What it takes to be a top PM today | Robby Stein (Google Search)
频道: Lenny's Podcast
视频: https://www.youtube.com/watch?v=sTgM_sbLMNg
原文语言: en
统计: 共 16 轮 · Robby 15
[0:06] Robby
Everyone, it's so great to be here. And I love being here to get to talk about one of my favorite things ever. Being a PM, product manager, and builder. And what's pretty incredible is that it's never been so interesting to be building products. And after 20 years almost of building, particularly in consumer journeys, it's very upsetting to be able to say out loud. Um but after almost 20 years of building, been fortunate to build all kinds of different products over my career. Built startups, and I've also helped lead teams and products through really big changes like Instagram and Google Search, where now billions of users are using products that our teams have created over the years. Also, I've built some things that zero people use today, by the way. Absolutely zero. And what I've learned throughout this process is that there is a little bit of a pattern of the kinds of things that we've done to make the products really great. Sometimes it works out, sometimes it doesn't, but there has been a repeatable pattern that I've seen emerge from the things that have done really well and the things that have not. And that's really what I'm here to share today uh and impart to all of you. There's lots of things to talk about, but I wanted to give you just a little bit of extra background on me. So, I spent some years at Instagram, where my team uh has built out Instagram Stories and shipped it originally, and built all of the experiences around it. We have Reels, uh feed ranking, and direct messaging. So, a whole bunch of products that came out of that era. Uh and then now at Google Search, where I am working through a brand new uh experience with the team to help think through what is Google Search in the AI era, so that you can truly ask anything on your mind and get the most helpful information imaginable across everything, whether it's videos or audio, text. And you can tap into all of the knowledge of the of the web and of
大家好，能来这儿真是太好了。我特别高兴能在这里聊一件我最爱的事：当 PM，也就是产品经理，当一个 builder（亲手把东西做出来的人）。很不可思议的是，做产品从来没有像现在这么有意思过。我做了将近 20 年产品，尤其是消费级这一块——这个数字大声说出来，还挺让人难受的。嗯，不过这将近 20 年里，我很幸运，职业生涯里做过各种各样的产品。创过业，也带着团队和产品经历过特别大的变革，比如在 Instagram 和 Google Search，如今有几十亿用户在用我们团队这些年做出来的产品。顺便说一句，我也做过一些今天一个用户都没有的东西。真的是零。这一路下来我学到的是：我们把产品做到真正出色时所做的那些事，其实是有那么点规律可循的。有时候成了，有时候没成，但从那些做得特别好的和没做成的东西里，我看到了一个可以重复的模式浮现出来。这就是我今天来这儿想分享、想交给各位的东西。要讲的很多，不过我想先多交代一点我自己的背景。我在 Instagram 待过几年，我的团队在那里把 Instagram Stories 做了出来、完成了最初的发布，还做了围绕它的所有体验。我们还有 Reels、信息流排序（feed ranking）和私信。所以那个时期出了一大堆产品。然后现在在 Google Search，我正和团队一起打磨一套全新的体验，一起想清楚 AI 时代的 Google Search 到底是什么——让你真的可以把心里想到的任何事拿来问，拿到你能想象的最有帮助的信息，覆盖所有形态，不管是视频、音频还是文字。而且你可以调用所有的知识——来自整个网络的，以及来自
[1:54] Robby
Google and connect the information that's out there. But, today I'm not talking about the transformation of product, although that is something I really deeply enjoy. I'm talking about the craft of product. What does it take to build? What does it take to have a product that people love and to shape that and make that? And how do you do that in the age of AI that ultimately, and from my opinion, any one of you can do now with just a handful of agents or kind of AI assistants which took us years and whole teams of people to do um in process here. Let's uh just mention one thing I wanted to say around the the role of PM before we get started. And that is really the the fundamental shift that we're all feeling is that we used to be, I think as PMs, a big part of our value was around getting stuff done. Was around organizing, creating energy and momentum, and being almost like a project manager and a product manager. And I think in some cases that's still true, but really if in the world of today, almost anything you can think of can get built, it's more clear to me than ever that the actual true value of PM is actually around judging. It's around taste. It's around doing something extremely well.
Google 的，并把外面那些信息串联起来。不过今天我不讲产品的转型，虽然那是我打心底里很享受的事。我要讲的是做产品的手艺。做产品到底需要什么？要做出一个大家真心喜欢的产品，去塑造它、把它造出来，需要什么？在 AI 时代又该怎么做——说到底，在我看来，这些事现在你们任何一个人只靠几个 agent、或者说 AI 助手就能做到，而我们当年要花好几年、动用一整个一整个团队，才能把这套流程走下来。正式开始之前，我想先就 PM 这个角色说一点。那就是，我们所有人都感受到的那个根本性转变：过去，我觉得作为 PM，我们很大一部分价值在于把事情推到做完。在于组织协调，制造能量和势头，差不多既是项目经理又是产品经理。我觉得在某些情况下这依然成立，但说真的，如果在今天这个世界，几乎任何你想得到的东西都能做出来，那我比以往任何时候都更清楚：PM 真正的价值其实在于判断。在于品味。在于把一件事做到极致。
[3:03] Robby
And I think what's something I've thought about a lot is that people say the PM's main craft is that you do lots of things pretty well. And I think that it is somewhat true, but I think above all else actually the craft of PM is one simple thing that you should be the absolute best at. And that is decision making. It's about making decisions and you should be students of what it takes to make great decisions. And it's a humbling thing cuz most decisions, many decisions are wrong, and these are just the things that I've learned over the years that have helped me and the teams make decisions that have led to some of the great experiences that are now out there today. There's kind of a book that's been unwritten uh in my mind that I've been kind of thinking about um three chapters around the process that, you know, I've used and teams have used to build and I'm here to share that today really for the first time in this format. So, chapter one is understanding people deeply. It all starts with people, obviously. And you can figure out lots of things that you can build, but it comes back to the very root of what are the fundamental needs of a person. And I love the book um Clayton Christensen wrote Competing Against Luck. I'm sure many of you have read it and are probably read the Jobs to Be Done framework, which I really do love.
有件事我想了很多：大家都说 PM 的主要手艺是很多事都做得还不错。我觉得这话有几分道理，但我认为归根结底，PM 的手艺其实就是一件简单的事，而你应该在这件事上做到绝对的最好。那就是做决策。核心就是做决定，你们应该当个学生，去钻研做出好决策到底需要什么。这是件让人谦卑的事，因为大多数决策——很多决策都是错的。下面这些只是我这些年学到的东西，它们帮助我和团队做出了一些决策，最终带来了今天大家能用到的一些很棒的体验。我脑子里一直有一本没写出来的书，嗯，我一直在琢磨，分三章，讲的是我自己和团队用来做产品的那套流程。今天我来这儿分享它，真的是第一次用这种形式讲。第一章，深刻地理解人。一切都从人开始，这是显然的。你可以想出很多能做的东西，但最后都要回到最根本的一点：一个人的根本需求是什么。我很喜欢 Clayton Christensen 写的那本书，《Competing Against Luck》（《与运气竞争》）。我相信在座很多人都读过，大概也了解过 Jobs to Be Done 框架（待办任务理论，JTBD），我是真的很喜欢。
[4:14] Robby
The thought is that we don't use products, we hire them to do things for us, and we have to know what those things are. And it's best articulated how to extract those kinds of insights through a story. And so, actually um a team that we had uh at Instagram, we did a offsite with the people who created the Jobs to Be Done framework, and we actually went through this process. So, you can all think about this with me. What was the last major purchase that you made? So, quickly think about it. And for me, it was a bed. So, here are a bunch of beds. And then what you're supposed to do, and the book actually talks about how to do this, we've really internalized this with our process, is to put yourself back into that moment. And if you're talking to a person, ask them details about their life. What was it like that day? Where were you? What were you reading? Who were you with? What were you wearing? Were you with your kids? Were you by yourself? Were you in a store? Were you online? And you really deeply put yourself in the position of that user. And for me, I'm looking at this sea of beds with my wife. And basically, we were kind of trying different things out, and at the end of the day, three or four felt great, but my wife weirdly turned to me and said, "Can you do me a favor and jump on this?" Basically. Uh she's like, "Do a barrel roll." I was like, "What?"
它的想法是：我们不是在「使用」产品，而是「雇用」产品来替我们办事，而我们必须知道那些事到底是什么。至于怎么把这类洞察挖出来，用一个故事来讲最清楚。其实，嗯，我们在 Instagram 的一个团队，跟创造 Jobs to Be Done 框架的那批人一起做过一次外出研讨（offsite），我们真的把这套流程走了一遍。所以各位可以跟我一起想：你上一次大件消费买的是什么？快速想一下。对我来说，是一张床。你们看，这是一堆床。接下来你该做的——书里其实讲了怎么做，我们也真的把它内化进了自己的流程——是把自己放回那个时刻。如果你是在跟一个人聊，就问对方生活里的细节。那天是什么样的？你在哪儿？在读什么？跟谁在一起？穿的什么？带着孩子吗？一个人吗？是在店里，还是在网上？你要真的、深深地把自己放到那个用户的位置上。拿我来说，我和我太太正对着一大片床。基本上我们就是这张试试、那张试试，到最后有三四张感觉都很好，但我太太很奇怪地转过来对我说：「你能帮我个忙，往这上面跳一下吗？」大意是这样。她说：「你上去滚一圈。」我说：「啥？」
[5:21] Robby
Uh all right. And so, I did like a huge flop, I guess, and she was like, "We're getting this bed right now. Thank you. We're done." And I was like, "What? This is so weird." She's like, "No, the main thing that I need is a bed that will not wake me up if you like move." And if you think about it, that's like a little offensive, but also I was like, "Okay, it's real." And if I was a PM for beds, then I might be like classically thinking, "Quick, what's the road map for beds?" Right? You might be thinking, "Okay, uh how about cooling? Everyone's doing cooling." Like, yeah, cooling, sure. Or like eco uh or just pure price. Like, those things do matter, but for us, the thing that made the decision on that day is a concept of transference. And if you didn't talk to me and you didn't learn that, the PM, the team, they would never know that. And so, that's what it means to discover the job to be done for someone and how products, things, customers hire what your services do, your products do to help them. And very quickly, just talking two quick examples in the digital world, one from Google. Early in our AI experiences, we realized that people actually came for inspirational journeys. But 2 years ago, chatbots were only text. So, we were one of the first teams to look at this need and say, "How do you get inspired by what to buy or or what kind of uh couch to get through text? How do you describe a couch to someone?" You can't.
呃，好吧。于是我就，算是整个人重重地往上一扑吧，然后她说：「这张床我们现在就买。谢谢。搞定了。」我说：「啥？这也太奇怪了。」她说：「不是，我最需要的就是一张你一动也不会把我弄醒的床。」你细想一下，这话有点冒犯人，但我也觉得：「行吧，这是真需求。」假如我是个做床的 PM，我大概会按老套路去想：「快，床的路线图是什么？」对吧？你可能会想：「好，嗯，做凉感怎么样？大家都在做凉感。」行啊，凉感，当然可以。或者环保，嗯，或者纯拼价格。这些确实重要，但对我们来说，那天真正决定买哪张的，是「动静传导」（transference）这个概念。如果你没跟我聊过、没了解到这一点，那 PM、那个团队就永远不会知道。所以，这就是替一个人找出他的「待办任务」的含义，也就是产品、东西——顾客是怎么「雇用」你的服务、你的产品所做的事来帮自己的。再很快讲两个数字世界里的小例子，一个来自 Google。在我们做 AI 体验的早期，我们发现大家其实是为了找灵感的那类旅程而来的。但 2 年前，聊天机器人只有文字。所以我们是最早一批盯着这个需求发问的团队之一：「你怎么靠文字来获得该买什么、或者，嗯，该买哪种沙发的灵感？你怎么向别人描述一张沙发？」描述不了。
[6:35] Robby
And so, we were one of the earliest teams to really invest in state-of-the-art multimodal understanding, retrieval, and knowledge systems, so you could actually ask about a couch, get an incredible visual response, and do multi-turn conversation with it to say, "Oh, I actually want emerald green or I want fuzzy." Like, what does fuzzy mean? Or like that throw pillow, how does it create attention on the throw pillow? And do a turn and then now do a grid of throw pillows. Like, these were really hard things many years ago, and obviously you see them now in many places, but that was a core need and a core insight that we saw. And then last example was very early in our AI mode exploration, which is a way that you can use Google Search now to ask anything and have a conversation with Google, was people came and it had all this incredible information, like, here this is a neighborhood and like a food question, but it was missing something. It was missing something that people were coming to Google for and were expecting, which was the knowledge and the accuracy and the experience of actually seeing the places you go, seeing them on the map, getting the Google level authority of the star rating, the closing time, what kind of food it is, how expensive it is. It's completely missing from the product. And when we brought that together, so that now you can talk to Google, ask about food, and then get this rich experience that brings all the goodness of seeing the food and clicking on an entity now and actually seeing the description that you can trust, that was the magical product. And it's still one of the most popular things on the product today based on thumbs up ratings and other things that we follow. But again, back to a human inside of the why behind they were using the experience and what caused them to use it in the first place.
所以我们是最早一批真正投入做最前沿的多模态理解、检索和知识系统的团队之一，这样你才真的可以问一张沙发，拿到一个惊艳的视觉化回答，并且跟它多轮对话，说：「哦，我其实想要祖母绿的」或者「我想要毛茸茸的」。那「毛茸茸」到底指什么？或者那个抱枕，它怎么把注意力落到那个抱枕上？然后来一轮对话，接着再给我出一整面抱枕的网格。这些在很多年前都是非常难的事，当然现在你在很多地方都能看到了，但那是我们当时看到的一个核心需求、一个核心洞察。最后一个例子发生在我们探索 AI Mode 的很早期——AI Mode 是现在用 Google Search 的一种方式，你可以问任何问题、跟 Google 对话——当时大家来了，它里面有各种很棒的信息，比如，这是某个街区，再比如一个关于吃的问题，但它缺了点东西。缺的正是大家来 Google 想要的、预期会有的东西：那些知识、那份准确，以及真正看到你要去的地方的体验——在地图上看到它们，拿到 Google 级别权威的星级评分、打烊时间、是什么菜系、贵不贵。这些在产品里完全没有。等我们把这些合到一起，现在你可以跟 Google 对话、问吃的，然后得到这样一个丰富的体验，把所有好东西都带进来：看到食物，点开一个实体，真的看到一段你信得过的描述——那才是那个有魔力的产品。按点赞评分和我们跟踪的其他指标看，它到今天仍是这个产品上最受欢迎的东西之一。但还是那句话，要回到对人的洞察：他们用这个体验背后的「为什么」，以及当初到底是什么让他们来用的。
[8:07] Robby
And so AI can actually change a lot of this and a lot of the talk today is how AI can apply to these. I think on this one, I'm still thinking about some of these uh myself, but one of the things I'm excited to try more is how you can actually take the transcript data of these user interviews and actually have it specifically sourced jobs to be done and scale that much more. Another idea I've been thinking about um is actually an agent that can interview people with the exact methodology that I just described so that it can really ask a user more at scale to talk about why they're using the product they're using and what those underlying problems were. So, that is understanding people. Next up, you have a seed. You have something you're working towards. The next step, diagnosing root causes. So, the way I think about it is you have this initial vision of this problem you're solving for people, this initial understanding of what people want. And then you make a thing. And that thing is usually not great. And most of the things that I've ever been a part of building started not great. Everything did, really. Um and what's incredible about this process is by having a clear analytical view of the exact problems that are in your product space, ranking them in priority and understanding them deeply, and then actually fixing them and then asking it recursively again and really looping through, you basically are doing your own training process, kind of like an epoch of model training a little bit.
而 AI 其实能改变这里面的很多做法，今天这场演讲很大一部分就是讲 AI 怎么用到这些环节上。在这一条上，有些我自己也还在想，但我很想多试试的一件事是：把这些用户访谈的转录数据拿过来，让 AI 专门从里面提炼出「待办任务」（jobs to be done），并且把这件事的规模放大很多。我一直在想的另一个点子，嗯，是做一个 agent，它能用我刚才讲的那套方法原样去访谈用户，这样就能真正大规模地去问用户，请他们讲讲为什么在用手上这个产品、背后那些深层问题是什么。好，这就是理解人。接下来，你手里有了一颗种子。你有了一个要努力去实现的东西。下一步，诊断根因。我是这么看的：你对自己要替人解决的这个问题有了一个最初的愿景，对人们想要什么有了一个最初的理解。然后你做出一个东西。而这个东西通常不怎么样。我参与做过的东西，大多数一开始都不怎么样。说实话，全都是。嗯，这个过程不可思议的地方在于：对你产品空间里到底有哪些问题，有一个清晰的、分析式的认识，按优先级给它们排序、把它们吃透，然后真的去修掉，接着再递归地问一遍、一圈一圈真的循环下去——你基本上就是在跑一套你自己的训练过程，有点像模型训练的一个 epoch。
[9:26] Robby
And each update you do with this change list gets you closer to the market fit you want until you have that fit, and then it works. There's two stories I wanted to kind of articulate here, both uh from Instagram. Uh one was when we originally launched stories it actually did take off. It was one of the earliest things that worked pretty well out of the box, but we had many internal versions where stories were fit into the feed that didn't work great. So there was a ton of headache to get to the the final version. But one thing that was particularly clear was that not everyone was adopting it or felt comfortable sharing. And really the whole premise was hey this is an opportunity to share your life throughout the day. Why aren't you doing this? And the process that we followed was pretty simple. We asked a very clear root cause of why you were not able to do the thing that was our goal. So it's kind of the inverse. The first one's like why did you do the thing? Now you have a thing and it's like well why aren't you doing the thing? And it turns out that if you ask that question people answered think of all the leaf nodes like a cloud behind me of all of the answers people had. And then if you were to group them by theme it turned out that the top one when you sorted them was something around audience problems more or less. Like my ex is on it. My teacher is on it. My aunt is on it. My sister is on it. I'm not going to like just put a goofy thing on like people are on it. And so what we realized in seeing that and then what we did was we then did a very large quantity survey thousands of people taking that as an input and then we were able to quantify the actual weights across them. And what we figured out was that if we couldn't overcome that burden there's no creative tool there's no like interesting product idea that would work. And so the team actually spent two years cycling through different ideas in the space until
你拿着这份改动清单每更新一轮，就离你想要的市场契合更近一步，直到真正契合上，然后产品就成了。这里我想讲两个故事，都来自 Instagram。第一个是，我们最初上线 Stories 的时候，它其实确实起飞了，算是最早那批一上来就跑得不错的东西之一。但在那之前，我们内部做过很多版把 Stories 塞进 feed 流里的方案，效果都不怎么样，所以走到最终版本之前，是吃了一大堆苦头的。不过有一点特别明显：并不是所有人都在用它，也不是所有人都能放心地往上发。而它整个出发点本来就是——嘿，这是个让你把一天里的生活随时分享出来的机会，你怎么不发呢？我们走的流程其实很简单：针对“你为什么没能做到我们希望你做的那件事”，去问一个非常明确的根因。所以它算是反过来的。第一种问法是“你为什么做了这件事？”；现在东西已经有了，问的就变成“那你为什么不做这件事？”结果发现，你把这个问题问出去，大家的回答——你可以把所有叶子节点想象成我身后的一团云，全是大家给出的各种答案。然后你要是把它们按主题归类，排完序会发现，排在第一的差不多就是受众方面的问题。比如：我前任在上面，我老师在上面，我阿姨在上面，我姐在上面。我才不会随手发个犯傻的东西上去，人都在上面看着呢。看到这些之后我们意识到了这一点，接下来我们就把它当作输入，做了一轮规模非常大的定量问卷调研，有几千人参与，这样就能把各项原因实际占的权重量化出来。我们得出的结论是：这道坎要是迈不过去，什么创作工具、什么有意思的产品点子，都不会管用。所以团队实际上花了两年时间，在这个方向上把各种点子轮番试了个遍，直到
[11:00] Robby
we came across close friends and then it ended up working. And there were all kinds of weird permutations in this like multiverse where like it's called favorites and it also has on your profile a back door that you can like see a like a private version of your profile and you can also post to your feed under that kind of close friends label. Super confusing no one understood it. Like why would I put this like very raw photo in the middle of like a glitzy feed of all this stuff that's happening. The only thing that worked was within stories. So we cut out the entire product relaunched it as a stories experience and it worked. Reels was the second example. We had a We had a a launch for Reels in Brazil and we thought at the time that if you're doing goofy things and dances and posting them, why would you want them to live on your profile for everyone to see forever, which made sense and the team felt that way. So, we launched it, huge failure. So, it turns out that if you invest a bunch of time doing an awesome dance, you don't want it to go away in a day, which was an ephemeral format. It's very obvious in hindsight. But, the the bigger insight was the people who were posting them, even if they were young, you know, people, they wanted to be entertainers. They wanted to make businesses out of their content and they wanted it to live on and go viral. And so, we did a massive update. We talked about this change list, where now it's its own format. Instead of being ephemeral, it lives like forever and it can be something your whole school sees instead of the opposite, like, oh, I don't want my whole school to see it.
我们摸到了 Close Friends，然后它最终跑通了。这中间出过各种奇奇怪怪的排列组合，简直像个多元宇宙：比如有一版叫 Favorites，在你的个人主页上还开了个后门，能看到一个私密版的个人主页，你还可以打着类似 Close Friends 这样的标签往 feed 流里发帖。特别让人犯晕，没人看得懂。大家会想：我干嘛要把这么一张特别原生态的照片，夹在一个光鲜亮丽、什么热闹都有的 feed 流中间？唯一行得通的，是放在 Stories 里面。于是我们把整个产品砍掉，重新作为 Stories 里的一种体验上线，结果就成了。Reels 是第二个例子。我们当时在巴西上线了 Reels，那会儿我们的想法是：你要是在做些搞怪的事、跳跳舞然后发出来，干嘛会想让它们一直挂在你的个人主页上、永远让所有人看见呢？这听着挺有道理，团队也是这么觉得的。于是我们就这么上线了，结果一败涂地。事实证明，你要是花了一大堆时间跳出一段很棒的舞，是不会希望它一天之后就没了的——而当时它恰恰是个限时消失的形态。事后看这再明显不过。不过更大的洞察是：发这些内容的人，哪怕是些年轻人，他们想当的是艺人。他们想靠自己的内容做成生意，想让内容一直留着、还能爆火。所以我们做了一次大改。我们刚才讲到的改动清单，就是这个：现在它成了一个独立的形态。它不再限时消失，而是差不多永久保留，而且可以是你全校都能看到的东西，而不是反过来那种“哎呀，我可不想让全校都看到”。
[12:20] Robby
It's like, no, that's actually what you want. You want everyone to get the joy from this product. And then, obviously, that take has taken off and done quite well over the years. So, two really clear examples um of how that's happened. And with AI, this is actually possible to do a lot easier now using internal tools and I think things that you can all do every day. And we use, you know, internally Google anti-gravity, but we have a system that effectively lets users that have opted into this, they want to talk to us more and they've decided that they're they're comfortable sharing as they're using Google products with us. And so, what will happen on the side is we will then ask them, "Hey, what did you think of this? Did it Where did it miss the mark?" And they will put comments in. And so, we imagine this huge library of feedback from people now that's available. Well, now, the model can take that and can both take qualitative feedback like above. This is not This is a way a shopping example for backpacks. It's like, it's awesome that you knew about each backpack and it's awesome that you knew which was larger and the sizing cuz Google has all this canonical information, but you should have asked me about my son and the height. Like, that's a pretty key thing or it's going to be too heavy for my kid. And it was so obvious that when you stack rank them, this was a very common kind of feedback. so this concept of being collaborative and and creating a conversation so you can actually help the the person make progress with to the ultimate objective. They don't actually know all the things that they need to know to answer it was absolutely critical to getting the user to vote and say this was awesome. And so we invested a ton of energy in this and it's been one of the biggest things we've done in terms of engagement, in terms of usage, in terms of value and helpfulness, you know, that we've seen out of the
其实是：不，这恰恰就是你想要的。你就是想让每个人都从这个产品里得到快乐。然后，很显然，这个路子后来起飞了，这些年表现相当不错。所以这是两个非常清楚的例子，说明这件事是怎么发生的。而有了 AI，现在做这件事其实容易多了，用内部工具就行，而且我觉得是你们每个人每天都能做的事。我们内部用的是 Google 的 Antigravity，不过我们还有一套系统，实际上是面向那些主动选择加入的用户——他们愿意多跟我们聊聊，也确定自己可以放心地在使用 Google 产品的过程中把情况分享给我们。于是在一旁会发生这么一件事：我们会问他们，“嘿，你觉得这个怎么样？它哪里没答到点子上？”他们就会把评论写进来。所以你想象一下，现在手头有这么一个巨大的用户反馈库。好，现在模型可以把它拿过去，既能吃进像上面这样的定性反馈——这是一个买背包的购物例子。用户说的是：你对每款背包都了解，这很棒；你知道哪个更大、尺码是多少，这也很棒，毕竟 Google 有这么多权威信息；但你本该问问我儿子的情况、问问身高的。这可是个挺关键的事，不然对我家孩子来说会太重。而且特别明显的是，你把这些反馈排个序，会发现这是非常常见的一类。所以“协作”这个理念，也就是把它做成一场对话，这样你才能真正帮这个人朝着最终目标往前推进——他们其实并不知道回答这个问题需要了解的所有东西——这对于让用户投上一票、说“这太棒了”是绝对关键的。所以我们在这上面投入了大量精力，从参与度、从使用量、从价值和有用程度来看，这都是我们做过的最重大的事情之一，你懂的，是我们在这个
[13:50] Robby
product. So that's kind of chapter two, uh, which is around understanding the root causes. Now, if you're lucky, you have a great seed idea, you've done this process of finding the market fit by going looping through, seeing your problems, analyzing them rigorously, and fixing them aggressively. Maybe you get to the stage now where the product is ready for primetime. You're going all out you're going to all your customers, you're sending the blast email. And really this is about the fit and finish and final elements of the product, which is around craft. And I think this is often overlooked, but basically the way I think about craft is the feeling that I think everyone has a general sense and I believe people can detect whether the creators of a product should care about the product that they created, basically. And there's two things that I think drive this. One is does the product work perfectly? And is there any user pain, basically, in that experience? And the second is how does it make me feel? Do I feel something, joy, excitement using it? And so I'm going to talk about each real quick. Um, on the pain side is really not just about obsessing over it and you all being power users of your own product, but actually how to scale that now with AI, which is really incredible. And so I actually built a tool that I'm sharing here. It's a real tool that I built, um, internally using the anti-gravity system to basically look at all of the potential questions we might ask. So here's a sample here on the side.
产品上看到的。这差不多就是第二章，讲的是理解根因。接下来，如果你运气好，你有了一个很棒的种子想法，也走完了寻找市场契合的这个过程——一圈一圈地循环，看见问题、严谨地分析问题、再大刀阔斧地把它们修掉。也许你现在就到了这样一个阶段：产品可以正式登台亮相了。你要全力出击，要推给所有客户，要群发那封邮件。而这个阶段真正关乎的，是产品的做工、收尾和最后那些细节，也就是手艺。我觉得这一点常常被忽视。我对手艺的理解，基本上是这样一种感觉——我想每个人大体上都有感知，而且我相信人们能察觉出来：做这个产品的人，到底在不在乎自己做出来的这个东西。我认为有两件事在驱动这种感觉。一是产品是不是运转得完美无误？这段体验里，说白了，有没有任何让用户难受的地方？二是它让我有什么感受？我用它的时候有没有感觉到点什么，快乐也好，兴奋也好？所以这两点我各快速讲一下。痛点这一面，其实不只是说你要对它死磕、你们每个人都要当自己产品的重度用户，而是现在怎么用 AI 把这件事规模化，这真的很不可思议。所以我其实做了一个工具，在这儿分享给大家。这是我真的自己做出来的一个工具，是在内部用 Antigravity 做的，基本上就是去把我们可能会问的所有潜在问题都过一遍。旁边这里就是一个示例。
[15:10] Robby
Use the product as an agent, so like literally like go to Google, type in a question, go to the AI and ask a follow-up, and then record that experience, screenshot it, and then evaluate what you see based on a really comprehensive rubric. And what we found out was that the model actually was incredibly capable to look at things like, "Hey, there's a math question." or this is like a school homework thing. Is it doing a great job rendering math in LaTeX? Or is it, you know, emitting weird uh characters? Or if you ask about bioluminescence, is it adding a visual tray, which is what the response should have if you're asking about something visual? Or is it just text? And you can go through and you not only can you get to the best response, but you can see breakages really easily and the team easily and the team has now invested in a bunch of agents that can go figure out when stuff is just broken or off spec, and then increasingly fix it itself. And you have this feedback loop and where used to have maybe a QA team, you you may need a PM, or you might need people on your team who are obsessively using every single part of the product. I think many of those days are now over, where we can scale this process much more elegantly for everyone. And then the last thing I'm going to say is around, you know, delight and really sparking joy. And this is a little ex- perience. This is what the Google search box looks like now. We we reinvented this the search box after 20 years, announced it at I/O, what an AI-based search box should be. It should grow with you. It should be able to upload things, and have context around files, around videos, around photos. What What should that be like? And And how daunting of an experience is that? And what we did was we came up with a a system that said, "Okay, we wanted Google to still feel Google." And but we wanted it to be this upgrade that felt like AI is here. It's in There's AI
以 agent 的身份去用这个产品，也就是真的去打开 Google，敲进一个问题，再到 AI 那里追问一句，然后把这段体验录下来、截图，再按照一份非常全面的评分细则（rubric）去评估你看到的东西。我们发现，模型在看这类事情上其实能力强得惊人，比如：“嘿，这是一道数学题”，或者这是学校作业一类的东西，它用 LaTeX 渲染数学公式渲染得好不好？还是说，在吐一些奇怪的字符？又或者你问的是生物发光，它有没有加上一个视觉内容栏（visual tray）——你问的要是视觉性的东西，回答里就该有这个——还是说只有文字？你可以这样一路过下去，不光能找到最好的回答，还能非常轻松地看到哪里坏了。团队现在已经投入做了一批 agent，它们能自己去发现哪些东西压根坏了或者不合规格，然后越来越多地自己把它修好。这样你就有了这么一个反馈闭环。以前你可能得有个 QA 团队，可能需要一个 PM，或者需要团队里有人着了魔似的把产品的每一个角落都用一遍。我觉得这样的日子很多都已经过去了，我们现在可以用优雅得多的方式，替所有人把这个流程规模化。然后我要讲的最后一点，是关于愉悦感，关于真正点燃快乐。这里有一个小小的体验。这是 Google 搜索框现在的样子。时隔 20 年，我们把搜索框重新发明了一遍，在 I/O 上发布的——一个基于 AI 的搜索框应该是什么样。它应该能跟着你一起伸展变大。它应该能上传东西，能围绕文件、视频、照片拥有上下文。那该是什么样？这样的体验又会有多让人望而生畏？我们的做法是，想出了一套体系，它的意思是：“好，我们希望 Google 依然有 Google 的感觉。”但我们又希望它是一次升级，让人感到 AI 来了。它就在——有 AI
[16:48] Robby
inside that's helping you. And so there's this concept of this kind of gradient of color that really represents the AI kind of power behind the product. And you can actually see a couple nice details here that the design team brought together. One was actually in the cursor, if you watch it blinking, it's cycling through the different Google colors. And I I can't tell you how many times I've seen this pop up on X that like someone noticed this and and and and and just and felt like posting it. It just blows me away every time. Or if you tap in the search box, it's kind of excited, it like jumps out kind of and then loads the ring. It's like, "Ooh, like we're getting ready to load. Like we're we're here, you know?" It has this kind of feeling. And it actually has real feeling with haptics. So, you actually see Imagine all that coming together as an experience. I mean, this is one small example of just by using the product and through motion, through color, through feel, through touch, you know, and the intentionality of the experience, you communicate the humanity behind the product and you give it a soul. And I think many people overlook this, you know, part of the the product creative process.
在里头帮着你。所以就有了这么一个概念：用一种颜色的渐变，来真正代表产品背后的那股 AI 力量。你其实能在这儿看到设计团队凑到一起的几个漂亮细节。一个其实在光标上：你盯着它闪，会发现它在 Google 的几种颜色之间轮流切换。我都数不清有多少次在 X 上看到这个冒出来——有人注意到了，然后就是想发一条。每次都让我特别震撼。再比如你点一下搜索框，它有点兴奋似的，像是往外蹦了一下，然后把那圈光环加载出来。就好像在说：“哦，我们准备要加载啦，我们在这儿呢，你懂吧？”它有这么一种感觉。而且它还真的有实打实的手感，带触觉反馈（haptics）。所以你其实能看到——想象一下所有这些合在一起，成为一个体验。我的意思是，这只是一个小例子：光是通过使用这个产品，通过动效、通过色彩、通过手感、通过触碰，还有体验里那份用心，你就传达出了产品背后的人味儿，给了它一个灵魂。我觉得很多人都忽视了产品创作过程里的这一部分。
[17:43] Robby
And then really, when you bring this kind of all together, this is kind of the the wrap here. I think that this really means that you can have a a playbook for what it means to be a PM in this new, you know, AI frontier. And it's you have to still do all the same things. Like, a lot of this is not new, but it's it's possible in a totally new light with AI. So, you have to understand people deeply, get at their core needs and problems that they're feeling. You have to be rigorous and analytical at sorting out, "What are the things I need to fix in order to get the product to win and to work for people against that objective?" And I need to build it so that it's awesome and it's simple to use and it works great and it makes people feel awesome using it. And that's really the playbook. And I think that now it's exciting that the craft of PM is so critical. I think it's never been more important time to be a PM and to do product and to be a builder. And I think the skill of making great decisions that can lead to great products has also never been, you know, more important. So, hopefully all of you can take a little bit of something. This is just one way. There's so many ways to build products. You've probably heard 10 other ones today that were just as valid as this one. You should decide on what makes sense for you, but hopefully you take one thing away um that you can take away with you and to build as well. And hopefully the next time you're thinking about your products, you're spending a little bit less time thinking about, "Hey, here's cool features kind of out of the blue or here's some tech we're going to throw at it or here's a new way we're going to get it in front of people." And you spend more of your time thinking about the actual people and the humans on the other side of the screen that you're building for.
然后，把这些全都合到一起——这差不多就是收尾了。我觉得这真正意味着，在这个新的 AI 前沿，当一个 PM 意味着什么，你是可以有一套打法的。那就是：你仍然得做所有那些一样的事。这里面很多都不是新东西，但有了 AI，它能以一种全新的方式做到。所以，你得深入地理解人，摸到他们的核心需求，摸到他们正在感受的问题。你得严谨、有分析力地理清楚：“对照那个目标，为了让产品赢、让它对人们管用，我需要修掉的是哪些东西？”然后我得把它做出来，做得很棒、用起来简单、运转顺畅，让人用的时候感觉很棒。这其实就是那套打法。我觉得现在令人兴奋的是，PM 这门手艺变得如此关键。我认为当 PM、做产品、当 builder，从来没有哪个时候比现在更重要。而且我认为，做出好决策、进而通向好产品的这项能力，也从来没有像现在这么重要过。所以，希望你们每个人都能带走一点东西。这只是其中一种方式。做产品的方式太多了。你们今天多半已经听了另外 10 种，跟这一种同样站得住脚。你应该自己决定什么对你来说行得通，但希望你能带走一样东西，带回去也用来造点什么。也希望下次你思考自己的产品时，能少花一点时间去想“嘿，这儿有些凭空冒出来的酷功能”，或者“这儿有项技术，我们往上一怼”，或者“这儿有个新办法能把它送到人们面前”，而把更多的时间花在想那些真实的人上——屏幕另一头、你在为之打造产品的那些人。
[19:08] Robby
Thank you so much.
非常感谢大家。
[19:10]
[applause]
（掌声）