconds":14,"level":"C1","overallScore":8,"rationale":"지식자산의 축적 가치를 강하게 설명함."},{"segmentIndex":51,"text":"And it's that kind of asset that's nearly impossible to replicate because nobody else has read what you've read or saved what you've saved.","startTime":2124.56,"endTime":2131.44,"durationSeconds":7,"level":"C1","overallScore":8,"rationale":"개인화 자산의 희소성을 설득력 있게 말함."},{"segmentIndex":4,"text":"And that's the problem. This is genuinely the most useful AI setup I've seen in months and implemented in Claude.","startTime":28.88,"endTime":36.399,"durationSeconds":8,"level":"B2","overallScore":6.6,"rationale":"가치 판단이 분명한 핵심 주장."},{"segmentIndex":12,"text":"And with a health check and that loop, it just keeps improving over time.","startTime":79.759,"endTime":83.68,"durationSeconds":4,"level":"B2","overallScore":6.6,"rationale":"자기개선 원리를 간결히 제시함."},{"segmentIndex":13,"text":"So, by the end of this video, you'll know exactly what the system is, why it beats every Obsidian Plus plug-in setup you can find for simplicity, and how to build your own right now with Claude.","startTime":83.68,"endTime":94.479,"durationSeconds":11,"level":"B2","overallScore":7.2,"rationale":"가치·비교·결과를 선명히 제시함."},{"segmentIndex":26,"text":"Now, the best bit is those then get fed back in and help to refine it, plus one file at the root.","startTime":165.84,"endTime":171.28,"durationSeconds":5,"level":"B2","overallScore":6.8,"rationale":"피드백 루프의 핵심을 보여줌."},{"segmentIndex":29,"text":"No database, no obsidian, no vault setup, just folders and text files on your computer.","startTime":181.92,"endTime":187.2,"durationSeconds":5,"level":"B2","overallScore":7,"rationale":"단순성의 강점을 선명히 강조함."},{"segmentIndex":33,"text":"If it works for one of the most respected AI researchers alive, it'll probably work for your business.","startTime":202.959,"endTime":208.72,"durationSeconds":6,"level":"B2","overallScore":6.8,"rationale":"권위 기반 일반화가 설득력 있음."},{"segmentIndex":34,"text":"The best thing like what I've done in my notion agent OS is you can then point a custom agent at that knowledge base and it becomes a specialist agent expert that you can speak with.","startTime":208.72,"endTime":219.12,"durationSeconds":10,"level":"C1","overallScore":7.2,"rationale":"응용 가능성과 표현 다양성이 큼."}],"generatedAt":"2026-06-24T23:43:10.982Z","keyClipsTotalSec":696},{"videoId":"ib74sLgjIBM","chunkIndex":1,"totalChunks":4,"title":"I Built Karpathy's AI Knowledge Base in Claude: Try it! — Part 2 of 4","thumbnail":"https://i.ytimg.com/vi/ib74sLgjIBM/maxresdefault.jpg","duration":2175,"uploader":"Systems Made Better","youtubeUrl":"https://www.youtube.com/watch?v=ib74sLgjIBM","keywords":["knowledge-management","second-brain","ai-workflow","claude","notion","obsidian","productivity","markdown","automation","personal-knowledge-base"],"normalizedKeywords":["프로덕트","교육","기술 트렌드"],"targetAudience":[{"who":"지식노동자","why":"노트와 자료를 쌓기만 하고 재활용하지 못하는 문제를 해결하는 데 유용함"},{"who":"창업자","why":"팀이나 개인의 지식 자산을 AI로 자동 정리하는 운영 방식에 도움됨"},{"who":"크리에이터","why":"콘텐츠·자료·레퍼런스를 빠르게 축적하고 재구성하는 워크플로를 배울 수 있음"}],"normalizedAudience":["지식노동자 일반","창업자·스타트업","크리에이터·작가"],"summary":"이 영상은 Claude를 활용해 개인 지식베이스를 자동으로 구축하고, RAW에 자료를 덤프한 뒤 wiki를 생성하는 워크플로를 실제로 시연한다. 핵심은 사람이 직접 정리자(librarian)가 되는 대신, AI가 분류·링크·요약·인덱싱을 맡게 만들어 '두 번째 뇌'가 실제로 작동하게 하는 데 있다.\n\n발화자는 Notion, Obsidian, Xcode, markdown 파일, 웹 클리퍼, PDF/이미지 첨부 같은 실무 도구를 섞어 구체적인 구축 방법을 설명한다. 또한 AI 글쓰기 스타일 가이드를 넣어 산출물의 품질을 관리하고, 이런 자동화가 개인 생산성뿐 아니라 팀·비즈니스의 지식 운영에도 확장 가능하다고 주장한다.","insights":["두 번째 뇌의 핵심은 저장이 아니라 재사용이다.","정리 부담을 AI에 넘길 때 지식베이스가 살아난다.","RAW는 예쁘게 꾸미는 곳이 아니라 무조건 쌓는 곳이다.","문서형 지식은 markdown으로 갈수록 자동화와 궁합이 좋다.","AI에게 스타일 규칙을 주면 내용만이 아니라 품질도 관리된다."],"keyClips":[{"clipId":"ib74sLgjIBM:c1:1-9","startSegmentIndex":1,"endSegmentIndex":9,"startTime":602.07,"endTime":668.079,"durationSeconds":66,"preview":"자동화 메모리 구조","mustSee":false},{"clipId":"ib74sLgjIBM:c1:10-17","startSegmentIndex":10,"endSegmentIndex":17,"startTime":668.079,"endTime":720.079,"durationSeconds":52,"preview":"두 번째 뇌의 함정","mustSee":false},{"clipId":"ib74sLgjIBM:c1:18-31","startSegmentIndex":18,"endSegmentIndex":31,"startTime":720.079,"endTime":804.8,"durationSeconds":84.7,"preview":"노션을 붙이는 법","mustSee":false},{"clipId":"ib74sLgjIBM:c1:32-41","startSegmentIndex":32,"endSegmentIndex":41,"startTime":804.8,"endTime":861.04,"durationSeconds":56.2,"preview":"RAW는 덤프 공간","mustSee":false},{"clipId":"ib74sLgjIBM:c1:42-54","startSegmentIndex":42,"endSegmentIndex":54,"startTime":861.04,"endTime":948.56,"durationSeconds":87.5,"preview":"마크다운 실무 팁","mustSee":false},{"clipId":"ib74sLgjIBM:c1:57-68","startSegmentIndex":57,"endSegmentIndex":68,"startTime":958.72,"endTime":1054.08,"durationSeconds":95.4,"preview":"위키 생성의 가치","mustSee":true},{"clipId":"ib74sLgjIBM:c1:69-88","startSegmentIndex":69,"endSegmentIndex":88,"startTime":1054.08,"endTime":1198.96,"durationSeconds":144.9,"preview":"스타일과 확장성","mustSee":false}],"curatedSegments":[{"segmentIndex":14,"text":"Now, I found that day one of running it, your knowledge base is pretty basic, but day 100, it's a company asset that nobody else has.","startTime":94.479,"endTime":101.92,"durationSeconds":7,"level":"B2","overallScore":7.6,"rationale":"시간에 따른 자산화 통찰이 큼."},{"segmentIndex":77,"text":"Flag contradictions between articles, find topics mentioned, list claims not backed by a source, etc. Please write your proposed top level second brain MD and then knowledgebased MD for the productivity knowledge base example we're building.","startTime":484.879,"endTime":500.639,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"검증 기준과 작업 지시가 매우 구체적."},{"segmentIndex":86,"text":"I'm going to change that. This knowledge base is focused on the ethos of doing less but better, finding a balanced approach to a greater contribution to the world, deeper thinking, and stronger output whilst managing health, happiness, and balance in your life.","startTime":556.48,"endTime":575.92,"durationSeconds":19,"level":"C1","overallScore":8,"rationale":"생산성 철학이 잘 드러난 통찰 문장."},{"segmentIndex":16,"text":"This is the simplest way, but this is the point about a second brain as well. We find something brilliant, we save it, and then we lose it.","startTime":710.48,"endTime":716,"durationSeconds":6,"level":"B1","overallScore":7.6,"rationale":"문제 핵심을 선명하게 짚음."},{"segmentIndex":35,"text":"The point is it's a folder for capture. The organization is the AI's job.","startTime":820.88,"endTime":825.92,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"역할 분담 원칙을 명확히 설명."},{"segmentIndex":60,"text":"And then you basically walk away and let it do the job. What you come back to is information organized topic pages with summaries, connections between ideas you didn't know existed, an index that makes everything searchable in second.","startTime":985.519,"endTime":999.199,"durationSeconds":14,"level":"B2","overallScore":7.8,"rationale":"결과 가치와 효용을 잘 요약함."},{"segmentIndex":64,"text":"What I think Kapathy has figured out with this approach using LLMs is that the AI becomes the librarian.","startTime":1016.079,"endTime":1022.8,"durationSeconds":7,"level":"B2","overallScore":8,"rationale":"핵심 관점 전환을 담은 문장."},{"segmentIndex":66,"text":"And by the end, it's learning and improving on its own, helping you actually apply the knowledge to output.","startTime":1028,"endTime":1035.36,"durationSeconds":7,"level":"B2","overallScore":7.8,"rationale":"성과와 활용 방향을 제시하는 문장."},{"segmentIndex":8,"text":"So now we need to ask it questions and get things out of it. And it's this process that actually changes everything because every time you ask the agent a question you like the answer to, you can then save that back into raw or into the wiki and the system gets smarter the more you use it.","startTime":1242.95,"endTime":1257.2,"durationSeconds":14,"level":"B2","overallScore":8,"rationale":"학습 원리 설명이 선명하고 표현도 풍부."},{"segmentIndex":15,"text":"You can't win both simultaneously. Trying to is what produces burnout.","startTime":1300.559,"endTime":1304.799,"durationSeconds":4,"level":"B2","overallScore":7.6,"rationale":"번아웃 원인을 압축해 말한 통찰 문장."},{"segmentIndex":41,"text":"You'll save it back and the next answer quietly builds on a mistake.","startTime":1452.4,"endTime":1457.84,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"오류 누적 메커니즘을 잘 보여주는 문장."},{"segmentIndex":50,"text":"Every meeting transcript that mattered, every answer you've saved back into the system becomes a carefully cured, cross-referenced, linked, and summarized set of information that you can query with the librarian themselves.","startTime":2110.64,"endTime":2124.56,"durationSeconds":14,"level":"C1","overallScore":8,"rationale":"지식자산의 축적 가치를 강하게 설명함."},{"segmentIndex":51,"text":"And it's that kind of asset that's nearly impossible to replicate because nobody else has read what you've read or saved what you've saved.","startTime":2124.56,"endTime":2131.44,"durationSeconds":7,"level":"C1","overallScore":8,"rationale":"개인화 자산의 희소성을 설득력 있게 말함."},{"segmentIndex":4,"text":"And that's the problem. This is genuinely the most useful AI setup I've seen in months and implemented in Claude.","startTime":28.88,"endTime":36.399,"durationSeconds":8,"level":"B2","overallScore":6.6,"rationale":"가치 판단이 분명한 핵심 주장."},{"segmentIndex":12,"text":"And with a health check and that loop, it just keeps improving over time.","startTime":79.759,"endTime":83.68,"durationSeconds":4,"level":"B2","overallScore":6.6,"rationale":"자기개선 원리를 간결히 제시함."},{"segmentIndex":13,"text":"So, by the end of this video, you'll know exactly what the system is, why it beats every Obsidian Plus plug-in setup you can find for simplicity, and how to build your own right now with Claude.","startTime":83.68,"endTime":94.479,"durationSeconds":11,"level":"B2","overallScore":7.2,"rationale":"가치·비교·결과를 선명히 제시함."},{"segmentIndex":26,"text":"Now, the best bit is those then get fed back in and help to refine it, plus one file at the root.","startTime":165.84,"endTime":171.28,"durationSeconds":5,"level":"B2","overallScore":6.8,"rationale":"피드백 루프의 핵심을 보여줌."},{"segmentIndex":29,"text":"No database, no obsidian, no vault setup, just folders and text files on your computer.","startTime":181.92,"endTime":187.2,"durationSeconds":5,"level":"B2","overallScore":7,"rationale":"단순성의 강점을 선명히 강조함."},{"segmentIndex":33,"text":"If it works for one of the most respected AI researchers alive, it'll probably work for your business.","startTime":202.959,"endTime":208.72,"durationSeconds":6,"level":"B2","overallScore":6.8,"rationale":"권위 기반 일반화가 설득력 있음."},{"segmentIndex":34,"text":"The best thing like what I've done in my notion agent OS is you can then point a custom agent at that knowledge base and it becomes a specialist agent expert that you can speak with.","startTime":208.72,"endTime":219.12,"durationSeconds":10,"level":"C1","overallScore":7.2,"rationale":"응용 가능성과 표현 다양성이 큼."}],"generatedAt":"2026-06-24T23:43:41.045Z","keyClipsTotalSec":696},{"videoId":"ib74sLgjIBM","chunkIndex":2,"totalChunks":4,"title":"I Built Karpathy's AI Knowledge Base in Claude: Try it! — Part 3 of 4","thumbnail":"https://i.ytimg.com/vi/ib74sLgjIBM/maxresdefault.jpg","duration":2175,"uploader":"Systems Made Better","youtubeUrl":"https://www.youtube.com/watch?v=ib74sLgjIBM","keywords":["ai-workflow","knowledge-base","claude","automation","productivity","prompt-engineering","information-management","quality-control","skill-building"],"normalizedKeywords":["프로덕트","엔지니어링","커리어·성장"],"targetAudience":[{"who":"지식노동자","why":"AI로 개인 지식베이스를 만들고 반복적으로 개선하는 방법을 배울 수 있음"},{"who":"콘텐츠 제작자","why":"자료를 구조화해 검색·재활용 가능한 지식 시스템으로 바꾸는 감각을 얻을 수 있음"},{"who":"업무 자동화 관심자","why":"보고서 생성, 정기 점검, 스케줄링까지 자동화하는 흐름을 볼 수 있음"}],"normalizedAudience":["지식노동자 일반","크리에이터·작가","엔지니어·개발자"],"summary":"이 영상은 Claude로 개인용 AI 지식베이스를 만들고, 그 지식베이스가 질문에 답하게 한 뒤, 그 답변을 다시 raw/wiki/outputs에 반영해 점점 똑똑해지게 만드는 과정을 보여준다. 처음에는 인덱스와 위키 구조가 잘 작동하는지 테스트하고, 이후에는 질문 응답 결과가 outputs에 자동으로 저장되지 않는 문제를 발견해 시스템 규칙을 수정한다.\n\n이후에는 지식베이스의 품질을 유지하기 위한 월간 health check를 설계한다. 위키의 모순, 누락, 출처 부족, 오래된 글, 끊긴 링크, 새 글 후보까지 점검하는 7단계 감사(audit)를 skill과 scheduled task로 자동화하며, AI가 스스로 개선되도록 만드는 운영 방식이 핵심 메시지다.","insights":["지식베이스의 가치는 '저장'보다 '재사용'에서 커진다.","질문-응답-저장 루프가 반복될수록 시스템은 더 똑똑해진다.","출력을 남기지 않으면 개선도 축적되지 않는다.","AI는 맞게 쓰는 것보다 주기적으로 점검하는 게 더 중요하다.","자동화는 편의 기능이 아니라 지식 품질 관리 장치다."],"keyClips":[{"clipId":"ib74sLgjIBM:c2:2-9","startSegmentIndex":2,"endSegmentIndex":9,"startTime":1204.799,"endTime":1263.6,"durationSeconds":58.8,"preview":"지식이 자라는 구조","mustSee":false},{"clipId":"ib74sLgjIBM:c2:11-19","startSegmentIndex":11,"endSegmentIndex":19,"startTime":1267.12,"endTime":1320,"durationSeconds":52.9,"preview":"질문으로 검증하기","mustSee":false},{"clipId":"ib74sLgjIBM:c2:20-30","startSegmentIndex":20,"endSegmentIndex":30,"startTime":1320,"endTime":1397.2,"durationSeconds":77.2,"preview":"출력까지 남겨라","mustSee":false},{"clipId":"ib74sLgjIBM:c2:31-37","startSegmentIndex":31,"endSegmentIndex":37,"startTime":1397.2,"endTime":1441.44,"durationSeconds":44.2,"preview":"빈틈을 찾아 개선","mustSee":false},{"clipId":"ib74sLgjIBM:c2:40-48","startSegmentIndex":40,"endSegmentIndex":48,"startTime":1448.799,"endTime":1510.799,"durationSeconds":62,"preview":"월간 품질 점검","mustSee":true},{"clipId":"ib74sLgjIBM:c2:54-68","startSegmentIndex":54,"endSegmentIndex":68,"startTime":1544.159,"endTime":1660.96,"durationSeconds":116.8,"preview":"스킬과 스케줄링","mustSee":false},{"clipId":"ib74sLgjIBM:c2:73-82","startSegmentIndex":73,"endSegmentIndex":82,"startTime":1677.84,"endTime":1730.24,"durationSeconds":52.4,"preview":"감사 결과 읽기","mustSee":false}],"curatedSegments":[{"segmentIndex":14,"text":"Now, I found that day one of running it, your knowledge base is pretty basic, but day 100, it's a company asset that nobody else has.","startTime":94.479,"endTime":101.92,"durationSeconds":7,"level":"B2","overallScore":7.6,"rationale":"시간에 따른 자산화 통찰이 큼."},{"segmentIndex":77,"text":"Flag contradictions between articles, find topics mentioned, list claims not backed by a source, etc. Please write your proposed top level second brain MD and then knowledgebased MD for the productivity knowledge base example we're building.","startTime":484.879,"endTime":500.639,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"검증 기준과 작업 지시가 매우 구체적."},{"segmentIndex":86,"text":"I'm going to change that. This knowledge base is focused on the ethos of doing less but better, finding a balanced approach to a greater contribution to the world, deeper thinking, and stronger output whilst managing health, happiness, and balance in your life.","startTime":556.48,"endTime":575.92,"durationSeconds":19,"level":"C1","overallScore":8,"rationale":"생산성 철학이 잘 드러난 통찰 문장."},{"segmentIndex":16,"text":"This is the simplest way, but this is the point about a second brain as well. We find something brilliant, we save it, and then we lose it.","startTime":710.48,"endTime":716,"durationSeconds":6,"level":"B1","overallScore":7.6,"rationale":"문제 핵심을 선명하게 짚음."},{"segmentIndex":35,"text":"The point is it's a folder for capture. The organization is the AI's job.","startTime":820.88,"endTime":825.92,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"역할 분담 원칙을 명확히 설명."},{"segmentIndex":60,"text":"And then you basically walk away and let it do the job. What you come back to is information organized topic pages with summaries, connections between ideas you didn't know existed, an index that makes everything searchable in second.","startTime":985.519,"endTime":999.199,"durationSeconds":14,"level":"B2","overallScore":7.8,"rationale":"결과 가치와 효용을 잘 요약함."},{"segmentIndex":64,"text":"What I think Kapathy has figured out with this approach using LLMs is that the AI becomes the librarian.","startTime":1016.079,"endTime":1022.8,"durationSeconds":7,"level":"B2","overallScore":8,"rationale":"핵심 관점 전환을 담은 문장."},{"segmentIndex":66,"text":"And by the end, it's learning and improving on its own, helping you actually apply the knowledge to output.","startTime":1028,"endTime":1035.36,"durationSeconds":7,"level":"B2","overallScore":7.8,"rationale":"성과와 활용 방향을 제시하는 문장."},{"segmentIndex":8,"text":"So now we need to ask it questions and get things out of it. And it's this process that actually changes everything because every time you ask the agent a question you like the answer to, you can then save that back into raw or into the wiki and the system gets smarter the more you use it.","startTime":1242.95,"endTime":1257.2,"durationSeconds":14,"level":"B2","overallScore":8,"rationale":"학습 원리 설명이 선명하고 표현도 풍부."},{"segmentIndex":15,"text":"You can't win both simultaneously. Trying to is what produces burnout.","startTime":1300.559,"endTime":1304.799,"durationSeconds":4,"level":"B2","overallScore":7.6,"rationale":"번아웃 원인을 압축해 말한 통찰 문장."},{"segmentIndex":41,"text":"You'll save it back and the next answer quietly builds on a mistake.","startTime":1452.4,"endTime":1457.84,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"오류 누적 메커니즘을 잘 보여주는 문장."},{"segmentIndex":50,"text":"Every meeting transcript that mattered, every answer you've saved back into the system becomes a carefully cured, cross-referenced, linked, and summarized set of information that you can query with the librarian themselves.","startTime":2110.64,"endTime":2124.56,"durationSeconds":14,"level":"C1","overallScore":8,"rationale":"지식자산의 축적 가치를 강하게 설명함."},{"segmentIndex":51,"text":"And it's that kind of asset that's nearly impossible to replicate because nobody else has read what you've read or saved what you've saved.","startTime":2124.56,"endTime":2131.44,"durationSeconds":7,"level":"C1","overallScore":8,"rationale":"개인화 자산의 희소성을 설득력 있게 말함."},{"segmentIndex":4,"text":"And that's the problem. This is genuinely the most useful AI setup I've seen in months and implemented in Claude.","startTime":28.88,"endTime":36.399,"durationSeconds":8,"level":"B2","overallScore":6.6,"rationale":"가치 판단이 분명한 핵심 주장."},{"segmentIndex":12,"text":"And with a health check and that loop, it just keeps improving over time.","startTime":79.759,"endTime":83.68,"durationSeconds":4,"level":"B2","overallScore":6.6,"rationale":"자기개선 원리를 간결히 제시함."},{"segmentIndex":13,"text":"So, by the end of this video, you'll know exactly what the system is, why it beats every Obsidian Plus plug-in setup you can find for simplicity, and how to build your own right now with Claude.","startTime":83.68,"endTime":94.479,"durationSeconds":11,"level":"B2","overallScore":7.2,"rationale":"가치·비교·결과를 선명히 제시함."},{"segmentIndex":26,"text":"Now, the best bit is those then get fed back in and help to refine it, plus one file at the root.","startTime":165.84,"endTime":171.28,"durationSeconds":5,"level":"B2","overallScore":6.8,"rationale":"피드백 루프의 핵심을 보여줌."},{"segmentIndex":29,"text":"No database, no obsidian, no vault setup, just folders and text files on your computer.","startTime":181.92,"endTime":187.2,"durationSeconds":5,"level":"B2","overallScore":7,"rationale":"단순성의 강점을 선명히 강조함."},{"segmentIndex":33,"text":"If it works for one of the most respected AI researchers alive, it'll probably work for your business.","startTime":202.959,"endTime":208.72,"durationSeconds":6,"level":"B2","overallScore":6.8,"rationale":"권위 기반 일반화가 설득력 있음."},{"segmentIndex":34,"text":"The best thing like what I've done in my notion agent OS is you can then point a custom agent at that knowledge base and it becomes a specialist agent expert that you can speak with.","startTime":208.72,"endTime":219.12,"durationSeconds":10,"level":"C1","overallScore":7.2,"rationale":"응용 가능성과 표현 다양성이 큼."}],"generatedAt":"2026-06-24T23:44:00.634Z","keyClipsTotalSec":696},{"videoId":"ib74sLgjIBM","chunkIndex":3,"totalChunks":4,"title":"I Built Karpathy's AI Knowledge Base in Claude: Try it! — Part 4 of 4","thumbnail":"https://i.ytimg.com/vi/ib74sLgjIBM/maxresdefault.jpg","duration":2175,"uploader":"Systems Made Better","youtubeUrl":"https://www.youtube.com/watch?v=ib74sLgjIBM","keywords":["claude","ai-workflow","knowledge-base","automation","productivity","personal-knowledge-management","prompt-engineering","information-retrieval","note-taking"],"normalizedKeywords":["프로덕트","커리어·성장","교육"],"targetAudience":[{"who":"지식노동자","why":"자신만의 자료와 회의 기록을 AI로 축적·검색하려는 사람에게 유용함"},{"who":"콘텐츠 제작자","why":"학습한 내용을 체계화해 재사용 가능한 지식 자산으로 만들 수 있음"},{"who":"AI 활용자","why":"Claude로 반복 업무를 자동화하고 지식베이스를 운영하는 방법을 배울 수 있음"}],"normalizedAudience":["지식노동자 일반","크리에이터·작가"],"summary":"이 영상은 Claude를 이용해 개인 지식베이스를 만들고, 이를 건강검진·문서 생성·인덱싱 업데이트까지 자동으로 굴리는 워크플로를 시연한다. 단순히 자료를 저장하는 수준이 아니라, 정기적으로 보고서를 생성하고 액션 리스트를 실행해 지식베이스가 스스로 정비되도록 만드는 것이 핵심이다.\n\n화자는 이런 시스템이 초반에는 별것 없어 보여도 시간이 쌓일수록 강력한 자산이 된다고 강조한다. 회의록, 답변, 기사 초안 같은 조각들이 서로 연결·요약·검색되면서, 남들이 쉽게 복제할 수 없는 개인용 지식 인프라가 된다는 점을 설득한다.","insights":["지식베이스의 가치는 축적과 연결이 쌓일수록 커진다.","자동화는 편의 기능이 아니라 지식 자산을 키우는 엔진이다.","초기엔 단순하게, 검증 후엔 전자동으로 옮기는 게 효율적이다.","비용이 드는 자동화는 빈도가 아니라 임팩트 기준으로 설계해야 한다.","반복해서 쌓인 개인 자료는 남이 복제하기 어려운 경쟁력이 된다."],"keyClips":[{"clipId":"ib74sLgjIBM:c3:1-8","startSegmentIndex":1,"endSegmentIndex":8,"startTime":1800.63,"endTime":1842.64,"durationSeconds":42,"preview":"자동 점검의 시작","mustSee":false},{"clipId":"ib74sLgjIBM:c3:11-20","startSegmentIndex":11,"endSegmentIndex":20,"startTime":1851.36,"endTime":1903.519,"durationSeconds":52.2,"preview":"건강검진이 핵심","mustSee":false},{"clipId":"ib74sLgjIBM:c3:25-34","startSegmentIndex":25,"endSegmentIndex":34,"startTime":1921.679,"endTime":2001.36,"durationSeconds":79.7,"preview":"반자동에서 전자동","mustSee":true},{"clipId":"ib74sLgjIBM:c3:37-46","startSegmentIndex":37,"endSegmentIndex":46,"startTime":2025.84,"endTime":2097.99,"durationSeconds":72.1,"preview":"질문이 자산을 만든다","mustSee":false},{"clipId":"ib74sLgjIBM:c3:47-54","startSegmentIndex":47,"endSegmentIndex":54,"startTime":2097.99,"endTime":2150.88,"durationSeconds":52.9,"preview":"100일 뒤의 가치","mustSee":true}],"curatedSegments":[{"segmentIndex":14,"text":"Now, I found that day one of running it, your knowledge base is pretty basic, but day 100, it's a company asset that nobody else has.","startTime":94.479,"endTime":101.92,"durationSeconds":7,"level":"B2","overallScore":7.6,"rationale":"시간에 따른 자산화 통찰이 큼."},{"segmentIndex":77,"text":"Flag contradictions between articles, find topics mentioned, list claims not backed by a source, etc. Please write your proposed top level second brain MD and then knowledgebased MD for the productivity knowledge base example we're building.","startTime":484.879,"endTime":500.639,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"검증 기준과 작업 지시가 매우 구체적."},{"segmentIndex":86,"text":"I'm going to change that. This knowledge base is focused on the ethos of doing less but better, finding a balanced approach to a greater contribution to the world, deeper thinking, and stronger output whilst managing health, happiness, and balance in your life.","startTime":556.48,"endTime":575.92,"durationSeconds":19,"level":"C1","overallScore":8,"rationale":"생산성 철학이 잘 드러난 통찰 문장."},{"segmentIndex":16,"text":"This is the simplest way, but this is the point about a second brain as well. We find something brilliant, we save it, and then we lose it.","startTime":710.48,"endTime":716,"durationSeconds":6,"level":"B1","overallScore":7.6,"rationale":"문제 핵심을 선명하게 짚음."},{"segmentIndex":35,"text":"The point is it's a folder for capture. The organization is the AI's job.","startTime":820.88,"endTime":825.92,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"역할 분담 원칙을 명확히 설명."},{"segmentIndex":60,"text":"And then you basically walk away and let it do the job. What you come back to is information organized topic pages with summaries, connections between ideas you didn't know existed, an index that makes everything searchable in second.","startTime":985.519,"endTime":999.199,"durationSeconds":14,"level":"B2","overallScore":7.8,"rationale":"결과 가치와 효용을 잘 요약함."},{"segmentIndex":64,"text":"What I think Kapathy has figured out with this approach using LLMs is that the AI becomes the librarian.","startTime":1016.079,"endTime":1022.8,"durationSeconds":7,"level":"B2","overallScore":8,"rationale":"핵심 관점 전환을 담은 문장."},{"segmentIndex":66,"text":"And by the end, it's learning and improving on its own, helping you actually apply the knowledge to output.","startTime":1028,"endTime":1035.36,"durationSeconds":7,"level":"B2","overallScore":7.8,"rationale":"성과와 활용 방향을 제시하는 문장."},{"segmentIndex":8,"text":"So now we need to ask it questions and get things out of it. And it's this process that actually changes everything because every time you ask the agent a question you like the answer to, you can then save that back into raw or into the wiki and the system gets smarter the more you use it.","startTime":1242.95,"endTime":1257.2,"durationSeconds":14,"level":"B2","overallScore":8,"rationale":"학습 원리 설명이 선명하고 표현도 풍부."},{"segmentIndex":15,"text":"You can't win both simultaneously. Trying to is what produces burnout.","startTime":1300.559,"endTime":1304.799,"durationSeconds":4,"level":"B2","overallScore":7.6,"rationale":"번아웃 원인을 압축해 말한 통찰 문장."},{"segmentIndex":41,"text":"You'll save it back and the next answer quietly builds on a mistake.","startTime":1452.4,"endTime":1457.84,"durationSeconds":5,"level":"B2","overallScore":7.8,"rationale":"오류 누적 메커니즘을 잘 보여주는 문장."},{"segmentIndex":50,"text":"Every meeting transcript that mattered, every answer you've saved back into the system becomes a carefully cured, cross-referenced, linked, and summarized set of information that you can query with the librarian themselves.","startTime":2110.64,"endTime":2124.56,"durationSeconds":14,"level":"C1","overallScore":8,"rationale":"지식자산의 축적 가치를 강하게 설명함."},{"segmentIndex":51,"text":"And it's that kind of asset that's nearly impossible to replicate because nobody else has read what you've read or saved what you've saved.","startTime":2124.56,"endTime":2131.44,"durationSeconds":7,"level":"C1","overallScore":8,"rationale":"개인화 자산의 희소성을 설득력 있게 말함."},{"segmentIndex":4,"text":"And that's the problem. This is genuinely the most useful AI setup I've seen in months and implemented in Claude.","startTime":28.88,"endTime":36.399,"durationSeconds":8,"level":"B2","overallScore":6.6,"rationale":"가치 판단이 분명한 핵심 주장."},{"segmentIndex":12,"text":"And with a health check and that loop, it just keeps improving over time.","startTime":79.759,"endTime":83.68,"durationSeconds":4,"level":"B2","overallScore":6.6,"rationale":"자기개선 원리를 간결히 제시함."},{"segmentIndex":13,"text":"So, by the end of this video, you'll know exactly what the system is, why it beats every Obsidian Plus plug-in setup you can find for simplicity, and how to build your own right now with Claude.","startTime":83.68,"endTime":94.479,"durationSeconds":11,"level":"B2","overallScore":7.2,"rationale":"가치·비교·결과를 선명히 제시함."},{"segmentIndex":26,"text":"Now, the best bit is those then get fed back in and help to refine it, plus one file at the root.","startTime":165.84,"endTime":171.28,"durationSeconds":5,"level":"B2","overallScore":6.8,"rationale":"피드백 루프의 핵심을 보여줌."},{"segmentIndex":29,"text":"No database, no obsidian, no vault setup, just folders and text files on your computer.","startTime":181.92,"endTime":187.2,"durationSeconds":5,"level":"B2","overallScore":7,"rationale":"단순성의 강점을 선명히 강조함."},{"segmentIndex":33,"text":"If it works for one of the most respected AI researchers alive, it'll probably work for your business.","startTime":202.959,"endTime":208.72,"durationSeconds":6,"level":"B2","overallScore":6.8,"rationale":"권위 기반 일반화가 설득력 있음."},{"segmentIndex":34,"text":"The best thing like what I've done in my notion agent OS is you can then point a custom agent at that knowledge base and it becomes a specialist agent expert that you can speak with.","startTime":208.72,"endTime":219.12,"durationSeconds":10,"level":"C1","overallScore":7.2,"rationale":"응용 가능성과 표현 다양성이 큼."}],"generatedAt":"2026-06-24T23:44:30.370Z","keyClipsTotalSec":696}]}