{"success":true,"count":6,"items":[{"videoId":"cMAs8z2dehs","chunkIndex":0,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 1 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["artificial-intelligence","agentic-ai","google-ai","coding-agents","search","product-strategy","developer-tools","gemini","automation"],"normalizedKeywords":["기술 트렌드","프로덕트","비즈니스·전략"],"targetAudience":[{"who":"AI 제품 기획자","why":"에이전트 기능이 기존 제품과 비즈니스에 미치는 영향을 이해할 수 있음"},{"who":"엔지니어","why":"Agent harness와 코딩 에이전트의 제품화 방향을 배울 수 있음"},{"who":"창업자","why":"AI가 플랫폼과 수익모델을 어떻게 재편하는지 감을 잡는 데 도움됨"}],"normalizedAudience":["프로덕트 매니저·기획자","엔지니어·개발자","창업자·스타트업"],"summary":"이 영상은 Google AI Studio와 Gemini API를 이끄는 Logan Kilpatrick이 Google의 'Agentic Gemini era'를 어떻게 보는지 설명하는 대담이다. 핵심은 언어모델 자체가 Google 제품의 공통 기반이었던 시대를 지나, 이제는 'agent harness'가 검색·Gemini 앱·Cloud·AI Studio를 잇는 새로운 공통 레이어가 되고 있다는 점이다. 단순한 답변 생성이 아니라, 사용자를 대신해 실제로 행동하는 제품으로 전환하는 과정에서 Google이 어떤 방향으로 제품을 재구성하고 있는지 보여준다.\n\n동시에 이 대화는 에이전트가 기존 비즈니스를 잠식할 수 있다는 우려에 대해, Google이 단기적으로는 'eyeball time'보다 고객의 최종 성과를 더 중시하며, 검색처럼 오히려 사용량을 늘리는 positive-sum 효과를 기대한다는 관점을 제시한다. 다만 Google의 방대한 사용자 기반과 책임 때문에 현재의 에이전틱 수준은 'crawl'에 가깝고, 실험적 환경에서는 'walk'에 가까운 경험이 가능하다는 식으로 현실적 제약도 함께 인정한다. 마지막에는 에이전트가 광고·검색·쇼핑 같은 인접 영역의 가치 포착 방식을 어떻게 바꿀지, 그리고 그 변화가 생각보다 급진적이지 않을 수도 있다는 시각을 공유한다.","insights":["에이전트 시대의 공통 레이어는 모델이 아니라 하니스다.","AI는 눈길을 빼앗기보다 고객의 목표 달성률을 높여야 한다.","에이전트 확산은 검색을 대체하기보다 사용량을 증폭시킬 수 있다.","대규모 소비자 제품은 혁신보다 신뢰와 점진성이 먼저다.","지금의 변화는 급격한 단절보다 기존 흐름의 누적일 가능성이 크다."],"keyClips":[{"clipId":"cMAs8z2dehs:c0:16-24","startSegmentIndex":16,"endSegmentIndex":24,"startTime":105.36,"endTime":191.44,"durationSeconds":86.1,"preview":"에이전틱 시대의 전환","mustSee":true},{"clipId":"cMAs8z2dehs:c0:25-39","startSegmentIndex":25,"endSegmentIndex":39,"startTime":191.44,"endTime":307.919,"durationSeconds":116.5,"preview":"코딩 하니스의 확장","mustSee":false},{"clipId":"cMAs8z2dehs:c0:40-50","startSegmentIndex":40,"endSegmentIndex":50,"startTime":307.919,"endTime":439.52,"durationSeconds":131.6,"preview":"눈길보다 결과","mustSee":true},{"clipId":"cMAs8z2dehs:c0:61-70","startSegmentIndex":61,"endSegmentIndex":70,"startTime":504.8,"endTime":568,"durationSeconds":63.2,"preview":"지금은 아직 걸음마","mustSee":false},{"clipId":"cMAs8z2dehs:c0:71-72","startSegmentIndex":71,"endSegmentIndex":72,"startTime":568,"endTime":582.399,"durationSeconds":14.4,"preview":"가까운 미래의 전선","mustSee":false}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T12:00:56.401Z","keyClipsTotalSec":1075},{"videoId":"cMAs8z2dehs","chunkIndex":1,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 2 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["ai","generative-ai","agentic-ai","coding","large-language-models","product-strategy","enterprise-tech","google-deepmind","machine-learning","model-training"],"normalizedKeywords":["엔지니어링","프로덕트","비즈니스·전략"],"targetAudience":[{"who":"엔지니어","why":"코딩 모델, 에이전트, 내부 도그푸딩 전략을 직접 다룸"},{"who":"프로덕트 매니저","why":"AI 제품 수가 늘어날지, 어떤 UX가 먹힐지 논의한다"},{"who":"창업자","why":"AI 경쟁 구도와 제품-모델-배포 전략 감각을 얻기 좋다"},{"who":"투자자","why":"모델 성능, 사용성, 생태계 전개 속도를 읽는 데 유용하다"}],"normalizedAudience":["엔지니어·개발자","프로덕트 매니저·기획자","창업자·스타트업"],"summary":"이 대화는 Google DeepMind의 관점에서 AI 제품과 코딩 모델, 장기 에이전트의 현재 위치를 점검하는 내용이다. 한쪽에서는 Google이 AI를 위한 제품 표면을 몇 개나 가져갈지, 범용 인터페이스가 정말 더 나은지에 대한 제품 철학을 논하고, 다른 한쪽에서는 실제로 가장 잘 먹히는 에이전트가 왜 아직 코딩에 집중되는지와 긴 호흡의 작업을 수행하는 모델을 만들기 위해 무엇이 필요한지를 설명한다.\n\n특히 Gemini와 DeepMind의 내부 개발 문화, 도그푸딩, pre-training과 post-training의 역할, 그리고 새 모델이 나와도 시장 인식이 얼마나 빠르게 바뀌는지를 강조한다. 핵심 메시지는 '모델 성능만으로는 부족하고, 그 성능이 실제 긴 작업을 굴리는 제품 엔진과 피드백 루프로 연결되어야 한다'는 점이다.","insights":["AI는 만능 표면보다 특화된 제품이 더 자연스러울 수 있다.","범용 인터페이스는 멋져도 인지부하를 늘릴 수 있다.","에이전트의 '성공'은 장기 자율작업이 되는가로 봐야 한다.","코딩 모델은 제품 엔진이 있어야 진짜 빨리 좋아진다.","내부 도그푸딩과 대규모 엔지니어 피드백이 경쟁우위다."],"keyClips":[{"clipId":"cMAs8z2dehs:c1:2-10","startSegmentIndex":2,"endSegmentIndex":10,"startTime":613.839,"endTime":735.36,"durationSeconds":121.5,"preview":"AI 표면은 몇 개일까","mustSee":false},{"clipId":"cMAs8z2dehs:c1:11-20","startSegmentIndex":11,"endSegmentIndex":20,"startTime":735.36,"endTime":862.88,"durationSeconds":127.5,"preview":"에이전트 성공의 기준","mustSee":true},{"clipId":"cMAs8z2dehs:c1:22-31","startSegmentIndex":22,"endSegmentIndex":31,"startTime":866.16,"endTime":940.88,"durationSeconds":74.7,"preview":"Gemini가 늦어 보인 이유","mustSee":false},{"clipId":"cMAs8z2dehs:c1:33-48","startSegmentIndex":33,"endSegmentIndex":48,"startTime":945.279,"endTime":1092.24,"durationSeconds":147,"preview":"코딩이 모델을 끌어올린다","mustSee":false},{"clipId":"cMAs8z2dehs:c1:49-55","startSegmentIndex":49,"endSegmentIndex":55,"startTime":1092.24,"endTime":1152,"durationSeconds":59.8,"preview":"도그푸딩의 힘","mustSee":false},{"clipId":"cMAs8z2dehs:c1:56-61","startSegmentIndex":56,"endSegmentIndex":61,"startTime":1152,"endTime":1199.2,"durationSeconds":47.2,"preview":"연구 가속의 조건","mustSee":false}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T14:59:58.037Z","keyClipsTotalSec":1075},{"videoId":"cMAs8z2dehs","chunkIndex":2,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 3 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["ai-coding","code-generation","world-models","video-games","productivity","developer-tools","generative-ai","machine-learning","ai-product"],"normalizedKeywords":["엔지니어링","프로덕트","기술 트렌드"],"targetAudience":[{"who":"엔지니어","why":"코딩 에이전트와 개발 생산성 변화가 실제로 어디까지 왔는지 이해할 수 있음"},{"who":"프로덕트 매니저","why":"모델 성능보다 제품 스캐폴딩이 경험의 완성도를 좌우한다는 점이 중요함"},{"who":"창업자","why":"AI로 게임·도구를 만드는 새로운 제품 기회의 간극을 읽을 수 있음"}],"normalizedAudience":["엔지니어·개발자","프로덕트 매니저·기획자","창업자·스타트업"],"summary":"이 영상은 AI 코딩이 이미 개발 생산성을 크게 끌어올리고 있으며, 적어도 코드 영역에서는 '좁은 슈퍼지능'처럼 느껴질 정도라고 말한다. 다만 화자는 AGI라는 거대한 서사를 앞세우기보다, 지금 당장 인간 개발자를 어떻게 증폭시키는지에 더 주목해야 한다고 본다. 동시에 이런 변화는 코딩에만 머무르지 않고, 수학·금융·과학처럼 검증 가능한 영역으로 먼저 확장될 가능성이 크다고 전망한다.\n\n후반부에서는 '누구나 비브코딩으로 게임을 만들 수 있는가'라는 질문을 중심으로, 게임은 모델만 좋아진다고 완성되지 않고 스캐폴딩, 오케스트레이션, 툴링, 신뢰성 같은 제품 계층이 필요하다고 설명한다. 현재는 게임 엔진+코딩 에이전트 조합이 단기적으로 더 유망하지만, 장기적으로는 world model과 coding agent의 경계가 흐려질 것이라고 본다. 핵심 메시지는 모델 능력 자체보다 이를 실제 사용 가능한 제품으로 바꾸는 구조 설계가 더 중요하다는 것이다.","insights":["AI 코딩의 가치는 AGI보다 현재의 생산성 증폭에 있다.","좋은 모델만으로는 게임이 안 되고, 스캐폴딩이 성패를 가른다.","검증 가능한 분야일수록 AI의 진전이 먼저 체감된다.","AI는 개발자를 대체하기보다 더 큰 야망을 갖게 만든다.","world model의 핵심은 이름이 아니라 실제 사용성이다."],"keyClips":[{"clipId":"cMAs8z2dehs:c2:1-16","startSegmentIndex":1,"endSegmentIndex":16,"startTime":1200.549,"endTime":1378.4,"durationSeconds":177.9,"preview":"코딩은 이미 한계 넘어","mustSee":true},{"clipId":"cMAs8z2dehs:c2:17-26","startSegmentIndex":17,"endSegmentIndex":26,"startTime":1378.4,"endTime":1452.48,"durationSeconds":74.1,"preview":"다음 슈퍼지능 분야","mustSee":false},{"clipId":"cMAs8z2dehs:c2:30-48","startSegmentIndex":30,"endSegmentIndex":48,"startTime":1469.6,"endTime":1681.039,"durationSeconds":211.4,"preview":"비브코드 게임의 조건","mustSee":false},{"clipId":"cMAs8z2dehs:c2:50-63","startSegmentIndex":50,"endSegmentIndex":63,"startTime":1684.64,"endTime":1800.88,"durationSeconds":116.2,"preview":"월드모델의 경계","mustSee":false}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T15:00:34.618Z","keyClipsTotalSec":1075},{"videoId":"cMAs8z2dehs","chunkIndex":3,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 4 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["generative-ai","multimodal-ai","video-editing","world-models","android-apps","ai-studio","agentic-systems","model-infrastructure","content-creation","product-strategy"],"normalizedKeywords":["기술 트렌드","엔지니어링","프로덕트"],"targetAudience":[{"who":"AI 제품 기획자","why":"멀티모달 모델과 제품화 전략이 어떻게 맞물리는지 이해할 수 있음"},{"who":"엔지니어","why":"단일 모델·에이전트 하네스·스캐폴딩의 구조 변화를 배울 수 있음"},{"who":"크리에이터","why":"AI가 콘텐츠의 형식은 바꾸되 사람성과 진정성은 유지하는 방식이 중요함"},{"who":"모바일 앱 개발자","why":"AI Studio로 안드로이드 앱을 빠르게 만드는 흐름과 기회를 볼 수 있음"}],"normalizedAudience":["프로덕트 매니저·기획자","엔지니어·개발자","크리에이터·작가"],"summary":"이 구간은 Google의 Omni 모델, AI Studio, 그리고 'model eats the harness'라는 개념을 중심으로 AI 제품의 구조 변화가 어떻게 일어나고 있는지 설명한다. 전통적인 월드 모델은 비싸고 비실용적이었지만, 이제는 세계에 대한 이해를 가진 단일 모델이 영상 편집 같은 실제 사용 사례를 먼저 풀어내며 멀티모달 경험의 방향을 바꾸고 있다고 말한다.\n\n또한 콘텐츠 제작에서는 AI가 사람의 말과 정체성을 바꾸기보다 배경, 세트, 장면 같은 비본질적 요소를 바꾸는 방향이 더 바람직하다고 본다. 코드 쪽에서는 AI Studio로 안드로이드 앱을 쉽게 만들게 하면서, Google 생태계의 기능을 한곳에 모아 개인 문제를 해결하는 소프트웨어 제작을 가속하고 있다는 점을 강조한다. 마지막으로 '모델이 하네스를 먹는다'는 말은, 모델이 단순한 가중치가 아니라 에이전트 도구·검색·실행 환경까지 포함한 확장 시스템으로 진화하고 있음을 뜻한다고 정리한다.","insights":["모델 경쟁의 중심은 단일 능력보다 통합된 경험이다.","진짜 혁신은 모델이 아니라 주변 하네스를 흡수하는 데서 온다.","AI는 사람을 바꾸기보다 비본질적 맥락을 바꾸는 쪽이 강하다.","개인 문제를 해결하는 작은 앱이 AI 시대의 핵심 수요다.","운영체제 안의 자연스러운 흐름에 AI를 넣을수록 사용성이 커진다."],"keyClips":[{"clipId":"cMAs8z2dehs:c3:3-21","startSegmentIndex":3,"endSegmentIndex":21,"startTime":1817.76,"endTime":1955.2,"durationSeconds":137.4,"preview":"단일 오모니 모델","mustSee":true},{"clipId":"cMAs8z2dehs:c3:23-45","startSegmentIndex":23,"endSegmentIndex":45,"startTime":1958.72,"endTime":2121.839,"durationSeconds":163.1,"preview":"콘텐츠는 사람 중심","mustSee":false},{"clipId":"cMAs8z2dehs:c3:46-62","startSegmentIndex":46,"endSegmentIndex":62,"startTime":2121.839,"endTime":2268.72,"durationSeconds":146.9,"preview":"앱 제작의 민주화","mustSee":false},{"clipId":"cMAs8z2dehs:c3:63-74","startSegmentIndex":63,"endSegmentIndex":74,"startTime":2268.72,"endTime":2383.839,"durationSeconds":115.1,"preview":"하네스를 먹는 모델","mustSee":true}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T15:01:13.435Z","keyClipsTotalSec":1075},{"videoId":"cMAs8z2dehs","chunkIndex":4,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 5 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["ai","large-language-models","agents","startup","google-deepmind","product-strategy","enterprise-software","engineering","leadership"],"normalizedKeywords":["비즈니스·전략","프로덕트","기술 트렌드"],"targetAudience":[{"who":"창업자","why":"모델이 하네스를 흡수할 때 어떤 차별화가 남는지 감을 잡을 수 있음"},{"who":"프로덕트 매니저","why":"플랫폼 락인, 추상화, 수직화 전략을 제품 관점에서 이해하기 좋음"},{"who":"엔지니어","why":"에이전트 하네스와 모델 능력의 경계가 어디로 이동하는지 볼 수 있음"},{"who":"리더십 관심자","why":"대기업과 연구조직이 어떻게 초점을 잃지 않으려 하는지 배울 수 있음"}],"normalizedAudience":["창업자·스타트업","프로덕트 매니저·기획자","엔지니어·개발자"],"summary":"이 영상은 AI 모델이 점점 더 많은 '하네스(harness, 외부 도구·작업 흐름)'를 흡수하면서, 지금 스타트업들이 쌓는 차별화가 12개월 뒤에도 유효할지에 대해 이야기한다. 핵심은 모델이 검색·코드 실행·에이전트 조합을 더 자연스럽게 처리하게 되면, 단순히 자기만의 하네스를 만드는 것만으로는 오래된 해자가 되기 어렵다는 점이다. 대신 하네스 호환성을 객관적으로 재는 벤치마크가 필요하고, 차별화는 점점 더 상위의 문제로 이동할 것이라고 본다.\n\n이어지는 논의는 '모델이 하네스를 먹어도' 왜 여전히 스타트업 기회가 큰가로 확장된다. 대기업은 사용자와 제품 수가 많아 한 도메인에 집중하기 어렵지만, 스타트업은 수직 도메인과 위험 감수 성향을 활용해 더 빠르게 움직일 수 있다는 주장이다. 후반부에서는 Google DeepMind 내부 문화, 리더십의 색깔, 그리고 거대한 조직 안에서 연구와 제품화를 동시에 추진하는 경험을 공유한다.","insights":["하네스는 장기 해자가 아니라 모델로 흡수될 기능일 수 있다.","차별화는 도구 제작보다 모델이 못하는 상위 문제로 이동한다.","모델의 하네스 적응력은 생태계 차원의 벤치마크가 필요하다.","스타트업의 강점은 규모가 아니라 좁은 문제에 대한 집중력이다.","대기업은 사용자와 제품이 많아 한 도메인에 올인하기 어렵다."],"keyClips":[{"clipId":"cMAs8z2dehs:c4:1-10","startSegmentIndex":1,"endSegmentIndex":10,"startTime":2400.23,"endTime":2498.56,"durationSeconds":98.3,"preview":"하네스의 해자 붕괴","mustSee":false},{"clipId":"cMAs8z2dehs:c4:11-24","startSegmentIndex":11,"endSegmentIndex":24,"startTime":2498.56,"endTime":2633.28,"durationSeconds":134.7,"preview":"스타트업 기회는 더 커진다","mustSee":true},{"clipId":"cMAs8z2dehs:c4:29-43","startSegmentIndex":29,"endSegmentIndex":43,"startTime":2664.88,"endTime":2831.839,"durationSeconds":167,"preview":"딥마인드 문화의 핵심","mustSee":false},{"clipId":"cMAs8z2dehs:c4:45-62","startSegmentIndex":45,"endSegmentIndex":62,"startTime":2840,"endTime":3004.8,"durationSeconds":164.8,"preview":"구글 안의 실행 현실","mustSee":true}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T15:01:36.063Z","keyClipsTotalSec":1075},{"videoId":"cMAs8z2dehs","chunkIndex":5,"totalChunks":6,"title":"Google DeepMind's Logan Kilpatrick: Why the Model Eats the Harness — Part 6 of 6","thumbnail":"https://i.ytimg.com/vi/cMAs8z2dehs/maxresdefault.jpg","duration":3069,"uploader":"Sequoia Capital","youtubeUrl":"https://www.youtube.com/watch?v=cMAs8z2dehs","keywords":["artificial-intelligence","google-deepmind","llm","agents","coding","world-models","human-ai","research-culture"],"normalizedKeywords":["기술 트렌드","엔지니어링","비즈니스·전략"],"targetAudience":[{"who":"AI 연구자","why":"에이전트, 월드 모델, 하니스 같은 연구 화두를 폭넓게 정리한 대화라서 유용함"},{"who":"엔지니어","why":"모델과 하니스의 관계, 실제 적용 관점의 뉘앙스를 이해하는 데 도움됨"},{"who":"창업자","why":"AI 제품과 기술 방향의 큰 흐름을 빠르게 파악하기 좋음"}],"normalizedAudience":["리서처·학자","엔지니어·개발자","창업자·스타트업"],"summary":"이 구간은 대화의 마무리로, 서로의 작업에 대한 감사와 호감을 주고받으며 인터뷰를 정리한다. 로건은 이번 대화가 에이전트, 코딩, 월드 모델, 하니스, GDM 문화까지 폭넓게 다뤘고 흥미로운 배움이 많았다고 요약한다. 또한 앞으로 사람들이 어디까지 발전시킬지 기대된다고 말하며, 마지막에는 가벼운 농담처럼 '개'가 등장하길 바란다는 식으로 친근하게 끝맺는다. 전체적으로는 기술적 결론보다, AI를 둘러싼 중요한 문제를 진지하면서도 인간적인 톤으로 다뤄온 대화의 온기를 남기는 엔딩이다.","insights":["기술 논의의 가치는 정확한 결론보다 넓은 맥락 정리에 있다.","좋은 인터뷰는 연구 주제와 문화적 감각을 함께 남긴다.","AI 담론은 차가운 분석만이 아니라 인간적 톤이 중요하다.","미래의 확실한 답보다 '다음에 무엇을 만들까'가 더 생산적이다."],"keyClips":[{"clipId":"cMAs8z2dehs:c5:1-3","startSegmentIndex":1,"endSegmentIndex":3,"startTime":3000.39,"endTime":3022.559,"durationSeconds":22.2,"preview":"대화의 전체 의미","mustSee":false},{"clipId":"cMAs8z2dehs:c5:4-6","startSegmentIndex":4,"endSegmentIndex":6,"startTime":3022.559,"endTime":3046.079,"durationSeconds":23.5,"preview":"따뜻한 마무리","mustSee":false}],"curatedSegments":[{"segmentIndex":47,"text":"Um, but from like my early feelings of how a lot of this is playing out, it does feel like it's very positive some from like an ecosystem value creation like how the human behavior aspect of it turns out I think is like somewhat clear in the next 1 to two years much less clear you know 3 to 5 years from now when the technology is improved and the products probably look a little bit different than the way that they do but ultimately like that is the success of product I think like we have a bunch of conversations with Demis all the time and it's like the point of building the technology is so that it can go and do stuff for you.","startTime":384.56,"endTime":418.479,"durationSeconds":34,"level":"C1","overallScore":7.8,"rationale":"현재와 미래를 잇는 핵심 관점 제시."},{"segmentIndex":62,"text":"Um, and I think Demis sort of like framed it to the world, uh, rightfully so, as a world model because of just like the level of understanding that it has of the world.","startTime":1781.679,"endTime":1790.88,"durationSeconds":9,"level":"C1","overallScore":7.8,"rationale":"세계모델로 보는 관점 설명이 핵심."},{"segmentIndex":73,"text":"um it was a set of weights and it was like really like how can you like as simple as possible send tokens in and get tokens out and I think we've just like progressively step by step we still call it the model we still call it you know Gemini 3.5 you still call it GPT whatever and claude whatever but like it's actually not just the weights anymore it's like an entire exp expanding sprawling system that's built around the weights um that sort of like enable a lot of these like next generation experiences from agentic tool calling to tool you know like all these hosted tools search code execution etc. Um you know the models are now being spun up in containers and sort of have an agent harness and all that stuff.","startTime":2334.96,"endTime":2373.839,"durationSeconds":39,"level":"C1","overallScore":8,"rationale":"모델이 시스템으로 확장된다는 핵심 통찰."},{"segmentIndex":13,"text":"both on one hand I everywhere I look I'm like there's never been more opportunity to go and build something at the same time obviously the models are doing more than they've ever done before um I think there's like you know there's that thread of capability overhang which I think there's a huge amount of alpha in there's the thread of the model companies are like going after these like very general problems um and there's just like so much value in these like verticalized domains if you have expertise in that domain you sort of like know the customers you know the ecosystem like it's you can really like run laps around even the best model labs because like focus is the like superpower of startups.","startTime":2513.28,"endTime":2550.16,"durationSeconds":37,"level":"C1","overallScore":8,"rationale":"스타트업 기회와 집중의 원리를 강하게 설명."},{"segmentIndex":9,"text":"and just like there's like so much subtle subtlety in getting that right and the model crushed it and it's just it very interesting and like still trying to like absorb and digest like what that means for you know the way we make content and all these other things.","startTime":50.8,"endTime":59.12,"durationSeconds":8,"level":"C1","overallScore":7.4,"rationale":"미묘함의 중요성을 길게 풀어냄."},{"segmentIndex":23,"text":"And so historically like prior to Gemini there actually like wasn't a through line for the you know probably subund number of Google products that we have the 50 Google products we have there wasn't a through line we had Gemini it became this through line everything is now sort of using Gemini in some way that's now becoming true for anti-gravity as sort of all of the products rebase um to become sort of like agentic native products and like actually taking action on behalf of users and helping them get things done.","startTime":153.519,"endTime":181.599,"durationSeconds":28,"level":"C1","overallScore":7.2,"rationale":"제품 변화의 방향을 길게 설명함."},{"segmentIndex":70,"text":"Like I think they have a lot of responsibility to actually do that in a way that it brings people along and doesn't just like change everything of how they interact with the internet and the way they associate with products and stuff like that.","startTime":557.519,"endTime":568,"durationSeconds":10,"level":"C1","overallScore":6.8,"rationale":"변화를 끌고 가는 책임을 설명함."},{"segmentIndex":3,"text":">> It's tough. I think a lot of this is actually baked in just like how humans consume products and my sense is that there's something nice about like having this like compartmentalization and this like specialization of products where like it becomes if you end up with a product that is like doing everything for you inherently there's more work involved in using that version of the product I think would be like the default state I think maybe somebody will spin together like the truly magic experience that doesn't make that true but I think the long tale of folks end up having to spend more mental energy and more time to actually like get the general purpose product to do the thing that they actually want to do versus like there's something nice about I click my calendar app, it just shows me my calendar like I don't need to worry and deal with anything else. I think a lot of the narrative was like Google has taken a huge leap forward um and made that happen.","startTime":621.44,"endTime":650.32,"durationSeconds":29,"level":"C1","overallScore":7.2,"rationale":"제품 분화에 대한 명확한 관점 제시."},{"segmentIndex":19,"text":"So even if the enterprises haven't felt it outside of coding like they are going to like this year um as sort of a bunch of those other use cases get much better as well >> from like a you know from the deep mind perspective do you think long horizon agents is like a KPI that matters is it they is it the KPI that matters >> it definitely matters um I think for deep mind like we're doing lots of things which we can talk more about later like there's you know a huge portfolio of different bets that are taking place longrunning agents obviously matters a And I think also like specifically coding agents in that matters a lot like it clearly is an accelerant of like every other part of your business if you have a great coding model.","startTime":820.16,"endTime":859.199,"durationSeconds":39,"level":"C1","overallScore":6.6,"rationale":"장기 에이전트의 전략적 중요성 강조."},{"segmentIndex":35,"text":"I think one of the takeaways is that it's actually really hard to make a great coding model for this like um for this developer use case of like really long running suite work if you don't actually have a product that does that.","startTime":955.04,"endTime":968.639,"durationSeconds":14,"level":"C1","overallScore":6.6,"rationale":"제품이 모델 성능에 미치는 영향 제시."},{"segmentIndex":10,"text":"Um, and so I spent a lot of time just like letting that fact sort of just like wash over me because I think it's e like obviously building AGI is super important and very interesting, but like building AGI if it sort of like takes away from the story of like the current present capability of the technology I think is actually like kind of a bad a think is actually like kind of a bad sort of like trade-off.","startTime":1270.799,"endTime":1292.48,"durationSeconds":22,"level":"C1","overallScore":7,"rationale":"현재 성능과 AGI 서사의 균형을 말함."},{"segmentIndex":16,"text":"and sort of it does it adds a different layer of sort of um responsibil or like some a different layer of burden actually because I'm like oh I can't just like do the sort of MVP of this like I actually need to like go 10 steps further because the technology enables me and like resetting my level of ambition I think is something that I I've also spent a bunch of time thinking about but I think that will happen in other these like vertical super intelligence domains um which will be interesting and it feels like we're going to get a bunch of those before we've like solved like it's almost like jagged super intelligence I jagged like jagged super intelligence I think is what we'll end up with.","startTime":1340.48,"endTime":1378.4,"durationSeconds":38,"level":"C1","overallScore":7.2,"rationale":"야망 조정과 책임 증가를 깊게 설명함."},{"segmentIndex":71,"text":">> What are your thoughts? >> Yeah, I think it's true. And I think part of this is like what we have historically thought of as the model is not the model anymore.","startTime":2317.04,"endTime":2327.599,"durationSeconds":11,"level":"B2","overallScore":6.6,"rationale":"모델 개념 변화에 대한 핵심 주장임."},{"segmentIndex":74,"text":"So the scaffolding is like oftent times a couple of steps ahead of like where the act what is like baked directly into the model and then what ends up happening is like the model eats that scaffolding and it becomes part of like the native model system and there's still value in having sort of the external scaffolding in certain cases and like search maybe is an example of this like there's lots of folks who use different search providers and there's different like use cases that you want and so like sure maybe the","startTime":2373.839,"endTime":2383.839,"durationSeconds":10,"level":"C1","overallScore":7,"rationale":"스캐폴딩이 모델에 흡수된다는 설명임."},{"segmentIndex":4,"text":"it won't be in sort of trying to spin your own harness because the model just like does it natively >> but I thought that the part of the reason why people are building their own harnesses is because if you use a harness from any given model provider you're locked in right so a lot of the application companies want flexibility which is why they're building their own harnesses >> yeah and I think that's part of the scaffolding story is like that starts out perhaps true but then as the model capability improves like it becomes less true over time actually I think the model the like you don't have a generalized model if it can't use another harness.","startTime":2430.32,"endTime":2445.839,"durationSeconds":16,"level":"C1","overallScore":7,"rationale":"잠금효과와 범용성 기준을 논리적으로 설명함."},{"segmentIndex":49,"text":"So it's like on one hand you have sort of like the deeprooted lab culture and the other hand you have sort of like all of these partners across the Google ecosystem that we're collaborating with everybody from Android that we talked about earlier to Google cloud to you know Gmail to workspace etc and so it's an interesting blend of like I think there's lots of research work happening but like there's tons of applied work that's happening to like actually like work with some of like the forefront customers like deploying Gemini to billion user products is a problem that like only two companies in the world have.","startTime":2860.8,"endTime":2875.52,"durationSeconds":15,"level":"C1","overallScore":6.8,"rationale":"연구와 제품 실행의 결합을 잘 보여줌."},{"segmentIndex":61,"text":"And so you take the like magic of Google and you water it down through like a lot of people and a lot of process and you actually yeah >> you miss the beautiful story which is like Google's doing the most interesting technology in the world and like helping our users with some of the hardest problems in the world.","startTime":2978.96,"endTime":2994.8,"durationSeconds":16,"level":"C1","overallScore":6.6,"rationale":"핵심 메시지와 자연스런 표현이 함께 있음."},{"segmentIndex":22,"text":"I think for us this agentic layer and I think we announced this actually at IO um sort of being powered by the anti-gravity agent harness is this like additional through line for Google that sort of connects all of our products that they're sort of like based on now.","startTime":138.56,"endTime":153.519,"durationSeconds":15,"level":"C1","overallScore":6.4,"rationale":"제품 전략을 하나의 축으로 설명함."},{"segmentIndex":27,"text":"You have a CLI. You have an SDK. But I actually think and I don't know how much we've framed it this way, but like it really is an ecosystem of stuff that we built and it's designed to sort of like meet developers wherever they are.","startTime":202.879,"endTime":212.4,"durationSeconds":10,"level":"C1","overallScore":6.2,"rationale":"대상 맞춤 전략을 잘 설명함."},{"segmentIndex":36,"text":"I think more generically too it's just like it is the agent harness. I think like coding as sort of like a specialized use case of the agent harness I think is obviously powerful but it is like coding has proved to be the general purpose agent harness in addition to also working really well for coding.","startTime":259.84,"endTime":273.6,"durationSeconds":14,"level":"C1","overallScore":6.2,"rationale":"용도 일반화에 대한 설명이 핵심임."}],"generatedAt":"2026-06-22T15:02:01.662Z","keyClipsTotalSec":1075}]}