本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
人们怀疑那些规模庞大、历史悠久、技术精湛且能力强大的工程组织能否大规模部署人工智能并产生实际效果。
People are skeptical that large, established, highly technical, highly capable engineering organizations can deploy AI at scale and get any effect.
但我认为你已经证明了这是可能的。
But I think you've proven it's possible.
这不仅是可能的，更是不适应就死亡。
It's not only possible, it's adapt or die.
这对团队来说简直是一种巨大的超级力量。
It's just been such a huge superpower for the team.
我们这里说的是多少工程师？
How many engineers are we talking about here?
一千多人。
A thousand plus.
所以我们这里可不是在玩票。
So we're not messing around here.
公司曾尝试采用其他人工智能工具。
The company tried to adopt other AI tools.
我们看到了采用率的上升。
And we saw this uptick in adoption.
人们打开了它，勾选了框，做了个类似‘你好，世界’的尝试，但并没有持续下去。
People opened it up, checked the box, did kind of like a hello world thing, but it didn't stick.
我最关心的是，怎样才能让这个该死的东西真正落地？
My biggest thing is how do I make this damn thing stick?
因为这里确实有东西。
Because there's something here.
我认为，在进行这种组织转型时，至关重要的是，领导层中要有一个人，拥有无比坚定的信念，并且能亲自上手实践。
I do think that it's really important when you're doing this organizational transformation that you have a single person with incredible conviction at leadership level who is also hands on the metal.
要让工程师们看到，而不仅仅是听到。
Show the engineers, not just tell.
任何工程负责人最糟糕的做法就是说：我命令你们必须使用AI。
And the worst thing any engine leader could do is just be like, I decree you must use AI.
拜托了。
Come on.
没人会听你的。
No one's gonna listen to you.
欢迎回到《我是AI》。
Welcome back to How I AI.
我是克莱尔·贝尔，产品领导者，也是个AI狂热者，我的目标是帮助你用这些新工具更好地构建产品。
I'm Claire Bell, product leader and AI obsessive here on a mission to help you build better with these new tools.
今天，我们邀请到Coinbase的高级工程总监钦廷·蒂拉基亚，他将向我们展示，是的，在一个拥有数千名工程师的工程组织中推动AI采用和提升开发速度是完全可能的。
Today, we have Chintin Thirakia, senior director of engineering at Coinbase, and he's gonna show us, yes, it is possible to drive AI adoption and higher velocity in an engineering organization of thousands of engineers.
他还将向我们展示工程经理和工程领导者的新期望：少开会，多写代码。
He's also gonna show us the new expectations for engineering managers and engineering leaders, which is less meetings and more code.
我们开始吧。
Let's get to it.
本集由WorkOS赞助。
This episode is brought to you by WorkOS.
AI已经改变了我们的工作方式。
AI has already changed how we work.
这些工具正在帮助团队编写更好的代码、分析客户数据，甚至自动处理支持工单。
Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically.
但这里有个问题。
But there's a catch.
这些工具只有在能够深度访问公司系统时才能良好运行。
These tools only work well when they have deep access to company systems.
你的代码助手需要查看你的整个代码库。
Your copilot needs to see your entire code base.
你的聊天机器人需要搜索所有内部文档。
Your chatbot needs to search across internal docs.
而对于企业客户来说，这引发了严重的安全顾虑。
And for enterprise buyers, that raises serious security concerns.
因此，这些应用从第一天起就面临IT部门的严格审查。
That's why these apps face intense IT scrutiny from day one.
为了通过审查，它们需要具备安全的认证、访问控制、审计日志等一系列企业级功能。
To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features.
从零开始构建所有这些功能？
Building all that from scratch?
这是一项巨大的工程。
It's a massive lift.
这就是Work OS的用武之地。
That's where Work OS comes in.
Work OS为您提供即插即用的企业功能API，让您的应用能够快速具备企业级能力并拓展市场。
Work OS gives you drop in APIs for enterprise features so your app can become enterprise ready and scale up market.
更快。
Faster.
可以把这想象成企业功能领域的Stripe。
Think of it like Stripe for enterprise features.
OpenAI、Perplexity和Cursor已经使用WorkOS来加速发展并满足企业需求。
OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands.
加入他们，以及数百家其他行业领军企业，访问workos.com。
Join them and hundreds of other industry leaders at workos.com.
今天就开始构建吧。
Start building today.
钦廷，非常感谢你加入我们。
Chintin, thank you so much for joining.
我喜欢我们今天要讨论的主题，因为我们花了太多时间谈论个人极客或非技术人员如何成为软件工程师。
What I love about what we're going to talk about today is we spend so much time talking about the individual vibe coder or the nontechnical person becoming a software engineer.
但人们仍然对大型、成熟、高度技术化且能力强大的工程团队能否大规模部署AI并产生实际效果持怀疑态度。
And still people are skeptical that large, established, highly technical, highly capable engineering organizations can deploy AI at scale and get any effect.
仍然存在很多怀疑。
There's still so much skepticism.
但我认为你已经证明了这是可能的，而且你很可能将为我们指明方向。
But I think you've proven it's possible, and you're hopefully gonna show us the way.
我认为这不仅是可能的。
I I think it's, not only possible.
这是适者生存。
It's, you know, adapt or die.
这简直就是团队的巨大优势，我们从中获得了极大的效率提升。
Like, it it's it's just been such a huge superpower for the team, and we've gotten so much efficiency out of it.
而且，其实有很多不同的方式可以去实施。
And and there's just, like, ways to approach it.
我昨天好像看到一条推文，讲的是微软内部一个非常长的故事，有人把Copilot引入了他们的组织。
I was I think I was reading a tweet yesterday just about a very, very long story at Microsoft and or someone, like, pulling Copilot into their organization.
那条推文只是轻松地表示：没错。
And it was just, like, just a fun tweet of just, yep.
我们要让图表向右上方攀升。
We're gonna make Graph go up into the right.
但实际的采用情况并不理想。
But, like, the actual adoption wasn't good.
所以过去一年，我一直在全神贯注地研究这个问题。
And and so, like, I've been spending the last year just absolutely obsessing about it.
你真的可以做到。
And you can do it.
人们是可以做到的。
People can do it.
那你们具体该怎么做到呢？
So how how can you do it?
你知道，我们这里说的是多少工程师吗？
Because, you know, how many engineers are we we talking about here?
一千多人。
A thousand plus.
是的。
Yeah.
我们可不是在玩票。
So we're not we're not messing around here.
这确实是。
This is Yeah.
我们可不是在玩票。
We're not messing around.
这确实是一个真正团队在开发真实产品，他们知道自己在做什么，并且已经打造出了优秀的软件。
Real this is a real team working on real products who know what they're doing, who have built great software.
那么，你是从哪个方面开始的呢？无论是文化上、产品角度，还是工具层面？
And so where did you start, either culturally, from a product perspective, from a tools perspective?
我觉得很多改变实际上是从去年这个时候开始的。
So I think a lot of it actually just started around this time last year.
我们进行了一些调整，以更好地协调我所负责的产品。
We had some changes to align, like, the product I'm responsible for.
其中很大一部分工作实际上是彻底从零开始重写整个产品，把它从一个自托管钱包转变为一个只是恰好使用加密货币的社会化消费类应用。
And a big part of that was effectively, like, rewriting the entire product from scratch, from turning it from a self custody wallet to actually a social consumer app that just happens to use crypto.
我们虽然使用的是 React Native，但当初为自托管钱包做了一系列决策。
And, you know, we're using React Native, but we made a lot of decisions for our self custody wallet.
但要转型为消费类应用，你就得重新思考一切。
But to become a consumer app, you gotta, like, rethink everything.
这是其中之一。
That was one.
第二，我们需要在六到九个月内完成。
Two, we needed to do it in, like, six to nine months.
所以我们直接与那些拥有数千人团队、领先十年的大型社交平台正面竞争。
So we were going head to head with like the big social players out there that have multi thousand person teams that have a ten year head start.
我们真的只是想做一件宏大、新颖又疯狂的事情。
And we were really trying to just do something big and new and crazy.
纯粹就是疯狂。
Like, absolutely just crazy.
而其中关键的一点是，我们如何重新设计这款应用，让它成为市面上最顶尖的消费级应用，并在如此疯狂的时间表内完成？
And and a big part of this is, like, how do we rewrite the app so that it is the best possible app out there, like, consumer grade and do it in this insane timeline?
团队非常出色。
And the team is cracked.
他们太棒了。
They're amazing.
但你知道，由于这些变革，我们的团队规模变小了。
But, like, you know, we we we became a smaller team as a result of of some of these changes.
于是我开始寻找各种加速的方法。
And so I started just looking at, like, ways to accelerate.
而且，你知道，我也不确定。
And and, you know, like, I don't know.
我的团队非常了解我。
My my team knows me well.
如果你了解我，就会知道我特别痴迷于效率。
And if you if you know me, like, I obsess about efficiency.
我认为这至关重要，要以合理的方式提升团队的开发速度，无论是通过工具还是使用工具。
And I think that's, like, so critical to, like, make teams accelerate their velocity, but in in in ways that make sense, for tool and using the tool.
大约在这个时候，我认为Cursor发布了他们的初始版本。
So at around this time, I think Cursor had come out with their sort of initial release.
那是在十一月左右。
It was around, like, November.
我们都试用了，2024年，但感觉挺差的。
We we all tried it, right, 2024, and it kinda sucked.
而且我并不是真的喜欢 Cursor。
And it it's not like I love Cursor.
我喜欢 Cursor。
I love Cursor.
当时的模型还不够成熟。
The models weren't there.
就是模型还不够成熟。
The just the models weren't there.
我觉得当时的模型甚至都写不出像样的单元测试，你知道的。
Like, I think the models couldn't even, you know, really write a unit test, right, well.
你知道，你是个工程师，你明白，一旦工程师试用一个工具后觉得不好用，就很容易很快地把它否决掉。
And, you know, you're an engineer, and you understand once an engineer tries a tool and they're like, Ah, this is not so good, it's very quickly and very easy to write it off.
对吧？
Right?
这种情况很常见。
It happens.
所以我们经历了这么一段低谷期，简直像在悲伤的深渊里徘徊。
And so we we kinda went through this, like, trough of sorrow.
就是觉得，好吧。
Was just like, okay.
该死。
Goddamn it.
AI工具还没到位。
AI tools are not here.
这些模型还不成熟。
The models aren't ready.
我们该怎么办？
What are we going to do?
就在这一事件发生前的一年里，公司还尝试过采用其他AI工具，比如GitHub Copilot。
And, you know, for even a year prior to this event, like, the company tried to adopt other AI tools like GitHub Copilot.
我们看到了使用率的明显上升。
And we saw this, like, uptick in adoption.
比如，人们打开了它，勾选了框，做了一个简单的Hello World测试，但并没有坚持下去。
Like, people opened it up, checked the box, did kind of like a hello world thing, but it didn't stick.
对吧？
Right?
而我最关心的是，我该怎么让这个该死的东西真正落地？
And and, like, my my biggest thing is how do I make this damn thing stick?
对吧？
Right?
因为这里确实有东西。
Because there's something here.
对吧？
Right?
我一直以来的思维模式都是，模型——那些基础的大语言模型——总会变得越来越好。
And my mental model was just always the models will the the foundational LLMs will always get better.
这就像是去健身房一样。
And it's like going to the gym.
你需要去实践，多尝试，这没关系。
You need to go and build your reps and try it, and that's okay.
而且做这件事的成本几乎可以忽略不计。
And the cost of doing it is like nothing.
只是浪费一点点时间而已。
It's just a little bit of waste of time.
我们现在不担心计算资源，因为还处于早期阶段。
We're not worried about compute right now because it's so early.
所以，从大约今年一月一直到2025年三月或四月，我彻底改变了心态和思维方式。
And so, like, from basically January all the way to, like, March or April 2025, I just changed the the mindset and the mentality.
我每天都泡在Cursor里，一天到晚都在用。
I I was, like, in cursor every single day, every single hour of the day.
我就在想，怎么才能让这东西真正奏效呢？
And I was like, how do I make this work?
对吧？
Right?
这很棒，因为我又开始写代码了。
Like, It was great because I was writing code again.
这很棒，因为它解锁了所有这些使用场景。
It was great because it was unlocking all these use cases.
我们正在面试候选人，只是我不想一定要把所有笔记都写下来。
We were doing interviews, interviewing candidates, just like, I don't wanna necessarily write up all the notes.
对吧？
Right?
这要花很长时间。
That takes a long time.
但我直觉上，我知道。
But I intuitively, I like I know.
我已经评估过了。
I've assessed.
对吧？
Right?
所以我会用它来处理一些战术性的日常文书工作，以加快我的进度。
So I would use it for, like, tactical day to day paperwork kind of things to accelerate me.
但从编程的角度来看，我会直接挑出一些bug，然后说：嘿。
But also from, like, a coding perspective, I would just pick up bugs and be like, hey.
我们试试这个。
Let's try this.
对吧？
Right?
会发生什么？
What's gonna happen?
我能学到什么？
What can I learn?
有哪些技巧和诀窍，能真正展示给工程师们，而不仅仅是告诉他们？
What are the tips and tricks to, like, show the engineers, not just tell?
任何工程主管最糟糕的做法就是直接说：我命令你们必须使用AI。
And the worst thing any engine leader could do is just be like, I decree you must use AI.
拜托吧。
Like, come on.
没人会听你的。
No one's gonna listen to you.
我必须理解这一点，因为我本人也管理着一个规模达数百人的工程团队，当时就亲身体验了这些工具的早期版本，并且发自内心地坚信它们必然会改变我们的工作方式。
I have to empathize with this because I also, running a large, like, multi 100 person engineering organization, you know, was experiencing even early versions of these tools and had such innate conviction that it would, of course, transform how we did work.
这对我来说非常明显。
Like, that was very obvious to me.
我不知道。
I don't know.
是因为经验而明显，还是因为它本来就很明显？
It's obvious because of experience or obvious because it was just obvious.
但确实如此。
But Yeah.
但作为领导者，你当时就是经历了这些，尤其是在大约一年前。
But then, you know, you just had these experiences as leaders, especially in the, you know, maybe twelve months ago.
一个工程师试了，但没成功。
One engineer tries it, doesn't work.
不只是那个工程师把它扔了。
It's not just that engineer throws it away.
而是其他人都说：我觉得，我信任他们的意见。
It's everybody else says, well, I think, you know, I trust their opinion.
如果他们说这行不通，那对我来说也行不通。
And if they say it's not gonna work, it's not gonna work for me.
我认为，当你进行这种组织转型时，非常关键的是，领导层中要有一个人拥有极强的信念，并且亲自参与实践。
And I do think that it's really important when you're doing this organizational transformation that you have a single person with incredible conviction at the leadership level who is also hands on the metal.
因为只有当你能说：我明白它对那个人没用，但它对这三件事有效。
Because until you can say, well, I understand it didn't work for that, but it worked for these three things.
或者我真正找到了让它对那件事生效的方法，因为我们尝试了A、B和C。
Or I actually figured out how to make it work for that because we tried A, B and C.
我认为，这是唯一能避免陷入空谈的方式。
I think it's just the only way you cannot be in philosophy.
你不能把希望寄托在未来的某个人身上。
You cannot be in somebody in the future.
你自己去搞明白。
You figure it out.
你必须真正地重新回到它上面。
You have to actually get back to it.
然后我觉得，这算是额外加分了。
And then I think, like bonus points.
我们很多工程领导者都被迫远离了编码工作。
So many of us in engineering leadership have, like, been pushed away from from coding.
我正打算重新投入进去。
I was having get back in it.
我就只是想再写代码了。
And I'm like, I just wanna code again.
给我一点乐趣吧。
Like, give me some joy.
给我点时间。
Give me some time.
是的。
Yeah.
而且而且
And and
这也是一种好处。
that's a benefit as well.
你必须展示，而不是告诉。
And you you have to show, not tell.
所以我这么做了。
And and so I did.
而且，我觉得我很快学到的是，好吧。
And, like, I think what I learned very quickly is like, okay.
这里有些东西。
There's something here.
那里就有一个。
There's there's a there.
对吧？
Right?
然后我们就开始挑选一两个使用场景。
And then we just started picking off, like, one or two use cases.
让工程师真正投入的最好方式，就是给他们工具，让他们不再做那些琐碎的工作，而是去构建他们真正热爱的东西。
And and the best way to get to an engineer is just give them the tools so they stop doing the shit work and so that they can build the stuff they love.
对吧？
Right?
对吧？
Right?
所以我们就会先挑出单元测试来处理。
And so, like, we would just, like, pick off unit tests.
我们会处理代码规范检查，还有这些琐碎的小事，它们就像不断的小伤口，一点点耗尽开发者的心力。
We'd pick off, like, linting, all these, like, little things that just, like, paper cut and suck the soul out of you as a builder.
但工程师和团队只是想更快地推进。
But the engineers and, you know, like the team just wants to move faster.
团队想构建更好的东西。
The team wants to build better things.
所以我们开始在这些方面采用光标规则。
And so we started leaning into like cursor rules for some of these things.
即使是再简单的事情。
Even the simplest thing.
我记得，我想我的那个时刻是，提交了一个bug报告，然后处理了它。
I I remember, like I think I remember my moment, which was, like, popping in some bug report, working through it.
然后我就没再想它了。
And then I didn't think about it.
我只是去做了。
I just did it.
我当时就想，直接创建一个草稿PR吧。
I was like, just create a draft PR.
这是工单。
Here's the ticket.
这是PR，你知道的，这是我想要的PR描述。
Here's kind of the PR like, and, you know, here's the PR description I want.
它就自动完成了。
And it just did it.
我再也不需要记住git status、git rebase了。
And I was like, I never need to remember git status, git rebase.
谁还干这种事啊？
Not like, why is anyone doing this anymore?
我们到底在做什么？
Like like, what are we doing?
有趣的是，我花了点时间才说服团队。
And it took funny thing is it took some convincing of me to the the team.
伙计们，直接输入“创建草稿PR”，它就会帮你搞定。
Like, guys, just type create draft a PR like, create a draft PR, and it'll be done for you.
而且，呃，你知道的，我有自己的工作流程。
And, like like, well, you know, I kinda have my workflow.
我当时就想，太棒了。
I was like, cool.
太棒了。
Cool.
太棒了。
Cool.
太棒了。
Cool.
我明白你的工作流程。
I get your workflow.
你可以修改它。
You can modify it.
你可以使用光标规则。
You can use cursor rules.
没关系。
It's okay.
没人会因为背下Git命令而加分。
Like, no one's getting bonus points for memorizing Git commands.
没错。
Exactly.
没错。
Exactly.
于是我们一点点推进，添加了很多规则，比如光标规则，这帮助太大了。
And and so, like, we chipped away, and we put in a bunch of rules, like cursor rules, and that that helps so much.
然后，我当时在感受。
And then, like, we I I was, like, sensing.
我觉得可以了。
I was like, okay.
我团队里已经有足够多的人表示认同了。
I have I have enough, like, folks on the team that are like, yep.
这正在解锁一些东西。
This is unlocking stuff.
他们会在团队频道里发帖。
And they would post in the team channel.
你看，居然还有一个频道叫“Cursor赢了”。
Like, look what had literally a channel called cursor wins.
每个人都在那个频道里发帖。
And, like, everyone was just posting in the channel.
我就刚写了20个单元测试，然后去喝杯咖啡。
Like, I just did, like, you know, 20 unit tests and then went and had a coffee.
这太棒了。
This was great.
我真的超喜欢。
Like, I love it.
于是大家开始亲眼看到它在实际中发挥作用。
And so people started seeing it in action.
然后我们到了这么一个节点。
And then we hit this, like, point.
我当时想，好吧。
I was like, okay.
我该怎么让整个团队快速上手呢？
How do I speedrun now the whole team?
这里有一点信念在支撑。
There's a there's a little bit of conviction here.
所以我只是，我记得这件事。
So I just and I remember this.
我觉得我当时已经落地了。
Like, I think I had landed.
我正要去东海岸。
I was going to the East Coast.
我下飞机后打了个优步，然后立刻加入了一个全员会议，直接开始了速成培训。
I landed for my flight, got into an Uber, hopped on like an entire team all hands, like speedrun.
我们称之为基本上是Cursor速通。
We call it it was like basically Cursor speedrun.
我在Uber里用Cursor，提交了PR。
And I was in the Uber using cursor, putting up the PR.
速通的目标是每个人只挑最简单的事情做。
And the goal of the speedrun was every single person would just pick up the most trivial thing.
可能是改个文字、修个bug，随便什么都行，然后直接提交PR。
It could be like copy change, a bug, whatever, and just put up the PR.
最后我们大概用了十五分钟，我觉得有一百人加入了。
And we ended up I think in fifteen minutes, I think a 100 people had joined.
十五分钟内，我们一共提交了大约七十个PR。
In fifteen minutes, we ended up putting up, like, 70 PRs.
我们甚至把GitHub搞崩了，这挺酷的，因为我们意识到我们的基础设施需要改进。
And we broke GitHub too, which was cool because we learned, like, our infrastructure needed improvement.
我想稍微停一下，因为再次强调一下，关于AI的一些战术技巧。
So I wanna I wanna pause real quick because, again, How AI, a little bit about tactical techniques.
你用过几个我也会用的技巧，比如由一位高信念、亲力亲为的领导者来推动：我们必须尽快接触这些工具，专注于减少琐碎工作。
And you've used a couple that I have used, which is like one high conviction leader with hands on the metal that just says, like, we just got to do this access to tools, focus on toil.
我觉得这非常重要。
I think it's very important.
你提到了代码规范检查，也提到了测试。
You called out linting, you called out tests.
另一个我想指出的是设计负债，也就是前端工程师或设计师们长期忍受着他们讨厌的某些应用模块。
Another one I would call out is like design debt, where, you know, front end engineers or designers have just lived with parts of the app they hate.
是的。
Yes.
这确实是另一个非常好的做法。
That is another really great one.
然后我们还建立了一个共享的 Slack 频道，我对你那个 Cursor 胜利频道的一个延伸建议是，我们把它改成了‘胜利与失败’。
But and and then a shared Slack channel and one like, you know, riff I would make on your cursor wins channel is we made ours wins and losses.
所以我们非常明确地要求：只要发布你做了什么，以及什么时候成功了、什么时候失败了。
And so we were very clear, like, just post what you did and when it worked and when it doesn't.
因为当事情没成功时，人们就会说，哦，是的。
Because when it didn't, people would be like, oh, yeah.
但你可以尝试X、Y、Z，或者我有个光标规则给你，等等。
But you could try X, Y, Z or I have a cursor rule for you or whatever.
但我还没听过一个值得大家竖起耳朵认真关注的想法，那就是‘PR速通’——也就是利用一段空闲时间，所有人启动工具，集中快速解决一些问题。
But what I haven't heard that I want people to just, like, perk their ears on and pay attention to is this, like, idea of a PR speedrun, which is like do a time downtime, everybody boot up whatever tool and just speed run some fixes.
因为一个组织要有多大的决心，才能从‘看吧，我经历过季度规划的拖延，这事儿得等四个月后再说’这种状态，突然转变成‘我们刚在三十分钟内把积压的70个PR全部发出去了’？
Because how much conviction does an org have to get going from, look, I've been there, the doldrums of like quarterly planning, and this will be in four months and blah blah blah blah blah to just, like, we just got 70 PRs that we've been sitting on out out the door in in thirty minutes.
这对任何边缘团队来说，都必须是一个变革性的时刻。
I just that has to be such a transformational moment for an edge team.
你知道吗，那些PR合并的成功率简直低得离谱。
You know, there was a success rate on those on merging those PRs, and, like, it was just like, shit.
这事儿是能做到的。
This is possible.
他们所有人都眼前一亮。
They're they like, everyone's eyes lit up.
这简直是对状态更新和长期构建的一种终结。
And it was really sort of a death to status updates, long lived building moment.
对吧？
Right?
是的。
Yeah.
还有另一件事我想强调，因为我觉得你们团队的文化真的很特别。
And and this is the other thing I wanna call out because I think you all have a really special culture there.
但很多时候，我们在产品、工程和设计团队中，会过于纠结于合作规则。
But so often, we in product engineering design orgs get, like, really wrapped around the axle on, like, the rules of engagement.
比如，我不被允许开发某个功能，除非产品经理说它很重要；或者我不能决定按钮的颜色，因为设计团队有话语权。
Like, I'm not allowed to build it unless the product manager says it's important or, like, I can't really make that decision about what color that button is because design has a weight in.
但我确实认为，像这样的时刻，当你彻底打破所有规则时，会让人恍然大悟：
And, like, I do think these moments where you just break all the rules and you're like, guess what?
记住，你完全可以直接上线代码。
Remember, you can just ship code.
就可以
Can just
你可以直接发布代码。
You can just ship code.
别管人工智能了。
Like, put AI aside.
人工智能也许能促成这一点，让成本低得多。
AI maybe enables it and makes it like a much less costly, know, expense.
但仅仅是这样做，就能极大提升效率，我认为还能提升质量。
But, like, just doing that is so powerful for velocity and for I also think for quality.
人们会更彻底地对事情负责。
Like, people just take more radical ownership of things.
所以我百分之百要采纳这个做法。
So I'm gonna a 100% steal this.
我的意思是，我希望每个人都去借鉴。
You I mean, I want everyone to steal it.
就像你说的，我真的很喜欢你刚才的说法。
Like, you know, I I really like the way you just put it.
对吧？
Right?
现在正是我们应该打破规则的时候，因为AI正在为我们打破规则。
This is a moment where we should be breaking the rules because AI is breaking the rules for us.
如果我们不适应AI的使用方式，我们就完蛋了。
And if we don't adapt to how, like, we can use it, we're toast.
对吧？
Right?
而且，作为一个整体，任何不适应的人都会落后。
And and we as, like, a very collective, like, whoever's not adapting is gonna fall behind kind of thing.
对吧？
Right?
所有这些最终所释放出来的，就是协调成本的降低。
And what all of this, like, ends up unlocking is is, like, the reduction in coordination overhead.
所以，我最近一直特别纠结的一件事是，好吧。
So, like, one thing I've been obsessing about a lot is like, okay.
酷。
Cool.
太好了。
Great.
速度通关做得不错。
Good job on the speed run.
是的。
Yes.
我们完成了很多事情。
We got a lot of stuff done.
我们开始看到这些成果了。
We started then seeing those wins.
越来越多的人开始采用。
More and more people adopted.
然后布莱恩，我们当时正在和布莱恩分享一些关于采用进展的信息，接着我们就举办了一次全公司的速度挑战赛。
Brian then you know, we were we were sharing some information with Brian, like, how adoption's going, and then we just did a company wide speedrun.
就在那一刻，大约有800名工程师在通话中，我们最终在三十分钟内推送了三四百个拉取请求。
And at that moment, like, there were, like, 800 engineers on the call, and we ended up pushing up for, like, three, four hundred PRs in thirty minutes.
而且，是的，我们又把GitHub搞崩了。
And, yes, again, we broke GitHub.
但这没关系。
And that's fine.
这很好。
That's good.
这其实是在压力测试。
Like, this is pressure testing.
我们应当设计自己去打破规则。
We should be designing ourselves to break the rules.
对吧？
Right?
但我一直纠结的问题是，你到底该怎么衡量这些成果呢？
But the thing I've been obsessing about is, like, how do you how do you measure any of this, like, in terms of output?
对吧？
Right?
这里存在一种矛盾：我们用的AI越多，这算不算是在替代人力？
There's there's this, like, tension where, okay, the more AI we use, well, does that count as a replacement for people?
而我坚定地认为，绝对不是。
And, like, I'm in the camp of absolutely not.
AI是一种加速器。
AI is an accelerant.
AI是一种加速器，因为总会有更多工作等着去做。
AI is an accelerant because there will always be more work to do.
对吧？
Right?
所以，我看待这个问题的方式——至少对我团队和我推动的整个方向来说——就是从创建工单到变更上线并交付给用户所花费的时间。
And so the way I think about it, at least for my team and what I'm pushing across the board, is really time from ticket to when the change lands to the user.
这实际上涵盖了你需要的每一个部分。
Like, that actually encompasses every single piece you need.
对吧？
Right?
而今天，即使你面对的是积压的工单之类的问题，比如，我该不该像你说的那样，优先处理这个？
And today, like, even if you go from, like, ticket backlogs and stuff like that, like, there's, oh, do I should I like like you said, should I prioritize this?
这个重要吗？
Is this important?
我去问问产品经理，或者问问项目负责人，不管叫什么吧。
Let me ask my PM, or let me ask the pro the pro program product manager project manager, whatever.
而现在，整个团队，从那时到现在，我们只要看到有人给我们反馈。
And now the whole team, like, fast forward from back then to now, we just see someone give us feedback.
几乎在几秒钟内，我们就会启动计时器，因为我们开发了一个内部机器人。
And literally within, like, seconds, we're like, clock like, we we built this internal bot.
我很高兴向你们展示。
I'm excited to show you.
几秒钟内，PR 就开始撰写了。
And within seconds, like, the PR's being authored.
对吧？
Right?
一个代理会立即处理，几秒钟内反馈就被执行了。
An agent picks it up, and within seconds, that feedback is, like, acted on.
所以我们压缩了响应时间，从创建工单到 PR 准备好接受审查，再到审查时间——我的开发人员总是抱怨，审查时间太长了。
And so we crunch the time to action, the time then from ticket to the the PR being ready for review, then the review time, like all my devs complain, review times take too long.
我们其实找到了一些解决方案。
We found some solutions actually.
我觉得我们以前的平均周期时间大约是 150 个小时，因为工作量太大了。
I think we were doing average of like one hundred and fifty hours, like was the cycle time for a PR review because there was so much.
我们把它缩短了十倍，降到大约 15 个小时左右。
We reduced it by 10 x down to, like, fifteen hours or so, roughly.
最后一步是，从合并之后，你如何进行 OTA 更新？
And then the last piece is, like, from that merge, how do you do, like, that OTA update?
然后你再次压缩整个周期，团队就会真正地获得惊人的速度。
And you squeeze that whole cycle again, and then the team is, like, just literally unlocked with sheer velocity.
是的。
Yeah.
然后我们就能把产品呈现给客户。
And then we get stuff in front of customers.
对。
Yes.
接着你就拥有了真实市场想法的快速迭代能力。
And then you have the velocity of, like, actual market ideas.
对。
Yes.
你会收到这些反馈，而且我们也在极度关注如何能最快地采纳现实生活中的反馈。
And you get that feedback and, like, the the we're obsessing also about how fast can we take, like, in real life feedback
对。
Yep.
展开剩余字幕（还有 480 条）
然后当场就把它修复了。
And then actually just fix it right then and there.
对。
Yep.
我觉得还有另一个关键时刻。
I think I think there is another moment.
我曾经和我们产品的一个用户开过一次会。
I was on a call with with, like, a a user of our product.
对吧？
Right?
他们说：嘿。
And they're like, hey.
如果你们能改一下 x、y 和 z，那就太好了。
It'd be cool if you, like, changed x, y, and z.
就在通话过程中，我直接提了一个 PR 并推送了更新。
And, like, literally, while I was on the call, I just put up a PR and pushed it.
他们说，在通话结束前，才过了三十分钟。
And they're like, before the call ended, it was thirty minutes.
我当时就说，你知道的，重新加载一下应用。
I was like, just, you know, reload the app.
已经修复了。
It's fixed.
好的。
Okay.
在我们进入长达一小时的讨论之前，让两位终端产品负责人说说‘快速发布’这件事。
Before we put this into an hour of, you know, like two two end product leaders being like, just ship really fast.
我们会深入探讨缩短PR周期时间的种种好处，那些有趣的内容。
We'll go into the merits of reducing PR cycle time, all that fun stuff.
让我们实际展示一下你开发的几个功能，因为我觉得，关于‘在工程团队中可以这样做’，这种元层面的评论很有意义。
Let's actually show a couple of things you built, because I think the kind of meta commentary on like, you can do this in engineering organizations.
这中间是有步骤的。
There are steps to it.
你可以采取一些措施，我认为这些都是每个人都能从中学习的东西。
There are measures you can take, I think, are things that everyone can learn from.
但你也一直在构建。
But you also have been building.
所以让我们谈谈你是如何实际使用Cursor来推动这一举措在组织中的落地，并理解AI的采用情况的。
So let's talk about how you used actually Cursor to drive how you drove this into the organization and understand adoption of AI.
是的。
Yeah.
当然。
For sure.
我认为很多东西都源于真诚的好奇心，以及找出瓶颈在哪里。
I think a lot of it just comes, like, from honest curiosity and figuring out where the bottlenecks are.
比如，为什么大家不采用呢？
Like, why aren't folks adopting?
人们是如何使用它的？
How are people using it?
诸如此类。
Etcetera, etcetera.
我想给你展示一下，我接下来要给你讲的这个疯狂想法是，我突然有了个异想天开的点子。
I wanna show you, I think the the kinda crazy thing I'm about to walk you through is like, just got this harebrained idea.
Cursor 有非常棒的分析功能。
Cursor has, like, great analytics.
对吧？
Right?
所以你进入管理面板，查看分析数据，很酷的是，它们允许你将数据下载为 CSV 文件。
And so you go to the admin panel, you look at the analytics, and, you know, awesomely, they let you download it into CSV.
我当时想，如果我用 Cursor 来弄清楚我的团队是如何使用 Cursor 的，而不是仅仅从像 AI 提交代码行数这种表面指标来看，会怎么样？
I was like, what if I just use Cursor to figure out what my team is doing in terms of using Cursor, but not in just like from a vanity metric point of view of like lines of code committed by AI.
我觉得那种方式其实有点误导性，我更想深入挖掘他们是如何使用 Cursor 的，以及我们如何复制那些高效用户的做法。
I think that's, like, kind of misleading, actually digging more into how they're using Cursor and how do we sort of, like, replicate power users.
我们来看看。
So let's see.
我们有一些数据。
We have some some data.
它就在这个文件里。
It's in this file here.
这是一个标准的CSV文件，来自Cursor，你可以从他们的网站、你的管理面板下载。
And it's just a, like, a standard CSV from Cursor that you can, like, download from their their site, like your admin panel.
这里还有一些不同的字段。
And then there's also here a bunch of different sort of fields.
比如接受的行数、聊天行数、删除的聊天行数，以及其他各种数据元素。
So, like, accepted lines, chat lines, chat lines deleted, various, like, data elements.
但一开始，我只是想了解Cursor的使用情况。
But, you know, one thing, like, I just sort of started with, I wanna understand the usage of cursor.
对吧？
Right?
我已经知道，我们从轻度用户到重度用户都有。
And I already know we have, like, light users all the way to power users.
我真的很想弄清楚的一件事是，使用上的自然分群是什么？
And one of the things I really wanted to figure out was, like, what are the natural clusters of usage?
你能在这支团队中找出这些分群吗？
Can you find them across the team?
将他们分组的最好方式是什么？
What is the best way to cohort them?
对吧？
Right?
我就先拿这个标准的分析文件，再这里加一个别的。
And I'm just gonna pick up the standard analytics file here, maybe pop in another one here.
我特别喜欢Opus High。
And then I love opus high.
我也喜欢计划模式，因为它能让你看到它的思考过程。
I also love plan mode because it gives you a chance to, like, see what it's thinking through.
所以我们先让它运行一下，看看它会返回什么结果。
So we can let this cook and see what it comes back with.
我想对工程经理或工程领导者强调的是，这种定量分析正是我们所有人曾经希望能在多个工程指标上实现的。
And what I wanna call out here for engineering managers or engineering leaders is this is the kind of quantitative analysis that we would all have loved to be able to do across a bunch of engineering metrics at at some point.
对吧？
Right?
比如，我们有多少次被董事会或上司问起：速度是多少？
Like, how often do we get asked by the board or our boss, like, what's velocity?
周期时间呢？
What cycle time?
我们的哪些工程师真正处于效率曲线的极端位置？
Which of our engineers are super you know, like, are are really at the far edge of the curve in terms of efficiency.
我们的初级工程师是如何逐步熟悉代码库的？
How are our junior engineers ramping into the repo?
所有这类问题。
All that kind of stuff.
而这类分析实际上非常繁琐且难以获取，原因在于数据的结构和分析本身的性质。
And that kind of analysis is actually really onerous and hard to get at because of the structure of the data and the nature of the analysis.
所以我特别喜欢LLM，尤其是使用Cursor这样的工具，因为它能让你以前所未有的方式，对人类行为和人类数据分析进行细致的分组分析。
And so what I love about just LLMs in general, and in particular using something like Cursor, is you can get to really nuanced cohorting analysis on human behavior and human analytics as a manager in a way that I think has been really challenging to do before.
是的。
Yeah.
我完全同意。
I totally agree.
而且，现在有了MCP和数据的可访问性，我觉得像Cursor这样的工具就是我的日常操作系统。
And, like, the beautiful thing is now with MCPs, with data accessibility, like, I think of tools like Cursor as just my daily operating system.
无论我有什么问题，不管是技术性的还是非技术性的，都没关系。
If I have a question, it doesn't matter if it's technical or not.
我只需要打开Cursor问它就行了。
I just go into Cursor and ask it.
因此，这种方式超级强大。
And so it's, like, super, super powerful that way.
好的。
Okay.
所以它在问我想要什么样的输出结果。
So it's asking me a little bit about, like, what outputs do I want.
我想丰富一下CSV文件，这样会更方便。
I do wanna enrich CSV just it makes it easier.
我也想做个静态仪表盘，纯粹为了好玩。
I do want a static dashboard just for fun.
我现在其实并不是真的想创建一个全新的仪表盘。
Like, I'm not really trying to create a brand new dashboard right now.
但我在这里的主要目标，老实说，就是找出自然的用户群体。
But my main goal here is just honestly honestly, like, find natural cohorts.
对吧？
Right?
所以它会试着识别出轻度、中度、活跃的超级用户。
And so it's gonna kinda try to do a light, moderate, active power superuser.
它会分析建议的指标，比如使用量、复杂度、代理模式、模型偏好、接受率和使用广度。
It's gonna look at line suggested, so volume, sophistication, agent mode, model preference, acceptance rate, and breadth.
他们都在使用哪些功能？
What features are they using?
我会输出一个CSV仪表板。
I'll spit out, you know, a CSV dashboard.
它还可能会生成一个我可重复使用的Python脚本。
It'll likely generate a Python script too that I can reuse.
所以我就先启动构建模式。
So I'm just gonna kick off build mode.
在它运行的时候，我想顺便去看看，它会用Python创建所有这些东西，为我生成脚本。
While that's cooking, I do wanna just maybe bop over to, like, it's gonna create all this stuff in Python, create the scripts for me.
太棒了。
Awesome.
但我们在这里可以看一下一些信息。
But we can look here at some of the the information.
对吧？
Right?
所以，这些都是一些随意编造的数据。
So, like, this is all sort of random made up data.
就像是示例数据。
It's like sample data.
但它在之前的运行中，分析了所有数据，生成了一个Python脚本，这很棒，非常简单。
But what it did was, in a previous run, it looked at all the data, generated the Python script, which is great, super simple.
它只是做了一些高层次的状态指标，比如AI代码占比、每周AI行数、composer行数，都是基于这些虚构的数据。
And it sort of just did some, like, high level status metrics, like AI code percentage, again, on all this made up data, AI lines per week, composer lines.
这是你在使用Cursor的代理模式时的情况，对吧？当你按下Tab键的时候。
This is when you're using the agent mode in cursor or tab lines, right, when you're hitting tab.
我的一个团队成员实际上获得了酷炫的Cursor Tab奖，这很棒。
One of my team members actually got the cool cursor tab award, which is which is great.
很好。
Great.
所以它把这些内容都逐一分解了。
And so it sort of breaks all this down.
而它真正细分出来的是重度使用代理的用户，也就是那些非常依赖代理功能的人。
And then what it really segmented around was agent heavy users, which is folks who really lean into agent usage.
还有重度使用 Tab 的用户。
There's also tab heavy users.
这是另一群不同的用户。
This is like a different cohort.
他们只是非常依赖 Tab 功能。
They just lean into tab usage.
他们可能只是想要更多的控制权，还没完全适应如何放手让代理来帮忙。
And they they maybe want really just a bit more control and maybe haven't gotten yet used to, like, how to let go with an agent.
你还有那些尝试两种方式的平衡型用户，然后可能还有一些对 Cursor 感兴趣、还没上瘾，或者目前还没被 LLM 影响的用户。
You have balance users that try both, and then you have sort of, like, maybe cursor curious or maybe not cursor pill or, you know, LLM pilled right now.
所以我生成了这个完整的脚本。
And so I generated this whole script.
这很棒。
It's great.
现在让我向你展示一下我打算在这里进行的更多分析。
And now let me show you sort of a bit more analysis I wanna do here.
我们来这么做吧。
So let's do this.
对我的一组示例用户运行分析，并生成HTML文件。
Run the analysis on I have a sample user set and generate the HTML as well.
实际上，这是由Python生成的分析脚本的输出结果，该脚本已经在并行运行中。
And let's we're actually like, this is sort of the output of the analysis script that was generated in Python, which is already cooking in parallel.
明白了。
Got it.
你在这里所做的，是从Cursor中提取了一些原始数据。
So what you've done here is you've taken some raw data from Cursor.
你让一个代理对用户群进行分组分析，并生成一个包含额外数据的增强型CSV文件。
You've asked one kind of agent to do a cohort based analysis and generate a enriched CSV essentially with some data.
然后你启动另一个代理，对这些数据进行分析，并生成一个HTML视图，以便可视化数据。
And then you're kicking off another agent to actually do the analysis on that and generate sort of an HTML view of it so you can visualize the data.
没错。
That's right.
没错。
That's right.
它所做的就是那个生成的Python脚本。
What it did was the Python script that was generated.
对吧？
Right?
它找到了这些自然的用户群体：超级用户、普通用户、高级用户、轻度用户和不活跃用户。
It found these natural cohorts, these natural cohorts of super user, regular user, power user, light, inactive.
再说一遍，这仅仅是示例数据，但基于真实的信息、真实的数据结构和真实的Cursor数据字段。
Again, this is just honestly sample data, but based on like real information, real schema, real cursor data fields.
它得出的结果是：样本数据中70%是代理密集型，20%是最低限度型，4%是平衡型。
And it came up with like 70% or an agent heavy in the sample data, 20% are minimal, 4% are balanced.
我们在样本数据上还有一些改进空间。
We have some room to improve here on the sample.
对吧？
Right?
像，用的人不够多。
Like, not enough people are using it.
所以它做了一个简单的分析，我挺喜欢的，算是对指标的一个回顾。
And so it does a bit of a breakdown, which I kinda like, you know, kind of a recap of metrics.
是的。
Yeah.
这些数据中有大量的代码行。
We have a lot of lines of code in this data.
我们有520个高级用户。
We have 520 power users.
同样是虚构的名字，但这个人表现得非常出色。
Again, made up names, but, like, this person is crushing it.
我想知道这个虚构的人物加布里埃尔·迪亚兹在做什么。
I wanna know what this made up person Gabriel Diaz is doing.
对吧？
Right?
这里有个很棒的地方。
Awesome thing here.
它生成了一个简单的可视化仪表板。
It generated a little visual dashboard.
没什么花哨的。
Nothing fancy.
只是非常简单地看一下。
Something just really simple to look at.
对吧？
Right?
总行数、composer行数、标签补全、一些细分，以及分层和使用情况的结构化。
Total lines, composer lines, tab completion, a little bit of breakdown, some structuring on the tiers and usage.
对吧？
Right?
但我真正想弄清楚的是，加布里埃尔·迪亚兹到底在做什么？
But what I really kinda wanna understand is, like, what is Gabriel Diaz doing?
对吧？
Right?
这个虚构的用户简直太厉害了。
This made up user who's just, like, crushing it.
是的。
Yep.
根据数据，为每个用户群体生成指导建议，告诉他们该如何进步并升级为超级用户。
How about based on the data, generate guidance for each user cohort, what, you know, they should do to advance and graduate to super user.
我在寻找明确的指导。
I'm looking for explicit guidance.
换句话说，我想把这变成一份操作手册。
Effectively, like, I wanna turn this into some type of playbook.
对吧？
Right?
那我们先让这个过程自然发展。
So let's let this cook.
同时，我也希望加入一些可视化内容，因为从数据本身来看，有一个很直观的点：我们都知道，成为超级用户的过程并不是简单地从不活跃到轻度使用，再到普通用户、高级用户，最后到超级用户这样线性推进的。
And then in parallel, what I also wanna do is I like visuals, and there's something intuitive here where, like, as we look at the data itself, right, we we know that the, like, the path to this super user over here it's it's not like you go inactive to to light to regular to power to super.
我们知道，这个过程并不是线性的。
We know it's not linear like that.
对吧？
Right?
对吧？
Right?
可能会有从轻度用户直接跳到高级用户的分支路径。
There may be, like, forks from light to straight to power user.
普通用户似乎在各个层级之间保持着平衡。
Regular user seems to be, like, balanced on the tiering.
但我想知道的是，这些人究竟在做些什么特别的事情，以及我该如何推动整个曲线的转变？
But what I wanna know is, like, what are the special things these folks are doing, and how do I sort of shift the curve?
对吧？
Right?
所以我也想同时提出另一个问题：为用户从轻度到重度使用的各种路径创建一个Mermaid图，我假设这些路径不是线性的。
And so I'm also gonna throw another question in parallel, like create a mermaid diagram for all the different sort of paths a user can take from light to power, and it's I'm assuming it's not linear.
我们来看看这会生成什么结果。
And let's just see what this cooks up to.
好的。
Okay.
这真的在全力运行。
This is really working hard.
真的在全力运行
Really working
运行。
hard.
四五。
Four five.
是的。
Yeah.
Opus 在这上面确实非常努力，不过，咱们还是看看它会发展成什么样吧。
Opus is Opus is really working hard on this, but, yeah, let's let's see where it goes.
嗯，这其实非常有趣。
Well, you know, it's really interesting.
我来给你一个更简短的技巧来处理这个。
I'll give you a a a shorter hack on this one.
所以我认为它生成的是一个可以分享的 HTML 操作手册，里面包含了一些内容。
So I think what this is generating is like an HTML playbook that you could share out that has has things.
我来告诉你在这种情况下我会怎么做，我之前在客户季度业务回顾中也做过几次：我会写一篇 Slack 帖子，发到我的工程频道里，分享一些数据，以及我们如何帮助人们从 A 走到 B。
I will tell you what I would do in this use case, and I've done this a couple times with, like customer QBRs is I say write a Slack post that I can put in my engineering channel on a couple of these stats and, you know, how we can get people to move from A to B.
它会帮我生成一篇简短的 Slack 帖子。
And it'll write me like a short little Slack post.
所以我特别喜欢这个想法：从一个 CSV 文件开始，生成深入的分析，再变成 HTML 可视化，最后提炼出三条可以发在 Slack 里的要点。
I So I love this idea of going from something like a CSV to a really deep analysis to an HTML like visualization to, like, three bullet points I can send in Slack.
是的。
Yeah.
作为管理者，每一个步骤过去都要花上很久才能完成。
And as a manager, each one of those steps would have taken just forever to do.
但现在你可以在 Cursor 里一次性完成所有这些步骤。
And now you can get them all done in Cursor.
是的。
Yeah.
你知道吗，这正是像工作流 Markdown 文件这样的工具的厉害之处，确实如此。
You know, that is that's, like, kind of the awesome thing is the power of something like a workflow markdown file Yeah.
这种力量非常巨大。
Is huge.
它确实非常强大，完全就是你在这里描述的那种情况。
It's absolutely huge, and it is it's exactly like the thing you're describing here.
认识 Rovo，你的 AI 同事，连接知识、人员与工作流程，让团队更聪明地工作，更快地推进。
Meet Rovo, your AI teammate, connecting knowledge, people, and workflows so teams can work smarter and move faster.
它通过搜索、聊天、智能代理和工作室，帮助人们安全地、结合上下文地找到答案、做出决策并自动化工作。
It helps people find answers, make decisions, decisions, and automate work securely and with context through search, chat, agents, and studio.
Rovo 基于 Teamwork Graph——Atlassian 的智能层，统一您第一方和第三方应用中的数据，确保没有任何知识被遗漏。
Rovo runs on the Teamwork Graph, Atlassian's intelligent layer that unifies unifies data across your first and third party apps so no knowledge gets left behind.
而且从第一天起，您就能获得个性化的 AI 洞察。
And you always get personalized AI insights from day one.
最棒的是，它已内置在 Jira、Confluence 和 Jira Service Management 的付费订阅中，因此 Rovo 的强大功能触手可及。
And the best news, it's already built into Jira, Confluence, and Jira Service Management paid subscriptions, so the power of Rovo is already at your fingertips.
你有没有那种感觉：AI 从工具变成了队友？
Know the feeling when AI turns from tool to teammate?
如果你用了 Rovo，你就懂了。
If you Rovo, you know.
探索 Rovo——由 Atlassian 驱动、了解您业务的 AI。
Discover Rovo, AI that knows your business powered by Atlassian.
立即前往 rovo.com 开始使用。
Get started at rovo.com.
那是 ro,vasinvictory,o.com。
That's ro,vasinvictory,o.com.
我们来看看。
Let's see.
我们看看它得出了什么结果。
Let's see what it came up with.
对吧？
Right?
而且，你知道，关键是，没人应该指望所有这些信息都完美无缺。
And, like, you know, the the thing is, like, no one should expect all this information is going to be perfect.
就像，任何人可能会想，哇哦。
Like, anyone is thinking, oh, wow.
如果 Cursor 能做所有这些，那我作为领导者的职责又是什么？
What is going to be my job as a leader if Cursor can do all all of this?
而实际上，你作为领导者的职责就是带领团队、推动变革并产生影响，对吧？
And it's like, well, your job as a leader is to lead, right, and to make change and and impact.
这加速了他们的进程。
And this accelerates them.
不活跃的用户，嗯，确实有点道理。
Inactive users, like, yeah, kind of true.
你还没安装。
You haven't installed.
你还没真正使用过AI功能。
You haven't really used AI features yet.
最难的部分是开始。
The hardest part is getting started.
所以我挺喜欢这个的。
So I kinda like this.
它提供了一些非常简单的提示。
It it gives like just some very simple prompts.
为你的下一个任务试试代理模式。
Try the agent mode for your next task.
发送一些非常简单的东西，轻量且紧凑。
Something very, very simple sent something lightweight, tight.
试试看。
Try it.
标签补全流程。
Tab completion flow.
我总觉得这个大语言模型真的想把这变成一个游戏，像一个小任务或类似的东西。
And I kinda feel like the the LLM really wanted to just turn this into a game, like a little like a little quest or something.
是的。
Yeah.
它有点游戏化了。
It's gamified a little bit.
对。
Yeah.
确实有点游戏化，还挺有趣的。
It is it is a bit gamified, and it's kinda fun.
好的。
Alright.
这很酷。
So this is cool.
它让我觉得，这会是我的Slack帖子的摘要。
It's kinda given me, like this would be my Slack post t l d r.
AI高级用户是其他用户的16倍。
16 x more AI line super users versus other users.
让我再放大一点。
Let me zoom in just a bit more.
高级用户的代理请求更多。
More agent requests for super users.
我喜欢。
I love it.
别打了。
Stop typing.
开始交付。
Start shipping.
这是深色模式，工程师们一定会喜欢。
It's dark mode, the engineers will just love it.
是的。
Yes.
对吧？
Right?
这简直完美。
It it's kinda perfect.
然后你安装了 Cursor，但还没用过 AI。
And then you installed Cursor, but you haven't used AI yet.
我们聊过这个。
We talked about this.
这很酷。
That's cool.
亮色模式。
Lightmode.
好的。
Okay.
你知道吗？
I you know what?
这个，真的很贴切。
This, like, resonates.
别再说‘修复这个bug’了。
Stop saying fix this bug.
实际上，就像你跟一个初级工程师说话那样跟它交流。
Actually, like, talk to it like you would maybe a junior engineer.
对吧？
Right?
Cursor 刚刚发布了 bugbots。
Cursor just did release bugbots.
我爱bugbot。
I love bugbot.
是的。
Yep.
太棒了。
It's awesome.
我爱我爱bugbot。
I love I love bugbot.
智能代理不适合处理复杂问题。
Agent isn't for hard stuff.
它适合所有事情。
It's for everything.
这些现在简直就是励志语录了，我想我们应该把它们做成海报贴在墙上。
These are, like, motivational quotes now that I think, like, we should just make posters for and put them up on the wall.
编写单元测试。
Write unit tests.
实际上读一下评论。
Actually read the comments.
好的。
Okay.
不错。
Cool.
现在说说高级用户。
Now power users.
你没问题。
You're good.
要变得优秀，就要想得更远，更用力地使用标签。
To be great, think bigger, and tab harder.
更用力地使用标签。
Tab harder.
好的。
Okay.
如果Cursor在听的话，我觉得这可能会成为你们的新周边系列，各位。
If Cursor is listening, I think this is, like, gonna be your new merch line, guys.
我需要一顶写着‘tab harder’的帽子。
I need a hat that says tab harder.
是的。
Yes.
好的。
Okay.
所以再总结一下，我们在这里为Cursor免费做产品推广。
So just to just to recap again, we're doing free product work for cursor here.
你最初的问题是，如何提升这些工具的采用率？
Your ultimate problem was like, how do I drive up adoption of these tools?
而你的想法是，当然要用这个工具来分析采用情况，然后找出提升采用率的方法。
And you're like, of course, I'm gonna use the tool to understand adoption and then figure out ways to drive adoption.
我们做了分析。
We did analysis.
我们创建了数据本身的可视化。
We created a visualization of the the data itself.
你识别出了用户群体和核心用户，如果手动来做的话，这会非常繁琐。
You identified cohorts and power users, which would have been very tedious to do if you were gonna do manually.
是的。
Yeah.
然后你还创建了一个托管的操作指南以及一系列激励性语句，我们可以免费提供给Cursor的朋友们，或者现在就注册商标，赚点小钱。
And then you created a hosted playbook as well as a series of motivational statements, which we can either give to our friends at Cursor for free or trademark right now and make a little a little money.
一切皆代理，无意识敲击Tab，始终开启bug.dot，迭代提示词。
Agent everything, tab without thinking, bug dot always on, iterate prompts.
太棒了。
Love it.
不错。
It's good.
而且你知道，再次强调，我觉得有趣的是，让我谈谈我觉得这部分有趣的地方。
And this you know, again, what I think is fun let me talk about what I think is fun about this.
第一，所有有过工程管理经验的人都知道，这类内容正是你被要求在董事会会议上汇报的。
One, everybody who has been in engineering leadership knows this is the kind of stuff you get asked to put in a board meeting.
你的老板会问你：我们有多少比例的工程师在使用 Cursor？
You get asked by your boss, like what percentage of our engineers are using Cursor?
我们有核心用户吗？
Do we have power users?
我们真的从中获得了价值吗？
Are we actually getting value?
我们现在讨论的是一个 AI 使用场景。
And we're talking about an AI use case right now.
但同样地，在管理层面，实际上有很多可量化的手段可以提升团队的绩效和效率。
But, again, across management, there are actually measurable things you can do about the performance and efficiency of your team.
是的。
Yes.
我觉得以前这根本是不可能实现的。
And I think it's so impossible to get before.
第二，如果你不能用代码来做这件事，那就会没意思，而你现在就可以用代码来做。
Two, it would be no fun if you didn't get to do it with code, which you get to do.
现在你可以
Now you get
用代码来做
to do it
用代码了。
with code.
这正是
That is
关键所在。
the thing.
你现在仅用代码就能解决问题了。
You can problems with just code now.
对吧？
Right?
你可以直接做事情。
You can just do things.
我，你知道，你说得对。
I I you you know, you're so right.
我觉得我之前低估了你现在的说法，我想再重复一遍。
Like, I I think this I underappreciated what exactly what you're saying right now, and and I just wanna repeat it.
因为通常情况下，你会被问到这个问题，然后你得去叫个IC来处理。
Because normally, you would be asked this, and then you would have to go pull an IC to do that.
我在想，什么？
I'm like, what?
什么？
What?
是的。
Yeah.
拜托了。
Like, come on.
不，不是这样的。
Like, no.
你可以直接去做事情。
You can just do things.
对吧？
Right?
不。
No.
而且，再说一遍，人们往往低估了完成一项有趣任务所带来的效率提升。
And and, again, it's, like, not the I I I think people underappreciate the velocity creation of a fun task.
是的。
Yeah.
说到底，这虽然有点傻，但那些小小的趣味环节确实让人感觉很棒。
Which is like, at the end of the day, like, this is silly, but also the, like, little fun bits of it, you're like, great.
我想更进一步，因为我从这个暗黑模式手册里得到了一点多巴胺的满足感，还挺有趣的。
I wanna go to the next level because I got, like, a little dopamine hit from this dark mode playbook that's kinda funny.
我认为人们低估了这种快速反馈循环所带来的迭代速度。
And I I think people underappreciate, like, that iteration speed that can just come with, like, a fast feedback loop.
是的。
Yeah.
当你在构建某样东西时，如果有一个快速反馈循环，而且你追求的是高质量，就像这种设计，它看起来比谷歌文档有趣多了。
When you're building something and the fast feedback loop when you're building something that has high quality against it, which, like, something designed like this does so much more fun to look at than a Google Doc
是的。
Yeah.
我完全同意。
I totally agree.
一个电子表格或仪表板。
A spreadsheet or a dashboard.
所以我们赢了，我们做到了。
So winning we did it.
我们做到了。
We did it.
我们俩，你和我，就像是双子星，我觉得。
We again, you and I are twin stars, I think, here.
所以我们可能能聊上一整天我们觉得有趣的事情。
And so we can probably go all day on the things that we find fun.
但让我们转向第二个人们将会看到的使用场景。
But let's go to a second use case that I think people are gonna see.
让我们看看我们能多快完成这个使用场景，也就是你谈到的反馈到功能的响应速度。
And let's see how fast we can do this use case, which is you're talking about the speed of feedback to feature.
你刚才说了一些颇具挑战性的话。
And you you said some fighting words out there.
你说我们真的在大幅压缩从反馈到功能实现的时间。
You're like, we're really compressing the time from feedback to feature.
那这到底是如何实现的呢？
So how does that actually work?
那些确实是颇具挑战性的话。
Those were those were some fighting words.
而且，你知道，我觉得你也明白这一点。
And, you know, I think you know this.
对吧？
Right?
你希望为你的用户打造这个功能。
You want wanna build this for your users.
对吧？
Right?
你希望尽可能快地做出最棒的产品。
And you want to create the best damn product out there as fast as possible.
而要让这个循环高效运转，关键就在于你对反馈的响应速度有多快。
And the way to, like, make that cycle work really well is genuinely how fast you can move on feedback.
好的。
Okay.
但我想先从反馈通常是怎么来的开始说起。
But I want to start from how does feedback even normally come in?
对吧？
Right?
通常来说，团队在文化上会进行自用测试或缺陷集中排查活动。
So normal teams and culturally, you'll have dogfooding or bug bash sessions.
对吧？
Right?
你们会开个会或者聚在一间房里，继续使用产品，诸如此类的各种操作。
You'll get on a meet or get in a room, keep using the product, blah blah blah, all that jazz.
然后有人得把这些问题收集到一个Google文档里，再把Google文档里的问题录入到工单系统中。
And then someone has to collect the bugs in a Google Doc and then take those bugs in a Google Doc and put them into a ticket system.
对吧？
Right?
好的。
Okay.
接着就会有一番讨论，关于这个问题重要不重要。
And then there's a whole discussion around, is this important?
这不重要吗？
Is this not important?
好的。
Okay.
我们这次迭代要处理这个问题吗？
Should we pick it up in this sprint?
我们要等到下一个迭代吗？
Should we wait for another sprint?
到那时，你的用户已经流失了。
And by that time, your user has churned out.
他们会说：你们都没修这个问题。
They're like, you guys didn't fix this.
我有点讨厌这样。
I kinda hate it.
继续吧。
Moving on.
对吧？
Right?
每个人的注意力都特别、特别、特别短。
Everyone's attention is, like, so, so, so short.
现在，整个团队都在为一次重大发布做准备，我们想聚在一起搞一次叫作‘冲刺’的活动。
And right now, like, the whole team, we're all preparing for a big launch, And we wanted to get together and do this thing called a surge.
在这种冲刺中，我们会把团队聚在一起，每天工作非常、非常长，利用所有AI工具，大量地交付代码。
And this is where we, like, just bring the team together, and we do very, very long days using all this AI and just shipping, like, massive amounts of code.
有趣的是，在这些冲刺期间，我们在相同时间内交付的PR数量会达到平时的三到四倍。
And fun fact, during these surges, we end up shipping more than three to 4x more PR volume in the same time.
但我们还想做的一件事是让员工回到办公室，于是我们设立了叫作‘反馈咖啡角’的活动。
But the other thing we wanted to do was bring people into the office, and we set up this thing called a feedback cafe.
我们会邀请外部用户、内部员工等，和他们一起试用产品，向他们展示应用。
So we'd invite externals, internals, etcetera, and we'd dog food with them, and we'd show them the app.
这里只是给你看几秒钟，你知道的，实际是什么样子。
And, like, here's just, like, a couple seconds of, you know, what it looks like.
我们就站在那里收集信息，进行这种实时的实地体验测试。
We're just standing there collecting information, doing all this, like, live dog footing.
但难点在于，尤其是在现实生活中。
And the hard part, though, is especially in real life.
你究竟该如何捕捉这些信息呢？
How do you actually capture that information?
因为这是语音和视频。
Because it's voice, it's video.
你如何将其转化为系统可处理的内容？
How do you translate it into a system?
好的。
Okay.
所以我花了整整半个周末，开发了一个用于实时收集反馈的工具。
So I just spent, like, half a weekend and built a tool to capture feedback live.
我们随便选一个吧。
Let's just pick something.
我要选一个新东西。
I'm gonna pick I'm gonna I'm gonna pick a new thing.
我是怎么用AI的。
How I AI.
和克莱尔一起测试。
Testing with Claire.
太棒了。
Awesome.
那我们就这么做吧。
So let's do that.
它会创建一个小型会话。
It's gonna create a little session.
完美。
Perfect.
非常简单。
Very simple.
我们有两种模式。
And we have two modes.
你可以像在手机上使用这个功能。
You can, like you can use this on your mobile phone.
团队在现实生活中就是这么做的。
That's what the team did when they were in real life.
但在这里，我只是想录一些音频，我们来看看。
But for this, I'm just gonna, like, capture some audio, and let's see.
你有没有遇到什么有趣的小bug，或者有什么产品问题是你想修复的？
What's what's actually, maybe I can just hear from you, like, a fun little bug or something of a product that you you think you wanna fix.
我们现在开始录制音频。
So we're gonna start capturing audio.
我用的一个AI聊天机器人，当我把账户切换为企业账户时，它强制我清空所有聊天记录。
There is a AI chatbot that I use where my account, when switched to business account, forces me to clear all my chats.
我觉得我们应该修复这个bug，让我能继续访问之前的聊天记录。
And I think we should fix that bug so that I can access my existing chats.
我们开始录制音频。
We're gonna start capturing audio.
好的，我们开始。
We're gonna Okay.
太棒了。
Cool.
我们录到了。
We captured it.
它基本上是在获取音频。
It's basically taking the audio.
我写了一个系统提示词，发送给了一个大语言模型。
I did a system prompt, sent it to an LLM.
嗯哼。
Mhmm.
然后我们做的就是，这个提示词基本上是说：去识别这些bug。
And then what we do is it the prompt is basically saying go and identify the bugs.
是的。
Yep.
对吧？
Right?
然后我会创建它。
And then I'll create it.
在它处理的时候，我要做一个。
I'm gonna do one while it's processing.
现在，我正在使用这个应用。
Right now, I'm using the app.
我在交易标签页，点击了来源字段，输入数字，但数字没有显示出来。
I'm on the trade tab, and I'm clicking the from field, and I'm typing in numbers, but the numbers are not showing up.
所以这导致我无法进行交易。
So that's not letting me make a trade.
所以我认为在我们的第一个例子中，音频很难捕捉，因为它是通过系统传输的，但让我们看看第二个例子。
So I think in our first example, the audio is a little hard to capture just because it's going through the system, but let's look at the second example.
它明确地指出了这个问题。
It calls it out really clearly.
在交易标签页中，在“来自”字段输入时，数字不会显示。
On trade tab, typing into from field does not display under numbers.
用户无法发起交易。
User cannot initiate a trade.
不错。
Cool.
非常简洁明了。
Really, really clean.
是的。
Yep.
我点击了创建线性票据。
I hit create linear ticket.
它甚至还会给出一个建议的标题。
It even gives, like, a suggested title.
我关心的用户流程是交易。
The user journey I care about for this is trade.
搞定。
Boom.
我创建了这个工单。
I create the ticket itself.
太棒了。
Awesome.
我切换过去。
I pop over.
工单全在这里。
The ticket is all here.
文件也在那里。
The file is there.
Linear 是一个了不起的工具。
Linear is a incredible tool.
它正在做一些初步分类。
It's doing some triaging.
但我现在想转到的是，我们要直接创建拉取请求。
But the thing I wanna now hop over to is we're gonna just create the PR.
所以我们自己开发了一个工具。
So we have this tool we built in house.
我们叫它Cloudbot。
We call it Cloudbot.
它实际上使用了各种底层模型。
It's actually, like, using all sorts of underlying models.
它并不是专门针对Claude的。
It's not something that is specific to to to Claude.
所以Cloudbot创建了拉取请求。
So Cloudbot create PR.
我知道这个仓库是Wallet Mobile，这是对应的工单。
I know the repo for this is Wallet Mobile, and here's the ticket.
哦，这不是那个工单。
Oh, that's not the ticket.
工单是boom。
The ticket is boom.
在这里。
Here.
很好。
Great.
不错。
Cool.
所以我刚从一个错误报告转到了工单
So I just went from a bug report to a ticket
到PR。
To a PR.
PR正在处理中。
To the PR is cooking.
好的。
Okay.
所以我得暂停一下，因为如果你刚接触我的AI方式，可能还没见过我特别喜欢某样东西时的标志性动作，就是这个。
So I have to pause because if you are new to how I AI, you have not seen my signature move when I really love something, which is this.
我刚才在想你左边那个小微应用，也就是实时用户反馈，完全非结构化的，对吧？
And I was doing this because I was just thinking about this little micro app that you have on the left side, which is, you know, live user feedback, totally unstructured, right?
视频或音频，跑一个小小的LLM模型。
Video or audio, run a little baby LLM on it.
不仅能总结问题，还能给出如何修复的优质建议。
Get not only a summary of the issue, but a good recommendation on how you might fix it.
快速 beep boop 一下 Linear。
Very quick beep boop to Linear.
我们非常喜爱 Linear 的团队。
We love our friends at Linear.
我认为它是为代理工具设计的绝佳平台。
I think it's a great platform for agents.
然后在你的Slack中设置一个自定义代理，可以读取这些Linear工单并直接执行。
And then a little custom agent in your Slack that can read those Linear tickets and just execute on them.
而且，可能因为过去经历得太糟糕了，这个流程原本需要有人手动总结
And again, so traumatized by the past, maybe, which is like, this process would have been, you know, somebody manually summarizing
是的。
Yep.
来自研究会议的结果。
What came out of a research session.
是的。
Yep.
某份文档的撰写。
Some document being written.
有人明确决定哪些内容要包含、哪些不包含。
Somebody actually making explicit decisions about what to include and not include.
我觉得这应该稍微有点
Think that should be a little bit of
比如，再做筛选了。
Like, filter anymore.
你不会觉得，嗯，如果我把它写成五页，没人会读的。
You don't get that like, well, you know, if I make this five pages long, no one's gonna read it.
所以我真的要专注于最重要的前十件事。
So I'm really gonna focus on the top 10 things.
就是，把所有东西都收集起来，然后快速处理掉。
It's like, let's capture everything and then just burn through it.
然后我得问你们，你们为什么要自己开发一个小机器人来做这件事？
And then I have to ask you, why did you all build your own little bot to do this?
开发这个机器人的优势是什么？
What was the advantage of of building the bot?
这个是我们内部自己开发的。
So this this is, like, in house, and we built it.
你知道，这一切都始于今年年中。
You know, it all started around, like, middle of this year.
我创建了这个，因为我当时对人工智能着迷得不行。
I created this, like I I was just obsessing so much about AI.
我在想，怎么才能为团队、为公司打造更好的工具，让每个人都能加速进步？
And I was like, how do I create better tooling for the team, for the company so everyone can be accelerated?
所以我实际上在Twitter上发起了一个招募。
So I invented actually I put a call out on Twitter.
我创造了一个叫‘超级构建者’的角色。
I invented this role called super builder.
超级构建者唯一、最重要的职责就是培养更多的超级构建者。
And the single job single most important job of a super builder is to create more super builders.
于是我们雇了第一位超级构建者，我们讨论了一些想法。
So we we hired our first super builder, and they we we talked about some ideas.
其中最重要的一点是，我们公司大多数人用的都是Slack。
And one of the biggest things because most of our company uses Slack.
我们都用Slack。
We're all in Slack.
Slack，你知道的，我坚信它其实就是一群人类假装自己是系统。
And Slack, you know, I am it's, like, strong believer is just a bunch of humans pretending to be systems.
对吧？
Right?
在Slack里发消息的成本是零，但回复消息的成本却极高，而且大部分都是噪音。
And the cost of writing something in Slack is zero, but the cost of answering something in Slack is enormous, and most of it is noise.
对吧？
Right?
所以我们当时就在想，怎么把我们熟悉的那些工作流程带进来，把它们捕捉下来，再在上面叠加AI？
And so one of the things was just like, how do we bring the workflows that we are also used to, and how do we, like, sort of capture that and then add AI on top of it?
所以我们当时有各种各样的考虑。
So we had, like, various reason.
我们知道很多公司都在用后台代理、Cursor之类的工具。
We know, like, lots of companies have background agents, cursor, etcetera, etcetera.
但我们目前的安全要求不同，暂时没法上线这些工具，这也没关系。
We just have, like, different sort of security requirements right now that we just couldn't launch with it, and that's fine.
所以我们内部开发了这个系统，并设置了这些反馈渠道。
So we we built this in house, and we have these, like, feedback channels.
对吧？
Right?
嘿。
Hey.
这里有个bug。
There's a bug here.
这里有个bug。
There's a bug here.
所以我们现在只需要让Cloudbot去处理这些问题。
And so now all we just do is, like, Cloudbot go and do something with that.
或者如果有人说，嘿。
Or if someone is, like, hey.
我们刚开完这个会。
We just got out of this meeting.
这是摘要后的转录内容。
Here's a summarized transcript.
我们觉得太棒了。
We're like, awesome.
在Linear Agent里，把这个拆分成几个任务。
At linear agent, go break this down into tickets.
然后你就知道，你刚才那个表情，对吧？
And then just like, you know you know the look you you showed, like right?
每个人都在用那个头爆炸的表情。
Like, everyone is just doing that emoji of, like, the head exploding.
对吧？
Right?
因为现在我们有了二十多个任务，然后我们就做些疯狂的事，比如启动大量调用，所以我们开发了这个计划模式。
Because then now we have, like, 20 tickets, and then we do fun things like this, which is just go like bonkers where we just fire off tons and tons of calls, right, to just and so we we built this plan mode.
这个机器人有个创建拉取请求的功能，我现在正在运行它。
So this bot has a create PR, which I'm it's cooking.
它还有一个功能，创建拉取请求的亮点在于完成后会自动回复。
It has a and, and also the cool thing about create PR is when it's done, it will respond back.
它会显示一个链接，指向使用 Cursor 深链接的分支。
It will show you a link to, like, the cursor branch using cursor's deep link.
当一次性构建完成时，它会显示二维码，你可以直接扫描并开始测试修复效果。
And when then the one off build is ready, it will show the QR code so you can just scan and start playing with the fix.
对吧？
Right?
还有一个计划模式，和 Cursor 的计划模式非常相似。
There's a plan mode, which is very much like Cursor's plan mode.
它会自动生成一个计划。
It just comes up with, like, a plan.
我们还提供了解释功能，比如我想调试某个问题。
And then we also have explain as well where it's like, oh, I wanna debug something.
所以，为什么 Ab 现在不能正常工作呢？
So, like, why is Ab not working right now?
举个例子。
As an example.
对吧？
Right?
而且它具备所有技能和所有MCP。
And it has, like, all the skills, all the MCPs.
所以我意识到，上下文才是最重要的。
And so the thing the thing I realized is context is the most important thing.
我们收集所有上下文的地方是Linear。
So the place where we capture all of our context is linear.
然后我们构建的这个代理，添加了技能和MCP。
And then this agent that we built, we added skills and MCPs.
如果我们能通过Linear捕获上下文，就可以利用Linear中的全部上下文来触发代理，然后它会访问所有MCP，比如Datadog、Sentry、Amplitude，以及我们内部的Snowflake数据库等等。
So if we can capture context through linear, then we can trigger the agent using all the context from linear, and then it goes off into all the MCPs like Datadog, Sentry, Amplitude, our internal Snowflake databases, etcetera.
它能够从公司其他地方提取上下文，并能在多个代码库之间协同工作。
And it has the ability to pull context from the rest of the company, and it can work across multiple code bases.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。