MrBeast 首席产品技术官：用 AI 把每支视频当作产品发布

简要描述Brief Description

Carlos 采访了 Shiva Rajaraman，探讨 Beast Industries 的定制操作系统、每日产品交付、AI 辅助探索以及机构记忆。Carlos interviews Shiva Rajaraman about Beast Industries’ custom operating system, daily product delivery, AI-assisted exploration and institutional memory.

目录Table of Contents

• 引言
  Introduction
• 一种创意作品，而非一个平台
  A creative output rather than a platform
• 内部工作流程与定制软件
  Internal workflows and custom software
• 受众互动与变现
  Audience engagement and monetization
• Jimmy与日常生产流程
  Jimmy and the daily production process
• 与嵌入式工程师每日交付
  Shipping daily with embedded engineers
• 创始人、稀缺性与探索
  Founders, scarcity and exploration
• 互联工作流与机构记忆
  Connected workflows and institutional memory
• 结语
  Closing

引言Introduction

Carlos：嘿，我是 Carlos，Product School 的首席执行官，也是这档产品播客的主持人。今天的嘉宾是 Shiva Rajaraman，Mr. Beast 的首席产品与技术官。看看这些标志。Carlos: Hey, this is Carlos, CEO at Product School and your host on the product podcast. My guest today is Shiva Rajaraman, chief product and technology officer at Mr. Beast. Look at the logos.

他在 Google 和 YouTube 工作了八年，之后先后供职于 Spotify、担任 WeWork 首席技术官、负责 Meta 的 Commerce 业务，并在 OpenSea 和 Uber 工作。如今，他打造的产品背后，有 5.2 亿订阅者，90 天内产生了 16 亿次观看。转变就在这里：在此前的每个平台上，他打造的都是供他人发挥创造力的画布。Eight years at Google and YouTube, then Spotify, CTO at WeWork, Commerce at Meta, OpenSea, Uber. Now he builds product behind 520 million subscribers and 1.6 billion views in 90 days. And here's the shift. At every platform before this, he built the canvas for other people's creativity.

在这里，产出本身就是产品。我们将探讨以下内容：他们为何弃用企业级 SaaS，并在 3 个月内打造自己的工具；前线部署工程师如何转而服务自己的团队；以及为什么现在一个想法就是一条提示词，而不再是待办事项，所以他们能够每天交付。Here the product is the output itself. Here are the things we'll cover. Why they threw out enterprise SaaS and built their own tooling in 3 months. Forward deployed engineers pointed inward at their own teams. shipping daily because an idea is now a prompt, not a backlog ticket.

创始人打造的产品往往是为自己设计的，而 AI 能解决这个问题。记忆层。这样一来，新员工就能继承过去 10 年的经验教训。我们来深入聊聊。Founders ship products designed for themselves and how AI fixes that. The memory layer. So, a new hire inherits 10 years of what went wrong. Let's get into it.

一种创意成果，而非一个平台A creative output rather than a platform

Carlos：嗯，嗯，感谢你过去这些年来做出的巨大贡献。我想我们是在你还在 Spotify 工作时认识的。我是说，你那时经历过这些公司，对吧？Spotify、WeWork，也许还有 Uber。Carlos: Well, well, thank you for being a big contributor over this last years. I think we met when you were Spotify. I mean, you had all the logos, right? Spotify, WeWork, and maybe I'm for Uber.

到目前为止，我可能已经忘了一些。Probably forgetting some at this point.

Shiva：是啊。没错。我爱上了很多东西。所以，就说到这儿吧。Shiva: Yeah. Yes. I've fallen in love with a lot of things. So, we'll leave it at that.

Carlos：不过现在，Right先生，我很想知道，这个高管职位与你在更传统的大型科技公司担任过的其他高管职位有什么不同。Carlos: But now, Mr. Right. So very curious to know how is this executive role different than the other executive roles you've had a more traditional big tech.

Shiva：是的。先说一下，我主要一直在做平台相关的工作。所以我的很多想法始终围绕着如何服务大量创作者；坦率地说，你会成为他人发挥创造力的画布，而不是创造力本身的源头。Shiva: Yeah. Let me start with you know I've largely worked on platforms. So a lot of my perspective has always been how do you serve a ton of creators and frankly you become a canvas for other people's creativity. You're not the source of that creativity yourself.

所以，这个职位的独特之处之一在于：我们要如何打造那些本身就是创意作品的产品？这可以是各种不同的东西，不过我们目前当然是在制作视频。所以，很多时候我们会想：哦，和一位已经在这方面做出出色成果、提出很棒想法的人合作时，你能做些什么？对吧？对方还想做更多。如何把这些创意产品化，才是真正的挑战。So, one of the unique things about this role is how do we build products that in and of themselves are creative productions and that could be a variety of different things, but of course we create videos right now. So, a lot of it is, ooh, what do you get to do when you're working with someone who's done fantastic, you know, great ideas on that, right? Wants to do more of it. How do you productize that is really the challenge.

但当你和那位创作者合作时，你能做哪些不同的事呢？你可以稍微更以那些目标为重，对吧？你可以深入了解工作流程，甚至改变工作流程。但在平台工作时，你就做不到这一点，对吧？But what do you get to do differently when you're working with that creator? You can be a little bit more selfish towards those objectives, right? You can start to get very deep in the workflow and maybe change the workflow. When you work at a platform, you don't get to do that, right?

这确实很难。所以你得认真思考。在我们的世界里，产品本身就是产出，而不只是一个促成产出的平台。这是一种截然不同的思维方式，我不想假装这很容易。我自己也在做很多调整。Like it's really hard. So you have to think and in our world, our product is the output, not just the platform that enables the output. So it's a very different way of thinking and I don't want to pretend like it's easy. I'm making a lot of adjustments too.

内部工作流程与定制软件Internal workflows and custom software

Carlos：那么，作为首席产品与技术官，你负责管理内部和外部产品。Carlos: So as a chief product and technology officer, you oversee both internal and external products.

Shiva：是啊。什么？再说一遍。内部的。Shiva: Yeah. What? Say it again. Internal

Carlos：内部和外部产品。Carlos: internal and external products.

湿婆：没错。没错。Shiva: Correct. Correct.

Carlos：那么，这些产品具体是什么？Carlos: And so what exactly are those products?

Shiva：是的。首先，我们在 Beast Industries 有一大批创意人才，他们每天都在创作内容。从编剧、剪辑师、布景设计师，到负责评估方案是否可行的人，各种人才都有。我们经常做一些全世界前所未有的事情。Shiva: Yeah. So first start with we have a ton of creative people at Beast Industries and they're creating content every day. And this is everything from writers to editors to set designers to people figuring out if that's feasible. We often do things that are first ever type of thing happening in the world.

那么，这在现实世界中到底能不能做到？这就变成了一个关于其物理可行性的问题。我们有这么多各有所长的人，他们学会了协同合作，为 MrBeast 制作出精彩的内容，每两周发布一期，但还有许多其他频道。So, can you even do it in the world? Like the physics of it becomes a question. So, we have all these specialized people who've learned to work together and produce a fantastic piece of content that drops every two weeks on Mr. Beast, but there's many more channels.

我们还有自己的工作室，正在加快拓展一些从 MrBeast 拆分出来的新频道。所以，其中一个工具要解决的问题是：我们该如何为所有这些人搭建工作流程，让大家能够快速推进，以最高质量完成制作，同时也能按时、按预算交付。过去，我想大家都知道，你们大概会直接买一款现成的企业级 SaaS 工具。绝无冒犯这个领域任何人的意思，但我们发现，这些项目管理工具在我们的领域里都不太够用。And we also have our studio that's ramping up new channels that are departed from Mr. Beast. So, one tool is how do we create the workflow for all of those people so we can move fast, produce at the highest quality, but stay on time and budget as well. And in the past, I think all of you know this, you'd probably get some enterprise SaaS tool off the shelf. no in no offense to anyone in that world but what we found is like every one of those project management tools it just falls short in our world.

所以我们当时的问题是：在 AI 时代，我们能不能干脆为自己打造完美的工具？这件事大约是三个月前开始的，现在你每天都能看到新东西上线。我们发现工作室有个问题。通常是 Jimmy 和工作室团队在想：“要是我们能做这个呢？”结果发现，现在用软件几乎什么都能做，而且不用担心成本，也很容易维护。于是，我们最终做的就是逐项审视这些任务，并围绕它们搭建一套操作系统。So our question was in the age of AI is can we actually just build the perfect tool for ourselves and that started about 3 months ago and now you're seeing a ship every day. So we see an issue with the studio. Usually it's Jimmy and the studio team is like what if we could do this and it turns out you can do just about anything in software right now and you don't have to worry about the cost of it and you can maintain it very easy as well. So what we've ended up doing is looking each of those tasks and building an operating system around it.

所以，作为一种内部工具，它令人兴奋的地方在于，你再也不必妥协了。比如，如果有人半夜在片场四处奔走、需要做些改动，他们只要对着手机说话，把内容录下来就行，而不必掏出笔记本电脑，也不用去找移动应用。过去，你没法轻易地创建这样的应用。你得跟供应商商量：“这在你们的路线图上吗？”他们会回来告诉你：“你们是唯一一家有人半夜四处奔走、而且只能用语音操作的客户，对吧？”So what's exciting about that as an internal tool is that you never have to compromise anymore. So if for example we have people running around a set at midnight trying to make changes, they can just talk to their phone and record that stuff rather than bust out a laptop or look for a mobile app. And in the past you couldn't just create that app easily. You'd have to negotiate with the vendor, is this on your road map? they'd come back and say, "You're the only customer that has people running around at midnight who can only use voice, right?

“所以，他们不会去做。”于是你就被困住了。但在当今世界，我认为这对产品经理来说是一个很大的机会。如果你假设任何事情都有可能，而且所有事情的成本几乎为零（并非完全为零，这一点我们可以讨论），那为什么不去做呢？所以，这就是我们对待内部工具的理念。对于外部工具也是如此。So, they're not going to do it." And so, you're stuck. But in today's world, I think this is a big opportunity for product managers. If you assume anything is possible and everything is almost zero cost, not entirely zero cost, which we can talk about, why not do it? So, that's the philosophy we're taking with internal tools. for external tools.

我不能透露太多，但我们正在推出一款面向消费者、广受公众关注的应用。令我们兴奋的是，想想看，在这款应用里，我们能做些什么，让现实生活中那些长期观看我们 YouTube 内容，或在 Instagram 和 TikTok 上观看我们制作内容的粉丝，能够更积极地参与进来？因此，我们正花时间探索和设计这方面的体验。你也可以想象，现在正值 AI 发展的巅峰时期，所以我们目前正在做的，就是借助 AI 快速制作原型、将想法变为现实，并迅速测试这些想法。I can't say too much about it, but we're launching a consumer app which is very public out there. And what we're excited about is okay, what can we do in that app that lets people in the real world who are our fans who've consumed YouTube content for a long time or content on Instagram and Tik Tok where we're also producing, what else can they do in this app that can be more participatory? So we're spending time defining that, designing it, and as you can imagine, this is like the peak AI period. So the ability to prototype, bring ideas to vision, test those ideas quickly is what we're working on now.

受众参与度与变现Audience engagement and monetization

Carlos：说到数字，据我上次查看，MrBeast 光是在 YouTube 上就有超过 5.2 亿订阅者，而仅在过去 90 天里，观看人数就超过 15 亿。Carlos: So speaking of numbers, last time I checked, Mr. Beast has over 520 million subscribers just on YouTube and over 1.5 billion viewers just in the last 90 days.

Shiva：是啊。1.6。对，没错。是啊，太疯狂了。是啊。Shiva: Yeah. 1.6. Yeah, exactly. Yeah, it's crazy. Yeah.

所以，这可不是件小事。这简直太疯狂了。So, it's not a small thing. It is it's pretty crazy.

Carlos：那么，我们在讨论注意力如何成为新的货币。我很好奇，你有什么愿景，能把这种注意力转化为一种商业模式，而不只是沿用我们所了解的 YouTube 盈利方式？Carlos: So, we're talking about how attention is the new currency and I'm curious to know what's your vision to convert that attention into a business beyond just what we know about how YouTube monetizes.

Shiva：其中一部分只是传统的变现方式。如果你拥有大量受众，同时身处一个依靠广告支持的平台，就可以通过深度赞助合作和广告来实现变现。我觉得大多数人都很清楚这一点。问题在于，当我们开始推出自己的直面消费者应用时，显然也可以尝试不同的方式，对吧？Shiva: Some of it is just the traditional way you monetize. So if you have a large audience when and you're sitting on an ad supported platform, you can just monetize that through deep sponsorships and advertising. So I think that's obvious for most people. The question is as we start to launch our own direct to consumer app, there's obviously different things we can play with there as well, right?

我们现在卖周边商品，对吧？我们也可以探索订阅之类的模式。但这些都只是次要的。眼下真正的问题是，我该如何吸引你，并真正打造出引人入胜的体验？如果这件事做起来了，你知道，我们就有了一个不错的支撑基础，可以探索这方面的各种尝试。我们会找到合适、自然且恰当融合的变现方式，但现在没必要非得把这件事放在心上。我觉得另一个优势在于，你知道，这个主题关乎我们能多快地发布和制作内容。但我认为，关键的一点是，我们也想测试这些变现途径，对吧？We sell merchandise today, right? We could explore things like subscriptions, etc. But all of that is secondary. Right now the real question is how can I engage you and really craft that engaging experience and if that takes off you know we've got a nice supportive foundation for exploring bets there we'll find the right organic way in a good blended way to monetize that but we don't necessarily have to obsess about it now I think the other advantage with you know the theme is like how fast can you ship and produce but I think one of the key things is we want to test those monetization vectors as well, right?

什么做法在这里行得通，什么做法在另一个市场行得通。我们是全球化的，你知道，我们的受众遍布各地。所以，在美国可能行得通的做法，在印度可能就大不相同。过去，你可能会把这些尝试分散到多年的开发过程中；我认为我们有机会在不同地区尝试不同做法，看看哪些能奏效，再把成功的做法推广开来，不仅是在产品核心体验上，也包括变现方面。What can work here, what can work in a different market. We are a global, you know, our audience is all over the place. So what might work in the United States might be very different than what works in India. In the past, again, you might spread that over years of development, I think we have a chance to try different things in different pockets, see what sticks and graduate the winners, not just on product core experiences, but on the monetization angle, too.

Jimmy 与日常生产流程Jimmy and the daily production process

卡洛斯：嗯，说到瓶颈，有一点我想到的是，MrBeast 品牌很大一部分其实就是吉米本人，对吧？所以，直接和他合作，而且他会参与很多细节，事实上还会出镜，这种合作体验怎么样？你又是怎么想办法确保他不会成为那个瓶颈的？Carlos: Well, one thing that comes to mind as we think about bottlenecks is that I mean a huge part of Mr. Beast brand is actually Jimmy, right? So when you're working directly with him and he's involved in a lot of the details and in fact he's on camera like how is it to work with him and like how do you find ways to also make sure that he doesn't become that bottleneck?

Shiva：问得好。首先，想想像 Jimmy 这样的人，他是一位企业家。多年来，他打造出了令人印象深刻的成果。这可不只是靠创造性思维就能做到的，对吧？Shiva: Good question. So first start off with when you think about someone like Jimmy, he's an entrepreneur. He's created something over several years now that is just impressive. And that doesn't come with just looking at creative thinking, right?

建立一个能够持续做到这一点的组织，培养这种能力，并制定相应策略来维持它。你也能从中看到很多产品思维，对吧？换句话说，哦，怎么讲，如果你们有人听过他讲话，那绝对值得一听。就是如何把创意上的成功拆解成分析性的配方？To build an org that can sustain that, to build a muscle and the tactics to sustain that. You see a lot of product thinking there too, right? In other words, oh what and you see if any of you have seen him talk, it's totally worth it. It's like how can you break down the creative successes into analytical recipes?

我们如何获得这种洞察？每天又如何审视这种洞察？就像你可能会在产品推出前进行产品评审，也可能会在产品发布后进行复盘。How do we get that insight? How are we looking at that insight every day? So just like you might launch a product, you might do a product review ahead of the product. You might do a retro after it ships.

每支视频都要经过同样的流程。所以，一方面，和这样的人合作其实并不难，因为他们骨子里就和我们有着相同的思维方式。另一方面，则是要在工作流程上展现出良好的判断力。世界上有些创意人士不太擅长审视创意背后的运作流程，也不擅长将其抽象出来，弄清哪些环节可以重复，以及可以用什么工具来实现。Every video goes through that same process. So one thing is like it's actually not hard to work with someone like that because they're thinking like we think right in their DNA. The other side of it is just showing good judgment on workflow. So there are people in the world creatively who aren't very good at looking at the operations behind that creativity and trying to abstract it and see what's repeatable and what tools can make it happen.

在我们的案例中，公司也对此非常执着。如果你回顾传统行业，看看一些创意工作流程是如何随着时间推移演变的，就会发现其中很多流程 50 年来几乎没有变化。如果你是在 YouTube 的环境中成长起来的，并且意识到这里的节奏和标准有多高，就会开始审视其中的每一个环节：你们如何让新人上手？In our case we have the company's obsessed around that as well. So if you look at historical industries and you look at how some of the creative workflows have basically evolved over time, they many of them haven't changed in 50 years. If you're native to YouTube and you realize the pace and the bar you need to have, you start looking at each one of those. How do you onboard people?

你如何培训他们？如何让人们尝试使用新工具？如何赋予他们超能力？如何去除那些琐碎乏味的工作，让他们专注于工作中那些神奇的部分？How do you train them? How do you have people experimenting with new tools? How do you give those people superpowers? How do you get rid of the mundane so they can focus on magical aspects of what they do?

这种执着非常好。公司里的人也都有这种本能，能够真正深入地理解这一点。其中一部分原因是，大家一直都在并肩协作。看到哪里出了问题时，你会问自己：这是个功能吗？That obsession is very good. And the instincts are there in the company to really understand that deeply. And part of that is just everyone running together all the time. And when you see something going wrong, you ask yourself, is that a feature?

如果是这样，那我们就把它做出来。我们团队有个奇怪之处：这件事会在当晚变成一场会议。大家会通宵进行一些氛围编程，做出原型，第二天就发布。不过，这样可能会让人筋疲力尽，除非轮换人员参与，或者安排一些身处不同时区的人来做这件事。If so, let's build it. Now, the weird thing about us is that will turn into a meeting that night. There'll be some vibe coding overnight. prototype what it is and it will ship the next day. Now that can be exhausting unless you rotate people into that or have some people working on different time zones who can do this.

不过，这种节奏加上对哪些做法有效、哪些无效的观察，把两者结合起来非常令人兴奋。我希望有更多公司能以这种方式开展工作。不过，这确实变得非常有意思了。But that pace combined with the observation of what's working or not. Bringing that together is very exciting. I hope many more companies are working like that. But it is it's become very interesting.

所以，换个角度来看这个问题——你也就此做过演讲——在我们公司内部，我们实际上在各个职能部门都配有前置部署工程师。有几个人可以驻在那里，为工作室制作原型、解决问题；我们也可以在营销部门这么做，在其他领域也一样。不过，已经有很多人和你的客户群谈起过 FDE。So one way to look at that and you give a talk on this too but internally within our company we effectively have forward-deployed engineers in the functions. So there's a couple that can sit there and prototype and solve problems for the studio and we can do that in marketing. We can do that in other areas as well. But it's been a lot of people have talked about FDEs with your client base.

我们可以想想内部部署的 FDE，这挺有意思的：他们能非常迅速地积累经验、获得洞察，同时又和自己支持的人朝夕相处。不同之处在于，他们可以很快地把某个想法做成原型，作为解决方案；而且做原型的成本很低，所以你可以尝试十种不同的方案，找出哪种有效，再把无效的方案丢掉。说实话，这种工作方式对我来说很新鲜，但我发现它让人很有自由感，也是我们都应该在各自公司里思考的一件事。It's sort of interesting for us to think about FDEs internally deployed and how can they pick up the speed and insights very quickly, but they live and breathe with the people they support. The difference is they can prototype something very quickly as a solution and because it's so cheap to prototype, you can try 10 different things, find which one works, and throw away the one that doesn't work. So that motion is very new to me actually, but I'm finding it to be very liberating and something we should all think about in our own companies.

与嵌入式工程师每日交付功能Shipping daily with embedded engineers

Carlos：我想再强调一下这一点，因为对我来说，每周发布一次都已经很快了，而你提到你们每天都在发布，对吧？那这到底是怎么做到的？Carlos: And I want to double down on that because even to me to ship on a weekly basis, it's already fast and you mention you're shipping on a daily basis, right? So how does that actually work?

Shiva：嗯，其中一点是，内部每天都在发布，这意味着你能非常及时地响应人们的需求，对吧？所以你会再次看到一个问题——我之前提过这个，我拿它来比较——我会一直拿企业 SaaS 举例，但我不是故意针对它；SaaS 也在不断发展。不过，关键的一点是，我记得整个职业生涯里都在和供应商打交道。如果你是供应商的重要客户，就会有季度 QBR 或评审会，你会告诉供应商，这是我们希望纳入路线图的事项。但太多时间都花在为这些需求辩解、讨论我们到底需要什么，以及为此陈述理由上了。在当今世界，我认为每天发布的重要性在于：嘿，有东西坏了，或者不好用了；要是有这个筛选器会怎样？要是有这种视图会怎样？要是这个视图跟甘特图完全不同，却更适合我们的团队使用，也更简单，这样他们扫一眼就能看明白，甚至可以在大屏幕上快速浏览，那会怎样？对吧？Shiva: Well, one is internally shipping on a daily basis just means you're being very responsive to people's needs, right? So you see an issue again I put this again I compare it I'll keep picking on enterprise SaaS but I don't mean to do it SaaS is evolving as well but one of the key things is I just remember throughout my career working with vendors and you'd have that quarterly QBR or review if you're a major client with your vendors here's things we want on the road map and too much of the time was spent on like justifying it and what do we need etc and making a case for it. In today's world, I think the importance of shipping every day is that hey, there's something broken or it's not working or what if we had this filter or what if we had this view? What if that view looked completely different than a Gantt chart but is much more usable for our teams and simpler so they can just glance on it and maybe glance at it on a big screen, right?

因为这就是我们工作的本质。我们可以直接把它发布出去。接下来就是发布、收集反馈。如果行不通，就把它砍掉。Because that's the nature of our work. We can just ship that. So what ends up happening is like ship it, get feedback. If it doesn't work, kill it.

不断迭代改进。换个方向尝试。这些现在都非常、非常便宜。所以，我觉得每天发布产品，可以这样理解：这相当于过去你把所有这些功能都列在一块板子上。Iterate on it. Take a different direction. That's all very, very cheap right now. So, shipping every day, I think the way to look at it is it's the equivalent of, you know, in the past, you take all these features and you put them on a board.

嗯，如果换个角度想，把那些功能都拿来用，你不是把它们放到看板上，之后再排优先级。你只是把那些想法转化成提示词，对吧？然后把提示词丢出去，看看会得到什么结果。Well, if instead you think about taking all those features, you're not putting them on a board and prioritizing them later. You're just taking those ideas and turning them into prompts, right? Then throw those prompts out there. See what spits back.

嘿，这还挺有意思的。我们再琢磨一下。好，我们把它提出来讨论吧。所以，你只是在压缩时间。Hey, that's kind of interesting. Let's iterate on it for a second. Okay, let's throw it out there. So, you're just compressing time.

这让产品面临的挑战更加重要。必须有人清楚你要解决什么问题。如果你同时推出 50 项功能，很快就可能让用户群感到困惑。所以，有时候需要把它们更好地打包整合起来。It makes the product challenge even more important. Someone has to know what problem you're solving. And if you have 50 things all going out there, you could confuse your user base very quickly. So, sometimes it's about packaging it up a little bit more.

但有时也会这样：嘿，这是为这个部门、这项技能以及一个非常具体的视图准备的。一共就10个。先发布出去，看看会发生什么。所以我们每天考虑发布时，都会权衡这些因素。But sometimes it's also, hey, this is for this department and this skill and a very specific view. It's 10 of them. Just get it out there and see what happens. So we just balance that as we think about shipping every day.

创始人、稀缺与探索Founders, scarcity and exploration

Carlos：不，你刚才这么说的时候，我就在想文化这件事。你曾亲身与一些传奇创始人共事，体验他们创造的文化，对吧？比如 WeWork 的创始人、Spotify 的创始人、Meta 的创始人，现在还有 MrBeast 的创始人。我觉得 Jimmy 也和他们是一个级别的。Carlos: No, as you were saying that, I was just thinking about culture and you've had firsthand experience working with some legendary founders and experiencing the culture they created, right? And thinking founder of WeWork, founder of Spotify, founder of Meta, now founder of Mr. Beast. I put Jimmy at that level.

那么，你在那里看到了什么，让这类文化如此特别？So what is it that you see there that makes those type of cultures so special?

Shiva：我觉得你看到的一个现象是，大家对此已经谈了很多，但我认为确实如此：驱动许多创始人的，和驱动优秀产品经理的是同一件事——我希望自己能更快地把产品推出去。我希望能试试五个想法，拥有更多选择。我希望能试试五个想法，而不是一个。我希望，你知道，我懂得怎么做某件事，而不必花钱买现成的；我可以自己把它做出来，并掌控它。Shiva: I think like one thing is you're seeing and this is there's been a lot of talk about this but I think it's true is what drives a lot of these founders is the same thing that drives a good product manager which is I wish I could ship faster. I wish I could try five ideas out optionality. I wish I could try five ideas out instead of one. I wish, you know, I knew how to do X, right, and not have to buy it, but I could build it and control it myself.

所以，你会发现很多这类创始人都认为事情本来就该是这样。我想说的是，在过去六个月里，他们终于都得到了自己梦寐以求的世界。但如果你从产品经理的角度来想，回想一下自己作为产品经理最艰难的时刻，你只能做一个选择。摆在你面前的选项一大堆，但由于工程资源有限、开发耗时，以及要把产品做得稳健安全所需付出的努力，你最终只能选一件事。So, what you see with a lot of those founders is that they believe that's just the way it should be. And I would argue in the last 6 months, they've all finally gotten the world they've wished for. But if you think about it as a product manager, if you've just think about your worst times as a product manager, you have to pick one choice. You have a huge set of options in front of you, but because of engineering scarcity, because of how long it takes, because of what it takes to make that robust and secure, you only get to pick one thing.

你知道，其实做三件事会更好。所以，优秀的创始人会在自己的组织里强行推动这件事，对吧？我们能不能做得更快？顺便说一句，这可能会非常烦人。And you know, it would have been better to do three things. So great founders sort of just force that, right, in their own organization. Can we do it faster? It can be very annoying, by the way.

感觉就像是在说，我们能不能做得更快？为什么不直接动手做呢？我现在什么都能靠氛围编程搞定，对吧？但如果你真的站在产品的角度考虑，我们已经努力扮演这个角色大概 25 年了，对吧？It's like, can we do it faster? Why can't we just do it? I can vibe code anything now, right? But if you really just put yourself in the product position, we've been trying to play that role for, you know, 25 years now, right?

就我而言，我当时就想：“天哪，这也太疯狂了。”所以我觉得现在我们确实得改变一下行为方式。比如我记得自己担任某些职位时，会想：“唉，我们得把 A、B 和 C 列为优先事项。”因此，和创始人的关系可能会很紧张，因为你得在一条路和五条路之间做取舍并为之辩解，毕竟资源确实有限。但在一个不再面临这种资源匮乏的世界里，你大概会问自己一个不同的问题。我们这里有三个想法。In my case, and I'm just like, "Oh my gosh, this is crazy." So I think right now what we do have to do is shift our behavior. Like I know like when I was in certain roles, I was like, "Ugh, we got to prioritize A, B, and C." So the founder relationship can be very tense because you're trying to justify one path versus five cuz you have real scarcity. But in a world where you don't have that scarcity anymore, you're probably asking yourself a different question. We've got three ideas here.

我们是否真的探索了所有可能的选项，还是三个想法就够了？这才是现在真正的问题。我们真的应该探索 10 个吗？我们能多快做到？指导这一过程的原则又是什么？Have we actually explored the full option space or is three ideas enough? That's really the question now. Should we really be exploring 10? And how fast can we do that? and what are our principles to guide that?

所以，你所扮演的角色会非常不同，而且再过一年，你甚至可能有机会探索更多选择。现在你面临的是另一个问题：判断力。其中一个问题是，你得看看创始人，判断他们的优势在哪里、盲点又在哪里。如果你是与这类人共事的产品负责人，就需要问问自己：当他们专注于发挥优势时，我该如何弥补他们的盲点？我共事过、或哪怕只是间接接触过的每一位创始人，都有很突出的优势。So it's a very different role you're playing and again a year from now you might even be able to explore more options. Now you've got a different issue which is judgment. So one of the questions is you have to look at your founder and say where are their strengths but where are their blind spots and if you're a product leader working with those types of people you need to ask yourself how do I fill the blind spots right while they focus on their strengths. And with every founder I've worked with or been close to even indirect, they all have major strengths.

它们也存在很大的盲区。大多数创始人最大的盲区之一，就是他们常常是为自己设计产品，而不是为更广泛的用户群体设计。所以我觉得，现在借助 AI，你可以这样做：我们来做这件事，但根据我对用户需求的理解，打造两个不同的版本。这样一来，三种选择就都有了。They also have major blind spots. And one of the biggest blind spots from most founders is they often design a product for themselves, not for their broader set of users. So what you can do now with I think with AI, it's like let's do that, but let's create two varieties based on what I understand about the user needs. And now we got all three.

而你不必做出那个选择。但过去，你不得不做出那个选择。猜猜在那个选择中谁会胜出？你们创始人的选择。And you don't have to make that choice. But in the past, you'd have to make that choice. And guess who wins in that choice? Your founders's choice.

对吧？而且，太多产品——我待过的每家公司都这样——推出时都是为创始人设计的，而不是为用户设计的。也正因如此，它们失败了。所以我觉得，这种情况也正在改变。Right? And too often so many products, and this is at every place I've worked, have shipped and launched a product that is designed for the founder, not for the users. And it failed for that reason. So I think that's what's changing, too.

说得有点多了。抱歉，我有点跑题，不过你明白我的意思。AI 的力量本质上赋予你一种探索的超能力，而这种能力牢牢掌握在产品团队和其他相关人员手中，对吧？And that's a lot. Sorry, I'm rambling, but you get the idea. It's it's the nature what the force of AI gives you is this superpower of exploration that is firmly in the hands of the product team. And others involved, right?

“产品”这个说法。现在如何利用它来取得更好的结果，确实是我们接下来一年左右需要弄清楚的事情。速度很重要，但选择空间的广度也很重要，而且还要认真思考各种选项；这也必须和速度结合起来，否则你可能只是很快地走上错误的道路，对吧？这就是我们需要弄清楚的部分。Product in quotes. How do you use that now to get a better result is really up to what we need to figure out over the next, you know, year. speed wins but also breadth of option space and really thinking through that also it has to go with speed otherwise you might just go down the wrong path very fast right so that's the part we need to figure out

互联工作流与机构记忆Connected workflows and institutional memory

Carlos：我想结合一些例子来聊聊，因为我记得，甚至在 AI 成为主流之前，Jimmy 就在说自己有多痴迷于尝试不同的 YouTube 缩略图、标题、图片和面部表情。这真的很不可思议。四年过去了，如今你有了这些工具，那么你具体是如何在流程中的某些环节运用 AI，来大幅加快这种探索的？Carlos: I want to explore that with some examples because I remember even before AI was mainstream Jimmy was talking about how he was obsessed over experimenting with the YouTube thumbnails and the title and the image different facial expressions something incredible So four years forward with now the tools at your disposal like how are you actually applying AI in specific parts of the process to turbocharge that exploration.

Shiva：我们使用 AI 的主要原因，大体上是要确保整个流程都衔接起来，对吧？所以你可以把它主要看作一个工作流，对吧？也就是说，你有一个创意团队，擅长你提到的所有事情，对吧？比如标题是什么、缩略图是什么，等等。Shiva: The main reason we use AI is largely that we need to just make sure that process is all connected, right? So you could think of it largely as workflow, right? Which is hey, you have an ideation team which is very good at all the things you talked about, right? Like what is the title, what is the thumbnail, etc.

然后，好吧，接下来我需要为此构思一大堆故事情节之类的。然后还得为它设计场景之类的。所以这里会发生一系列事情，而它们之间一直缺乏衔接。把这些都放进一个系统里，你就能看出，嘿，延误的原因是什么？And then okay, now I need to come up with a bunch of story lines for this, etc. And then I've got to design the sets for that, etc. So there's a sequence of things that happen here and it's been very disconnected. So putting that in one system so you can see hey what's the source of a delay?

我怎么才能了解到自己设计的某个特定套件？如果之后有一整个新团队要负责某个套件呢？除非大家都碰面，否则这些组织里就没有记忆层。如果你们推进得非常快，这就会成为一个问题。How do I learn that a particular set that I designed? What if a whole new team is working on a set later? There's no memory layer in these organizations unless you all meet. And if you're moving very fast that's a problem.

那为什么不把这一层记忆放到 AI 这边呢？比如说，嘿，我知道你正想布置这个场景，需要一个巨大的桶，然后我们会往里面装些东西，让它看起来像某种东西。接着告诉它：嘿，这里有三件事你得记住。顺便说一句，如果你做了这件事，那就行不通。So why not put that memory layer in the AI side of things? Hey, I know you're trying to do this set and it involves like a huge tub and we're going to fill it with something to make it look like whatever. Then, hey, here are the three things you need to remember. By the way, if you do this one thing, it's not going to work.

这应该是即时完成的。那么，有一种看待它的方式，就是想象有100个人整天四处奔走，不断遇到问题，也不断找到解决方案。只要把这些信息提供出来，下次你输入一个想法，它就会给你反馈，告诉你需要注意什么。这是一种非常具体的做法。我觉得这是最有意思的地方。That should just be instantaneous. Now, so one way to look at it is just saying there are 100 people running around all day running into problems, running into solutions. just making that available so the next time you pop in an idea it comes back and tells you what to watch out for. So that's a very concrete way of doing it. I think that's the most interesting thing.

当然，现在世界各地的人都在使用各种工具来开展创意工作流等。对我们来说，更有意思的是，如何把那些可能只存在于某个人脑海中的知识分享给所有人。这样，每个视频、每项制作中可能出错的地方，就不再令人措手不及了。如果你第一天加入团队，就应该能了解过去十年来所有参与过这项工作的人积累的经验，而且这些经验都应该对你开放。之后，我们在不断学习的过程中，也应该持续完善这些经验。Of course people are using you know tools everywhere in the world right now you know to also do creative workflows etc. What is more interesting for us is how do we take the knowledge that often might be sitting in one person's head and how do we now make that knowledge available to everyone. So every video right every production gets all the errors that can make it go wrong are no longer surprises. If you join the team on day one, you should have the history of everyone who's worked on this for 10 years and that should be available to you and then we should refine that history as we learn things too.

因此，AI 最有意思的一点，就是如何将对所有内容的强制采集制度化。这样，所有知识要点都会汇集到同一个地方，然后大家可以投票赞成或反对，判断这些知识是否适用于所有场景，还是只适用于某个特定场景。建立起这个基础后，你就会看到这些智慧带来的回报呈指数级增长。在我看来，这是我们正在做的事情中最重要的一点。So what's been very interesting with AI is how you should institutionalize that force capture of everything. So all those nuggets of knowledge are put into one place and then you're kind of voting up or down on whether that knowledge makes sense across everything or is localized to something. Once that base is built then you can you'll see like exponential returns on that wisdom. That's the most important aspect of what we're doing in my opinion.

结语Closing

卡洛斯：我们就到这里了。非常感谢你参与其中。好的。Carlos: We're going to leave you with that. Thank you so much for being part of it. Okay.

Shiva：谢谢。Shiva: Thank you.