{"success":true,"count":2,"items":[{"videoId":"8_gtzSXNDvU","chunkIndex":0,"totalChunks":2,"title":"Ivan Zhao, Co-founder & CEO of Notion at Adopt AI - Special Interview: The Future of Knowledge Work — Part 1 of 2","thumbnail":"https://i.ytimg.com/vi/8_gtzSXNDvU/maxresdefault.jpg","duration":971,"uploader":"Adopt AI - Grand Palais","youtubeUrl":"https://www.youtube.com/watch?v=8_gtzSXNDvU","keywords":["artificial-intelligence","knowledge-work","agents","saas","productivity","startup","software","automation","context-management","future-of-work"],"normalizedKeywords":["비즈니스·전략","프로덕트","기술 트렌드"],"targetAudience":[{"who":"창업자","why":"AI를 기존 제품 구조와 결합해 빠르게 전환하는 전략을 배울 수 있음"},{"who":"프로덕트 매니저","why":"지식노동용 AI가 어디서 먼저 먹히는지 제품 관점의 힌트를 얻을 수 있음"},{"who":"엔지니어","why":"컨텍스트 통합과 에이전트 인프라가 AI 제품 성능을 좌우한다는 점이 중요함"},{"who":"지식노동자","why":"AI 에이전트가 개인의 생산성을 어떻게 바꾸는지 이해하는 데 유용함"}],"normalizedAudience":["창업자·스타트업","프로덕트 매니저·기획자","엔지니어·개발자"],"summary":"이 영상은 Ivan Zhao가 Notion이 왜 '도구를 더 만드는 회사'가 아니라 작업 컨텍스트를 하나로 묶는 플랫폼이었는지를 설명하면서, 그 구조가 AI 시대에 큰 장점이 되었다고 말하는 인터뷰다. 핵심 논지는 AI의 한계가 모델 자체보다 주변 인프라와 컨텍스트 통합에 있으며, Notion은 이미 그 문제를 오래 풀어온 덕분에 에이전트 제품을 빠르게 만들 수 있었다는 것이다.\n\n또한 그는 agentic AI가 지식노동을 어떻게 바꿀지에 대해, 아직은 매우 초기지만 프로그래밍과 고객지원처럼 데이터와 도구가 한곳에 모인 영역부터 급격한 생산성 향상이 나타난다고 본다. 한 사람이 여러 AI 에이전트를 관리하며 출력이 수배로 늘어나는 모습이 앞으로 다른 사무직 일에도 확산될 수 있다고 보고, 조직 구조의 변화는 아직 단정하기 이르지만 장기적으로는 더 작은 팀과 더 높은 레버리지로 이동할 가능성을 제시한다.","insights":["AI의 성능은 모델보다 컨텍스트 통합에 크게 좌우된다.","하나의 범용 워크스페이스는 AI 에이전트의 최적 작업대가 된다.","에이전트는 사람을 대체하기보다 사람의 레버리지를 키운다.","먼저 자동화되는 일은 데이터와 도구가 한곳에 모인 업무다.","조직 변화는 개인 생산성 변화보다 느리게 따라온다."],"keyClips":[{"clipId":"8_gtzSXNDvU:c0:5-12","startSegmentIndex":5,"endSegmentIndex":12,"startTime":43.12,"endTime":111.439,"durationSeconds":68.3,"preview":"도구 파편화의 문제","mustSee":false},{"clipId":"8_gtzSXNDvU:c0:13-26","startSegmentIndex":13,"endSegmentIndex":26,"startTime":111.439,"endTime":268.96,"durationSeconds":157.5,"preview":"AI 전환의 핵심","mustSee":true},{"clipId":"8_gtzSXNDvU:c0:30-40","startSegmentIndex":30,"endSegmentIndex":40,"startTime":292.8,"endTime":348.479,"durationSeconds":55.7,"preview":"에이전트의 초기 신호","mustSee":false},{"clipId":"8_gtzSXNDvU:c0:41-56","startSegmentIndex":41,"endSegmentIndex":56,"startTime":348.479,"endTime":540.08,"durationSeconds":191.6,"preview":"40배 생산성의 사례","mustSee":true},{"clipId":"8_gtzSXNDvU:c0:57-62","startSegmentIndex":57,"endSegmentIndex":62,"startTime":540.08,"endTime":595.04,"durationSeconds":55,"preview":"조직 변화의 속도","mustSee":false}],"curatedSegments":[{"segmentIndex":28,"text":"Um but contrast with the machine skill city like Dallas or like Houston if you ever come to America those cities are designed around cars much bigger you can no longer walk from one end to the other end of the city you have to drive right um the city is much bigger people travel faster um it's a different sense of scale I would say we are the world by and large is still in Florence era of doing knowledge work everything is human scaled with people like my co-founder Simon start using AI to do more work and with we start redesigning organizational AI we're entering this machine scale the car scale of knowledge work economy here I don't want to create a judgment saying that car is better where Florence better than Dallas or vice versa I personally I think Florence is better than Dallas but um it's just going to be feel different and that's the future I think the world is heading towards right it's going to be faster we're going to be bigger something used to take a year to do you can done over the weekend and at the end of day applying to our value back to this world I want notion to provide a driving wheels for people to be a human centric way of providing tools so people can control their own destiny where they want to navigate to in this kind of larger scale knowledge war economy >> well thank you very much Ivanzo for being here with us at AdoptAI.","startTime":878.24,"endTime":963.839,"durationSeconds":86,"level":"C1","overallScore":10,"rationale":"장기 비유와 가치 연결이 매우 풍부."},{"segmentIndex":56,"text":"Um the reason why customer support and coding agent are the two first category where there's happens going back to the points I was mentioning earlier uh because the tools and contact are fairly unified in those two categories for customer support all the data is in one place right it's your in the email exchange with the customer for coding all the data is in your GitHub repo of the programmers so everything is fairly unified so when context is unified is easier for AI agent to do work for you but contrast with the rest of knowledge work where notion is the context extremely fragmented therefore it's much harder to build a knowledge agent that do work but that's what we're working on at notion >> and do you think we'll see more and more companies with tiny teams a CEO and uh yeah dozens of employees running AI engines in the future and even maybe a CEO with only AI agents.","startTime":478.72,"endTime":540.08,"durationSeconds":61,"level":"C1","overallScore":9,"rationale":"문맥 통합이 AI 성패라는 핵심 통찰."},{"segmentIndex":6,"text":"We're still in the watermill era of language models. I think we're in the early days.","startTime":642.8,"endTime":649.519,"durationSeconds":7,"level":"B2","overallScore":8.6,"rationale":"강한 비유로 현재 AI 단계를 통찰."},{"segmentIndex":23,"text":"So in 10 or 20 years what will it mean for you to you for notion to be a success and will you stay of course at the at this point >> like I said uh creating something beautiful and useful is at the heart uh will evaluate how much the success based on whether we have done that for the world um I think it's also uh helpful using this moment to reimagine how the world could be different right um earlier I mentioned how my co-founder Simon his individual productivity could change with language model with AI we also talk about how organization could be different and it's time for us to rethink how to redesign organizations if we zooming out the world economy knowledgeable economy could be different too um some metaphor we can use is the concept of uh driving a car in a city.","startTime":793.04,"endTime":856.399,"durationSeconds":63,"level":"C1","overallScore":9,"rationale":"성공 기준과 거시 비전이 밀도 높음."},{"segmentIndex":11,"text":"If you think about the last 15 years of SAS unbundle Microsoft Office and move all the workflow onto the cloud right now just too much fragmentation there's value to rebundle and that's what we're doing. Right?","startTime":96.72,"endTime":104.56,"durationSeconds":8,"level":"C1","overallScore":8,"rationale":"시장 흐름과 전략을 압축해 설명함."},{"segmentIndex":16,"text":"If you been if you've been building with language model or AI today you will know that the large limitation is not the model capability it's the infrastructure the plumbing that are around model that's lacking that's what blocking from the progress of AI applications and notion we have been our architecture has been pretty advantaged because we has been consolidating all the fragmentation of context for quite some time right you probably heard the term of context window.","startTime":139.52,"endTime":169.92000000000002,"durationSeconds":30,"level":"C1","overallScore":8,"rationale":"AI 한계를 인프라 관점에서 설명함."},{"segmentIndex":24,"text":"Notion is very flexible, very horizontal and in the past it's a little bit tricky how horizontal we are make our onboarding and setup difficult but with language model and with AI it turns our weakness into strength.","startTime":217.92000000000002,"endTime":234.159,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"약점이 강점이 되는 전환 통찰이 큼."},{"segmentIndex":48,"text":"In some sense he's elevating himself from a programmer from doing the work to manage agent to do work for him.","startTime":385.759,"endTime":394.4,"durationSeconds":9,"level":"B2","overallScore":7.8,"rationale":"노동에서 관리로의 전환 통찰이 큼."},{"segmentIndex":52,"text":"If we have been talking about that the world has been running on the internet which is information superighway for the past 40 years but if you think about it most of us still paddling bicycles on this inu information superighway.","startTime":414.639,"endTime":431.84,"durationSeconds":17,"level":"C1","overallScore":8,"rationale":"인터넷 비유로 현재 한계를 통찰함."},{"segmentIndex":53,"text":"But as Simon he start to manage encoding agent his throughputs really increase he's starting to driving cars and we're going to see this for the different sector of a knowledge work.","startTime":431.84,"endTime":442.479,"durationSeconds":11,"level":"C1","overallScore":8,"rationale":"자전거-자동차 비유가 인상적임."},{"segmentIndex":2,"text":"And the real productivity increase is when people replace reimagine how factory could be built, how organization could be different with this new technology.","startTime":616.959,"endTime":626.32,"durationSeconds":9,"level":"B2","overallScore":7.8,"rationale":"기술보다 재설계가 핵심이라 설명."},{"segmentIndex":7,"text":"We haven't start rethinking how organization could be different and I think in the next five years the world will be very different because we start thinking how to redesign our organization because of language model power. >> Okay.","startTime":649.519,"endTime":662.88,"durationSeconds":13,"level":"B2","overallScore":7.8,"rationale":"조직 재설계 전망이 뚜렷함."},{"segmentIndex":14,"text":"Um right now it's a such a fascinating time to build products and business and companies compared to the SAS era of you know few years ago I was always joking with my team the AI era make the SAS era feels like sleepwalk it's way easy back then but now it's way competitive but also way more interesting right um all strategy changes all the time but what we believe at notion is at the end of the What separates a good business with a mediocre business?","startTime":690.959,"endTime":722.56,"durationSeconds":32,"level":"C1","overallScore":8.2,"rationale":"시대 비교와 사업 본질 질문이 풍부."},{"segmentIndex":20,"text":"Most the modern technology including semiconductor come from Bell Labs and I was personally really inspired when I heard about that in the entryway of Bell Labs they have a quote and the quote says you can do anything you want to do in Bell Labs as long as fit into one of those two things.","startTime":746.399,"endTime":767.44,"durationSeconds":21,"level":"C1","overallScore":8,"rationale":"벨랩스 일화와 핵심 인용이 강함."},{"segmentIndex":21,"text":"It either has to be very useful or it has to be very beautiful and that has always been the why the values why we build notion.","startTime":767.44,"endTime":777.519,"durationSeconds":10,"level":"B2","overallScore":8,"rationale":"가치 기준이 명확하고 표현도 좋음."},{"segmentIndex":17,"text":"In order to do useful work with language model, you need all the context to be together.","startTime":169.92000000000002,"endTime":175.36,"durationSeconds":5,"level":"B2","overallScore":6.6,"rationale":"맥락 통합의 원리를 명확히 제시함."},{"segmentIndex":22,"text":"Even the creator of the language model doesn't know what it can and cannot do for a very horizontal technology like language model. It likes horizontal building blocks.","startTime":199.36,"endTime":207.68,"durationSeconds":8,"level":"C1","overallScore":7,"rationale":"기술 특성을 일반화해 설명함."},{"segmentIndex":25,"text":"So combining the consolidation of uh context and the horizontal nature of our building blocks we have a really good architecture to launch AI product fairly quickly.","startTime":234.159,"endTime":246.48,"durationSeconds":12,"level":"B2","overallScore":7,"rationale":"강점 결합 논리를 깔끔히 제시함."},{"segmentIndex":40,"text":"And we can use that as a hint to understand how it might change the rest of knowledge work.","startTime":342.72,"endTime":348.479,"durationSeconds":6,"level":"B2","overallScore":7,"rationale":"유추 방식의 설명이 통찰적임."},{"segmentIndex":45,"text":"He doesn't code anymore. Yeah. What he does instead he's managing a three and four coding agents.","startTime":369.199,"endTime":372.24,"durationSeconds":3,"level":"B2","overallScore":6.8,"rationale":"역할 전환을 직접적으로 보여줌."}],"generatedAt":"2026-06-25T00:36:16.952Z","keyClipsTotalSec":692},{"videoId":"8_gtzSXNDvU","chunkIndex":1,"totalChunks":2,"title":"Ivan Zhao, Co-founder & CEO of Notion at Adopt AI - Special Interview: The Future of Knowledge Work — Part 2 of 2","thumbnail":"https://i.ytimg.com/vi/8_gtzSXNDvU/maxresdefault.jpg","duration":971,"uploader":"Adopt AI - Grand Palais","youtubeUrl":"https://www.youtube.com/watch?v=8_gtzSXNDvU","keywords":["ai","knowledge-work","organization-design","leadership","business-strategy","startup","productivity","future-of-work","notion","technology-trends"],"normalizedKeywords":["비즈니스·전략","리더십·매니지먼트","기술 트렌드"],"targetAudience":[{"who":"창업자","why":"AI 시대에 제품보다 조직과 가치가 경쟁력을 만든다는 관점을 얻을 수 있음"},{"who":"프로덕트 리더","why":"AI를 기능이 아니라 업무 구조 변화로 봐야 한다는 시각이 유용함"},{"who":"지식노동자","why":"AI로 개인 생산성과 일하는 방식이 어떻게 바뀌는지 감 잡기 좋음"}],"normalizedAudience":["창업자·스타트업","프로덕트 매니저·기획자","지식노동자 일반"],"summary":"이 영상에서 Ivan Zhao는 AI, 특히 언어 모델의 영향력을 증기기관에 비유하며, 진짜 생산성 혁신은 단순히 기존 업무에 기술을 덧씌우는 데서 오지 않고 조직과 시스템 자체를 다시 설계할 때 나온다고 말한다. 지금은 아직 '물레방아 시대'처럼 초기 단계라서, 개인 생산성 향상보다도 조직 구조와 일하는 방식의 재구상이 더 중요하다고 본다.\n\n그는 Notion이 AI 시대에도 살아남는 방법을 단기 매출보다 가치와 문화에서 찾는다. Bell Labs의 '유용하거나 아름답거나'라는 원칙을 인용하며, Notion이 앞으로도 세계에 유용하고 아름다운 무언가를 만들고 싶다고 강조한다. 마지막으로, AI가 지식노동을 인간 스케일에서 자동차 스케일로 바꿔 더 빠르고 큰 경제를 만들 것이며, 그 변화 속에서 사람들에게 방향키를 쥐여주는 도구가 Notion의 역할이라고 정리한다.","insights":["기술 혁신의 큰 생산성은 도입이 아니라 재설계에서 나온다.","AI는 개인 도구를 넘어 조직 구조를 바꾸는 기술이다.","초기 AI 경쟁에서는 속도보다 가치와 문화가 생존을 좌우한다.","지식노동은 곧 더 큰 규모와 더 빠른 속도의 '기계 스케일'로 간다.","좋은 도구는 사용자를 통제하지 않고 자기 방향을 잡게 해야 한다."],"keyClips":[{"clipId":"8_gtzSXNDvU:c1:1-7","startSegmentIndex":1,"endSegmentIndex":7,"startTime":602.389,"endTime":662.88,"durationSeconds":60.5,"preview":"기술혁신의 본질","mustSee":true},{"clipId":"8_gtzSXNDvU:c1:8-13","startSegmentIndex":8,"endSegmentIndex":13,"startTime":662.88,"endTime":690.959,"durationSeconds":28.1,"preview":"AI시대 생존조건","mustSee":false},{"clipId":"8_gtzSXNDvU:c1:14-22","startSegmentIndex":14,"endSegmentIndex":22,"startTime":690.959,"endTime":793.04,"durationSeconds":102.1,"preview":"유용함과 아름다움","mustSee":true},{"clipId":"8_gtzSXNDvU:c1:23-29","startSegmentIndex":23,"endSegmentIndex":29,"startTime":793.04,"endTime":973.839,"durationSeconds":180.8,"preview":"지식노동의 재설계","mustSee":true}],"curatedSegments":[{"segmentIndex":28,"text":"Um but contrast with the machine skill city like Dallas or like Houston if you ever come to America those cities are designed around cars much bigger you can no longer walk from one end to the other end of the city you have to drive right um the city is much bigger people travel faster um it's a different sense of scale I would say we are the world by and large is still in Florence era of doing knowledge work everything is human scaled with people like my co-founder Simon start using AI to do more work and with we start redesigning organizational AI we're entering this machine scale the car scale of knowledge work economy here I don't want to create a judgment saying that car is better where Florence better than Dallas or vice versa I personally I think Florence is better than Dallas but um it's just going to be feel different and that's the future I think the world is heading towards right it's going to be faster we're going to be bigger something used to take a year to do you can done over the weekend and at the end of day applying to our value back to this world I want notion to provide a driving wheels for people to be a human centric way of providing tools so people can control their own destiny where they want to navigate to in this kind of larger scale knowledge war economy >> well thank you very much Ivanzo for being here with us at AdoptAI.","startTime":878.24,"endTime":963.839,"durationSeconds":86,"level":"C1","overallScore":10,"rationale":"장기 비유와 가치 연결이 매우 풍부."},{"segmentIndex":56,"text":"Um the reason why customer support and coding agent are the two first category where there's happens going back to the points I was mentioning earlier uh because the tools and contact are fairly unified in those two categories for customer support all the data is in one place right it's your in the email exchange with the customer for coding all the data is in your GitHub repo of the programmers so everything is fairly unified so when context is unified is easier for AI agent to do work for you but contrast with the rest of knowledge work where notion is the context extremely fragmented therefore it's much harder to build a knowledge agent that do work but that's what we're working on at notion >> and do you think we'll see more and more companies with tiny teams a CEO and uh yeah dozens of employees running AI engines in the future and even maybe a CEO with only AI agents.","startTime":478.72,"endTime":540.08,"durationSeconds":61,"level":"C1","overallScore":9,"rationale":"문맥 통합이 AI 성패라는 핵심 통찰."},{"segmentIndex":6,"text":"We're still in the watermill era of language models. I think we're in the early days.","startTime":642.8,"endTime":649.519,"durationSeconds":7,"level":"B2","overallScore":8.6,"rationale":"강한 비유로 현재 AI 단계를 통찰."},{"segmentIndex":23,"text":"So in 10 or 20 years what will it mean for you to you for notion to be a success and will you stay of course at the at this point >> like I said uh creating something beautiful and useful is at the heart uh will evaluate how much the success based on whether we have done that for the world um I think it's also uh helpful using this moment to reimagine how the world could be different right um earlier I mentioned how my co-founder Simon his individual productivity could change with language model with AI we also talk about how organization could be different and it's time for us to rethink how to redesign organizations if we zooming out the world economy knowledgeable economy could be different too um some metaphor we can use is the concept of uh driving a car in a city.","startTime":793.04,"endTime":856.399,"durationSeconds":63,"level":"C1","overallScore":9,"rationale":"성공 기준과 거시 비전이 밀도 높음."},{"segmentIndex":11,"text":"If you think about the last 15 years of SAS unbundle Microsoft Office and move all the workflow onto the cloud right now just too much fragmentation there's value to rebundle and that's what we're doing. Right?","startTime":96.72,"endTime":104.56,"durationSeconds":8,"level":"C1","overallScore":8,"rationale":"시장 흐름과 전략을 압축해 설명함."},{"segmentIndex":16,"text":"If you been if you've been building with language model or AI today you will know that the large limitation is not the model capability it's the infrastructure the plumbing that are around model that's lacking that's what blocking from the progress of AI applications and notion we have been our architecture has been pretty advantaged because we has been consolidating all the fragmentation of context for quite some time right you probably heard the term of context window.","startTime":139.52,"endTime":169.92000000000002,"durationSeconds":30,"level":"C1","overallScore":8,"rationale":"AI 한계를 인프라 관점에서 설명함."},{"segmentIndex":24,"text":"Notion is very flexible, very horizontal and in the past it's a little bit tricky how horizontal we are make our onboarding and setup difficult but with language model and with AI it turns our weakness into strength.","startTime":217.92000000000002,"endTime":234.159,"durationSeconds":16,"level":"C1","overallScore":8,"rationale":"약점이 강점이 되는 전환 통찰이 큼."},{"segmentIndex":48,"text":"In some sense he's elevating himself from a programmer from doing the work to manage agent to do work for him.","startTime":385.759,"endTime":394.4,"durationSeconds":9,"level":"B2","overallScore":7.8,"rationale":"노동에서 관리로의 전환 통찰이 큼."},{"segmentIndex":52,"text":"If we have been talking about that the world has been running on the internet which is information superighway for the past 40 years but if you think about it most of us still paddling bicycles on this inu information superighway.","startTime":414.639,"endTime":431.84,"durationSeconds":17,"level":"C1","overallScore":8,"rationale":"인터넷 비유로 현재 한계를 통찰함."},{"segmentIndex":53,"text":"But as Simon he start to manage encoding agent his throughputs really increase he's starting to driving cars and we're going to see this for the different sector of a knowledge work.","startTime":431.84,"endTime":442.479,"durationSeconds":11,"level":"C1","overallScore":8,"rationale":"자전거-자동차 비유가 인상적임."},{"segmentIndex":2,"text":"And the real productivity increase is when people replace reimagine how factory could be built, how organization could be different with this new technology.","startTime":616.959,"endTime":626.32,"durationSeconds":9,"level":"B2","overallScore":7.8,"rationale":"기술보다 재설계가 핵심이라 설명."},{"segmentIndex":7,"text":"We haven't start rethinking how organization could be different and I think in the next five years the world will be very different because we start thinking how to redesign our organization because of language model power. >> Okay.","startTime":649.519,"endTime":662.88,"durationSeconds":13,"level":"B2","overallScore":7.8,"rationale":"조직 재설계 전망이 뚜렷함."},{"segmentIndex":14,"text":"Um right now it's a such a fascinating time to build products and business and companies compared to the SAS era of you know few years ago I was always joking with my team the AI era make the SAS era feels like sleepwalk it's way easy back then but now it's way competitive but also way more interesting right um all strategy changes all the time but what we believe at notion is at the end of the What separates a good business with a mediocre business?","startTime":690.959,"endTime":722.56,"durationSeconds":32,"level":"C1","overallScore":8.2,"rationale":"시대 비교와 사업 본질 질문이 풍부."},{"segmentIndex":20,"text":"Most the modern technology including semiconductor come from Bell Labs and I was personally really inspired when I heard about that in the entryway of Bell Labs they have a quote and the quote says you can do anything you want to do in Bell Labs as long as fit into one of those two things.","startTime":746.399,"endTime":767.44,"durationSeconds":21,"level":"C1","overallScore":8,"rationale":"벨랩스 일화와 핵심 인용이 강함."},{"segmentIndex":21,"text":"It either has to be very useful or it has to be very beautiful and that has always been the why the values why we build notion.","startTime":767.44,"endTime":777.519,"durationSeconds":10,"level":"B2","overallScore":8,"rationale":"가치 기준이 명확하고 표현도 좋음."},{"segmentIndex":17,"text":"In order to do useful work with language model, you need all the context to be together.","startTime":169.92000000000002,"endTime":175.36,"durationSeconds":5,"level":"B2","overallScore":6.6,"rationale":"맥락 통합의 원리를 명확히 제시함."},{"segmentIndex":22,"text":"Even the creator of the language model doesn't know what it can and cannot do for a very horizontal technology like language model. It likes horizontal building blocks.","startTime":199.36,"endTime":207.68,"durationSeconds":8,"level":"C1","overallScore":7,"rationale":"기술 특성을 일반화해 설명함."},{"segmentIndex":25,"text":"So combining the consolidation of uh context and the horizontal nature of our building blocks we have a really good architecture to launch AI product fairly quickly.","startTime":234.159,"endTime":246.48,"durationSeconds":12,"level":"B2","overallScore":7,"rationale":"강점 결합 논리를 깔끔히 제시함."},{"segmentIndex":40,"text":"And we can use that as a hint to understand how it might change the rest of knowledge work.","startTime":342.72,"endTime":348.479,"durationSeconds":6,"level":"B2","overallScore":7,"rationale":"유추 방식의 설명이 통찰적임."},{"segmentIndex":45,"text":"He doesn't code anymore. Yeah. What he does instead he's managing a three and four coding agents.","startTime":369.199,"endTime":372.24,"durationSeconds":3,"level":"B2","overallScore":6.8,"rationale":"역할 전환을 직접적으로 보여줌."}],"generatedAt":"2026-06-25T00:36:48.388Z","keyClipsTotalSec":692}]}