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。
史蒂夫·耶格已经是一名软件工程师四十年了。
Steve Yegge has been a software engineer for forty years.
他在亚马逊和谷歌工作了数十年，以对行业直言不讳的批评和经常正确的判断而闻名。
He spent decades at Amazon and Google, is famous for his brutally honest rants about the industry, and for being right a lot.
他最近开发了开源AI代理协调器Gas Town，并与吉恩·金合著了《氛围编程》一书。
He recently built Gas Town, an open source AI agent orchestrator, and co authored the book Vibe Coding with Gene Kim.
在今天的对话中，我们讨论了史蒂夫提出的工程师AI采用八个层级，从完全不用AI到并行运行多个代理，以及为什么70%的工程师仍停留在最低层级。
In today's conversation, we discuss: Steve's eight levels of AI adoption for engineers, from no AI to running multiple agents in parallel, and why 70% of engineers are still stuck at the bottom levels.
AI正在对开发者造成一种吸血式的倦怠效应：你可以变得一百倍高效，但每天只能获得三个高质量的工作小时。
Why AI is creating a vampiric burnout effect on developers where you can be 100 times more productive but only get three good hours a day.
他预测大型科技公司正在悄然衰落，而由两到二十人组成的小团队将能与之匹敌产出。
His prediction that big tech companies are quietly dying and that small teams of two-twenty people will rival their output.
还有更多内容。
And many more.
如果你想了解未来日常软件开发的真实面貌，以及如何不被时代抛下，这一集就是为你准备的。
If you want to understand what the day to day of software looks like in the near future and how not to get left behind, this episode is for you.
本集由Statsig赞助播出，Statsig是一个集标志、分析、实验等功能于一体的统一平台。
This episode is presented by Statsig, the unified platform for flags, analytics, experiments and more.
请查看节目说明，了解更多关于Statsig以及我们其他资深赞助商Sonar和WorkOS的信息。
Check out the show notes to learn more about them and our other seasoned sponsors Sonar and WorkOS.
所以，Steve，很高兴你再次做客我们的播客。
So, Steve, really good to have you on the podcast again.
你最近在忙些什么？
What have you been up to?
Greg，很高兴回来。
Greg, great to be back.
已经十个月了。
It's been ten months now.
快一年了。
Closer to a year.
是的。
Yeah.
快一年了。
Close to a year.
是的。
Yeah.
天啊。
Boy.
感觉像过了好多年。
Seems like forever.
是的。
Yeah.
确实如此。
Sure does.
是的。
Yeah.
发生了很多事情。
It's there's been a lot going on.
我现在失业了，这可真是超级有趣。
Unemployed right now, which has been incredibly fun.
是失业了，还是‘快乐失业’？
Unemployed or funemployed?
我现在就是想干嘛就干嘛，这感觉真的很好。
I am just doing whatever I want is what I'm doing, which is real nice.
我还发布了几个软件，挺不错的。
And had a couple of software launches, which was nice.
去年我还出了一本书，感觉很棒。
I had a book launch last year, which was nice.
我一直在好好生活。
I've been living life.
是的。
Yeah.
很长一段时间以来，你一直以直言不讳著称，常常提出一些滑稽的、有时却令人不安的事实或观察。
So for a very long time, you've been known as this kind of truth teller of bringing in sometimes comical, sometimes really uncomfortable facts or observations, should I say.
你经常以非常有趣的方式写一些长篇大论，很多内容都引起了人们的共鸣。
You wrote, like, often in really kind of fun fun ways with rants, and a lot of them resonated with people.
你还记得哪一篇长篇大论特别突出吗？当时或之后有没有收到过特别好的反馈，让你觉得被认可了？
Do you remember what was a rant that really stood out and at any point in time that, like, you got some really good feedback either at that point or later you felt validated by it?
哦，很多人告诉我，那些熟悉的人最喜欢的是斯蒂维的博客《名词王国中的执行》。
Oh, well, so a lot of people tell me, well, those who know their favorite Stevie blog is actually execution in the kingdom of nouns.
你不记得那一篇了吗？
I don't know if you remember that one.
那是很久以前的事了。
Way back in the day.
那时我还在谷歌，刚起步的时候，我努力想让别人明白，Java的代码量增长与功能增长是呈线性关系的。
I was at Google, early days Google, and I was trying I was struggling to sort of, like, get this idea across to people that Java's growth was super linear with the amount of code.
也就是说，代码量的增长超过了功能的增长，这可不是什么好现象。
So the amount of code would grow more than the amount of functionality, which is not a good place to be.
从那以后，Java已经改进了很多。
And got Java's gotten a lot better since then.
对吧？
Right?
但我的帖子在索尼引起了很大争议，他们觉得：这人到底在抱怨什么？
But my post raised a lot of eyebrows at Son because they were like, what is this guy complaining about?
他干嘛不闭嘴呢？
Why is it he just shut up?
你知道的？
You know?
但我就是想用一种支持一等函数的语言。
But I was like, I wanna use a language that has first class functions.
于是我写了一篇非常非常非常特别的博文，叫《名词王国中的执行》。
And so I wrote a very, very, very unusual blog post called Execution in the Kingdom of Nouns.
人们非常喜欢它，因为那是一个故事。
People really loved it, where it was a story.
这是一个关于一个没有动词的国度的童话故事，没错。
It's just a a fairy tale about a land where there were no verbs, and it was a yeah.
这很有趣。
It was fun.
所以你有一篇不太为人所知的博客文章，对很多听众来说，它被称为《富程序员美食》。
So one of your lesser known blog posts or for a lot of listeners, it's called a rich programmer food essay.
富程序员美食。
Rich programmer food.
是的。
Yeah.
这篇文章是关于编译器的。
And this was about compilers.
你还记得你当时争论了什么，或者你提出了哪些观点吗？
Do you remember what you argued about or what the points you made?
当然记得。
Of course.
这是我最重要的博客文章之一。
That's one of my most important blog posts ever.
我得告诉你，我在纽约的SWIX人工智能工程大会上遇到一个人，他向我自我介绍。
I gotta tell you, I met a guy, okay, who he introduced himself at SWIX's AI engineering conference in in in New York.
他说：‘我一直想认识你，史蒂夫。’
And he's like, I've I've wanted to meet you, Steve.
我是你的玩家之一。
I'm one of your players.
明白吗？
K?
我当时心想：哇。
And I'm like, woah.
因为这家伙已经三十多岁了，而且玩过我的游戏。
Because this dude's, you know, in his thirties and, you know, know, he's played my game.
你根本不知道我写的这个游戏是什么吗？
You don't understand the game that I wrote?
这是一款大多数人能感受到的遊戲，但大多数人没看过，因为我没有开源它。
It's something most people Vibe and most people haven't seen it because I didn't open source it.
我总有一天会的。
I will someday.
这真是个麻烦事。
It's just a pain in the butt.
这是一件非常美好的事，它在玩家之间创造了大量的爱。
It's a really beautiful thing, and it and it created so much love in the players.
几十年来，他们都会回来。
For decades, they would come back.
对吧？
Right?
但这个人特别投入，他说他读了你的《富有的程序员美食博客》文章，然后决定成为编译器专家。
But this guy was so into it, and he's like, I read your I read your rich programmer food blog post and decided to become a compiler expert.
我拿到了博士学位。
I became a PhD.
他读这篇文章的时候还在上高中。
He was in high school when he read it.
获得了博士学位，并创办了自己的公司。
Became a PhD, started his own company.
他现在有一家初创公司，发展得非常好。
He's got a startup that's doing really really well now.
他说，这一切都源于那篇文章。
And he said it was all because of that post.
而且这篇文章谈到，我认为你曾主张，除非你了解编译器的工作原理，否则你不可能成为一名优秀的、高效的程序员。
And and this post talks about I think you argued that unless you know how compilers work, you're not gonna be a good programmer, an efficient programmer.
我不确定那个
I'm not sure what what the
短语是：在你所做的事情和计算机实际执行的操作之间，会存在一层魔法般的抽象，这将永远成为你的障碍。
phrase There's gonna be a layer of magic between what you're doing and what the computer is doing that is forever gonna be sort of a friction for you.
然后我认为你甚至指出，一些博士甚至都不理解编译器是如何工作的，这会让他们很难变得高效。
And then I think you even argued that some PhDs don't even understand how compilers work and this will make it really hard for them to be efficient.
当时，这确实是事实。
At the time, that was definitely true.
对吧？
Right?
你觉得那篇文章现在怎么样？
How do you think that post has aged?
因为那时候，我想大概是2012年左右。
Because at at that time, I think it was, like, 2012 or or so.
即使在那时，我也觉得说你需要懂汇编语言有点不寻常，因为当时已经是高级语言的时代了。
Like, even then, I would assume it was a bit unconventional to say, like, you need to understand assembly because it was high level languages.
对吧？
Right?
Java正处于鼎盛时期，C#，Ruby也开始兴起。
Java was was was in its prime, c sharp, Ruby was starting to come out.
我的意思是，连JavaScript都开始崛起，React几年后才会出现。
I mean, heck, JavaScript was starting to become big, React will start in a few years.
大多数开发者都会想，我为什么要懂编译器和汇编？
And most developers would have thought, why would I need to know compilers, assembly?
我是说，这不就是编译器该干的事吗。
I mean, that's what the compiler is for.
对吧？
Right?
对。
Yeah.
你这是在问一个非常非常非常基础的问题——耶格，你其实是在问我大学应该教些什么。
You're asking a really, really, really foundational quest you're asking me what universities should teach is what you're asking me, Yegge.
是吧？
Okay?
（这话）实在让人难以认同。
In despise.
而且你知道吗，从我80年代入行以来，这些要求标准每隔几年就会变一次。
And you know, that that those goal posts have moved every few years since I got into this game in the eighties.
好吧。
Alright.
要想成为一名软件工程师，过去你需要懂汇编语言。
What you need to know in order to be a software engineer, it used to be assembly language.
过去你需要了解很多位操作之类的东西。
It used to be like lots of bits and stuff like that.
随着时间推移，我和我的朋友们意识到，我们最爱的位操作问题，开始难倒那些连位是什么都没见过的候选人。
And over time, my buddies and I realized that our favorite bit manipulation questions were starting to bounce off candidates who'd never seen a bit before.
对吧？
Right?
后来我们认真反思了一下，就在二月份，我们开始想，现在你还真需要懂怎么用异或操作位和字节吗？
And we real you know, we did some soul searching in the February, you know, and we were like, yeah, do you really need to know how to manipulate bits and a bite with XORs and stuff like that anymore?
大概不需要了。
Probably not.
对吧？
Right?
这个发现让人挺沮丧的，因为我们一直为自己懂这些原理而自豪，但如今我们真的不再需要它们了。
And that was a depressing realization because we had prided ourselves in knowing how that stuff works, but we just don't need it anymore.
令人沮丧的现实是，我把自己很多的自尊和身份认同都寄托在了编译器背景上。
And the sad reality is that, you know, I I I had a lot of my own ego and identity wrapped up in my sort of compiler background.
这很有趣。
It's all it's interesting.
对吧？
Right?
但它在任何有意义的层面上都已经不再有用。
But it's it's not useful in any meaningful sense anymore.
是因为编译器在优化方面变得如此出色了吗？
And is is it not useful because the compilers have gotten so good at optimizing, for example?
是因为问题已经转向了更高层次的抽象吗？
Is it that the problems have moved on to higher layers?
是的。
Yeah.
你认为为什么会这样？
Why do think that is?
而且而且
And and
这是在提升抽象层次。
this is walking up the abstraction ladder.
就是这样。
That's all.
我们甚至还没谈到人工智能呢。
We're not even talking about AI just yet.
像这样的事情甚至在
Like, this happened even
你没提到人工智能。
You didn't say AI.
你提到人工智能了吗？
Did you say AI?
没有。
No.
还没有。
Not yet.
我们会提到的。
We will say it.
但就连在2010年代后期，我也记得，它根本没怎么被提及。
But this but even in the I remember, like, you know, late twenty tens, it didn't really come up.
在我的职业生涯中，我只记得有一次，知道编译器做了什么会很有帮助，但即便那样，可能也是个误导，说实话。
Like, in in my career, I can only remember one time where it would have been nice to know what the compiler did, but even then might have been a red herring, honestly.
你看，你必须掌握的东西一直在变化。
Look, what you have to know just keeps moving.
他们一直在调整课程内容。
They just they keep changing the courses.
他们不断改变教学内容。
They keep changing what they teach.
很多人没注意到这一点，因为他们只回看一两年或三年，稍微往前看一点，但我已经从事这个行业四十年了，我可以告诉你，现在教的东西和过去完全不同。
Many people don't see this because they're only looking a year or two or three back and you know, looking a little bit forward, but I've been doing this for forty years and I can tell you they teach you very different things now than they used to teach.
这是因为你需要掌握完全不同的知识。
And it's because you need to know very different things.
这一点在图形行业、计算机图形学的指数级发展过程中表现得最为明显。
And nowhere is it more evident than when we saw the exponential curve of the graphics industry, computer graphics.
看看今天的图形技术，和1992年我上大学学图形时相比，那时我必须亲手学习算法，来计算一条线上下一个像素的位置，以便进行渲染。
Look at graphics today compared to '19, you know, '92 when I was learning graphics in university and I had to learn how to literally, you know, do the algorithm to figure out where the next pixel goes on a line so I can render it.
最终，这被简化成了三角形，也就是多边形。
So eventually turned it into a triangle, which is a polygon.
但两年后，我再上同样的课程时，我们已经开始做动画了。
Meanwhile, two years later, I took the same course and we were doing animation.
是的。
Mhmm.
我当时甚至不知道什么是多边形。
I didn't even know what a polygon was.
我的意思是，我知道，但没达到那个深度。
I mean, I did, but not at that level.
对吧？
Right?
整个阶梯一直在往上移，工作内容也变了。
The whole ladder just kept moving up and the jobs changed.
最初需要能写设备驱动的人，后来需要能做游戏世界、物理引擎等各种东西的人。
Originally, needed people that could write device drivers, then they needed people and now they need people who can do game worlds and physics and all this stuff.
对吧？
Right?
图形学只是给我们指明了方向。
It's the they just graphics showed us the way.
这就是会发生的事情。
This is what happens.
软件工程岗位自iOS和移动云技术以来一直非常稳定。
And software engineering jobs have been very stable for, I don't know, since iOS, since mobile and cloud.
这两个是最后两大创新。
Those are the last two big innovations.
对吧？
Right?
是的。
Yep.
史蒂夫刚刚指出，这个行业经历了从原始像素到游戏引擎、从裸机到云的巨大成熟飞跃。
Steve just made the point that the industry goes through these massive maturity leaps from raw pixels to game engines, from bare metal to cloud.
如果你今天正在开发需要跃升至企业级的软件，那么有一款工具正好能处理这个问题。
And if you're building software today that needs to make that leap to enterprise grade, there's a tool that handles exactly that.
这是我们经验丰富的赞助商WorkOS。
This is our seasoned sponsor WorkOS.
如果你正在开发任何SaaS产品，尤其是AI产品，身份验证、权限、安全性和企业身份可能会悄然变成一项长期投资。
If you're building any SaaS, especially an AI product, authentication, permissions, security and enterprise identity can quietly turn into a long term investment.
SAML的边缘情况、目录同步、审计日志以及企业客户期望的所有功能。
SAML edge cases, directory sync, audit logs and all the things enterprise customers expect.
构建这些关键部分已经很费力，而维护它们还需要更多工作。
It's a lot of work to build these mission critical parts and then some more to maintain them.
但你不必这么做。
But you don't have to.
WorkOS 将这些组件作为基础设施提供，让你的团队能够专注于真正让产品与众不同的部分。
WorkOS provides these building blocks as infrastructure so your team can stay focused on what actually makes your product unique.
这就是为什么 Entropic、OpenAI 和 Cursor 等公司已经使用 WorkOS 了。
That's why companies like Entropic, OpenAI, and Cursor already run on WorkOS.
优秀的工程师知道什么不该构建。
Great engineers know what not to build.
如果身份认证正是你的其中之一，欢迎访问 workos.com。
If identity is one of those things for you, visit workos.com.
好了，让我们回到这个问题：软件工程上一次真正的创新到底是什么。
With that, let's get back to the question of what the last real innovation of software engineering actually was.
实际上，从那以后就没什么进展了。
And it's been kinda dead since then, actually.
是的。
Yeah.
我不想说AI，因为我们还没谈到这个，但我认为我们经历了一段停滞期，课程几乎没有变化，我们以为这就是我们永远需要知道的一切。
I don't wanna say AI because we're not talking about it yet, but but I think we went through a I think we went through a period where people stagnated a little bit, where the courses didn't change very much, and we thought this is all we're ever gonna need to know.
我觉得上一次大的创新是分布式系统，如果我错了请纠正我。
I I feel the last big innovation, correct me if I'm wrong, was distributed systems.
那是最后一个真正的难题，大概从二月优步把微服务引入以来，如何扩展服务、如何存储大量数据。
That that was the last kind of hard problem starting from like February when Uber brought brought microservices into there, how you scale services, how you store large amounts of data.
我觉得那是一个，就像
I feel that was a, like
我的意思是，那确实是个巨大而缓慢的转变。
I mean, it was a big it was a big slow.
是的。
Yeah.
但说实话，我觉得现在有很多迁移工作正在进行，新的React版本不断推出，开发者们为此感到困扰。
But, honestly, like, feel there's a lot of migrations happening, new React versions coming up and developers struggling with that.
苹果每年都会扔进一个螺丝刀，用新的破坏性版本卡住我们的进度。
Apple every year throwing in a, you know, like a a screwdriver in in in in the wheels with the new breaking version.
安卓开发者需要淘汰旧版本的安卓系统，并决定从哪里 cutoff。
Android developers needing to retire an Android old version and deciding, like, where to cut it off.
所以我觉得，一方面有这种迁移问题，另一方面商业环境也很好。
So I feel there was that, like, kind of, like, migrations thing and also business was just good.
对吧？
Right?
每个人都在成长。
Like everyone was growing.
我们当时都在疯狂招聘。
We were like, everyone was busy hiring.
好像明天就不存在了一样。
Like there's no tomorrow.
2021年的时候，市场非常火热。
There was a time in 2021, the market was so hot.
很多只有一三个月经验的培训班学员都拿到了offer。
A lot of boot campers with three months experience were getting offers.
这是一家不错的公司，因为当时每个人都急着招聘。
It's a pretty good company because everyone was so desperate to hire.
是的。
Yeah.
然后到了2022年，AI出现了。
And then came AI in in 2022.
我一直觉得你很务实，即使在2020年代甚至更早的时候也是如此。
One thing that always struck me about you, even in those, like, you know, in twenty twenties and even before, you're always pretty pragmatic.
你知道，你本来就是做编译器和调试工具的。
You know, you were by by trade, you were always into compilers, debugger tools.
你就是从这里开始的。
That's where you started.
你在亚马逊、谷歌都处理过棘手的问题。
You worked on hard problems at Amazon, at Google.
你从不回避那些复杂的技术难题，所有这些事情你都迎难而上。
Never shied away to getting into, like, hard technical problems and, you know, like, all all these things.
当AI出现时，我不记得你说过‘这太棒了’。
And when AI came out, I don't remember you saying, oh, this is amazing.
这将改变世界。
This is gonna change the world.
你当时是什么感受？
How did you feel?
你是不是在观察，持怀疑态度？
Were you kind of, like, observing, skeptical?
比如最开始，当你第一次接触到大语言模型时？
Like, at the very beginning, right, when you first came across LLMs?
那感觉怎么样？
How how was that?
我真的很惊讶，它们居然能写出相当连贯的Emacs Lisp函数。
I was pretty blown away that they could write fairly coherent Emacs Lisp functions.
嗯。
Like Mhmm.
比如ChatGPT，就是2022年12月那个原始版本。
Like ChatGPT, the original one in in December 2023.
2022年。
2022.
2022年？
2022?
好吧。
Okay.
时间过得真快啊。
Boy, time flies.
那时候已经能用一种奇怪的语言写代码了。
Could already write code in a weird language.
对吧？
Right?
写的代码也不多。
Not very much of it.
而且它很粗糙，但对我来说，那是我意识到‘原来如此’的起点。
And it was it was janky, but that was for me that was the beginning of oh, right?
你知道吗，因为我认识的AI领域的朋友已经说了二十年了，随时，马上就要来了。
You know, because I've had friends in AI for twenty years saying any minute now, any day now.
对吧？
Right?
他们总是给我们展示越来越完善、越来越好的东西。
And they'd show us something complete better and better and better.
而这次，我第一次真正明白了，哦，原来是这样。
And this was the first time it was oh, okay.
我现在懂了。
I I see now.
对吧？
Right?
但我还是像其他人一样持怀疑态度。
But I was still skeptical like everybody else.
我可以告诉你，因为去年年初关于Cloud Code的传闻出来时，对吧？据说Anthropic内部有个工具在帮他们写代码，还是个命令行工具？
And I can I can tell you because when when the rumors came out about Cloud Code in beginning of last year, right, that Anthropic had a tool internally that was writing code for them and it was a command line tool?
我和所有人一样，根本不信，觉得这不可能，就是直接拒绝。
I I along with everyone else went, no, it's not, you know, it's we were just like just flat out rejection.
绝对不可能发生。
Just absolutely not happening.
对吧？
Right?
直到我亲自用了之后，才恍然大悟。
Until I used it, and then I was like, oh, I get it.
我们都完蛋了。
We're all doomed.
对吧？
Right?
就在那之后，我其实写了篇关于初级开发者灭亡的文章。
And then I wrote death of the junior developer right after that, actually, I think.
天啊，我写《初级开发者之死》可能甚至是在GPT-4发布之后。
Gosh, it might have even been after after four o came out that I did death of the junior developer.
但一旦GPT-4发布，事情就迅速发生了变化。
But things changed really fast once that came out.
所以，我是个怀疑者吗？
So was I a skeptic?
是的。
Yes.
但我有关注趋势的变化吗？
But did I pay attention to the curves?
从一开始，我就想，如果ChatGPT-3.5能写出一个连贯的Emacs Lisp函数，那一年后会怎么样呢？
From the very beginning, I figured if chat g p t three five can write a coherent Emacs Lisp function, then in a year, let's see how they do.
一年后，GPT-4已经能写出上千行代码了。
And in a year, four o was writing a thousand lines of code.
上千行。
A thousand lines.
老兄，世界上大部分代码都存在于一千行或更少的文件中，这意味着它能做出可信的修改。
Dude, that's most of the world's code is in files of a thousand lines or less, which means that it can make credible edits.
在GPT-4发布之前，它还做不到这一点。
It wasn't able to up until four o came out.
对吧？
Right?
所以那时候我就想，好吧，我们正处在一条上升曲线上。
And so like, man, it was that point when I was like, okay, we're on a curve.
这是一段旅程。
This is a ride.
它不会停下。
It's not stopping.
让我们登上这趟列车，看看它会带我们去哪。
Let's get on the ride and see where it goes.
于是我全身心投入了。
And I dove in.
对吧？
Right?
我当时就想，我落伍了。
And I was like, I was behind.
我不懂人工智能。
I didn't know AI.
我不懂它的基本原理。
I didn't know, like, the the fundamentals of it.
我不懂那些术语。
I didn't know the lingo.
现在人人都懂这些东西了。
You know, everybody knows this stuff now.
对吧？
Right?
是的。
Yeah.
但我花了一年时间，只专注于阅读论文和追赶进度。
But I spent a year doing nothing but reading papers and catching up.
对吧？
Right?
所以在这本《Vibe Coding》里，我记得上次你上播客时，这本书即将出版，我当时正在读它的早期版本。
So in this book, Vibe Coding, I remember last time you were on the podcast, this book was about to come out and I was reading an early version of it or so.
但我刚看了封面背面，意识到你一定是一年前写的，里面说手工编码的时代已经结束了。
But the back cover, I just read the back cover and I realized that you must have written this about a year ago and it says the days of coding by hand are over.
你是什么时候意识到这一点的？
When did you realize this?
因为我最近才在Opus 4.5上意识到这一点，但你这想法可比那早多了。
Because I've realized this recently with Opus 4.5, but this was this was a while before well before that.
嗯。
Mhmm.
是的。
Yeah.
是一年前的事了。
It was a year ago.
让我想想。
It was let's see.
现在是什么时候了？
What is it right now?
一月吗？
January?
所以是一年多以前了。
So it was over a year ago.
我第一次意识到的时候，是十二、十三个月前。
It was twelve, thirteen months ago when I first realized.
而且那根本不是我的原话。
And and it wasn't that wasn't even my quote.
那是埃里克·迈耶博士说的。
That was that was doctor Eric Meyer.
对吧？
Right?
他是编程界众多发明的创造者。
The inventor of many, many, many things in in the programming world.
世界上最重要的编译器专家之一。
One of the most important compiler people in the world.
想想这个人。
That dude, think about it.
他一生都在为开发者构建能够编写代码的技术。
He spent his life building technology for developers to be able to write code.
而他却说，开发者以后不会再写代码了。
And he's saying developers aren't gonna write code anymore.
是什么让人会说，我毕生的工作其实并不对？
What would possess somebody to say, well, my life's work isn't really right?
这正是让吉恩·金和我都说‘对吧？’的原因。
And that's what caused actually Gene Kim and I both to go, Right?
你知道吗，这位发明家，你知道的，他对Visual Basic、C#、链接工具、Haskell，还有那个带猪的PHP都做出了巨大贡献。
You know, if the inventor of, you know, you you know, he he made huge contributions to to to Visual Basic and c sharp and and and link and and Haskell and p and and PHP with a pig.
是这么叫的吗？
Is that what it's called?
对吧？
Right?
全都是他做的。
All him.
但他却说：不。
And he's just like, no.
结束了。
We're done.
我们不再写代码了。
We're done writing code.
我的意思是，一个语言专家说出这种话，可真够重磅的。
I mean, that's that's that's that's pretty big words from a languages person.
世界上最有名的之一。
One of the most famous in the world.
对吧？
Right?
他看到了我们没看到的东西？
What does he see that we didn't?
他看到了那些曲线，伙计。
And he sees the curves, man.
就这么简单。
It's that simple.
就像指数曲线，它们会迅速变得极其陡峭，而今年我们正步入这个陡峭阶段。
It's like exponential curves, they get real steep real fast, and we're we're heading into the steep part this year.
所以，C# 和 Visual Basic 的发明者说，我们不用再写代码了。
So the inventor of c sharp and Visual Basic is saying that we're done writing code.
但即使 AI 写了所有代码，也还是得有人来验证。
But even if the AI writes all the code, someone has to verify it.
这就是我们经验丰富的赞助商 Sonar 发挥作用的地方。
And that's where our seasoned sponsor, Sonar, comes in.
Sonar 是 SonarQube 的开发者，它推出了以代理为中心的开发周期框架，这是一种专为 AI 生成代码的独特规模和速度设计的新型软件开发方法论。
Sonar, the makers of SonarQube, has introduced the agent centric development cycle framework a new software development methodology designed for the unique scale and speed of AI generated code.
这是一种转向更明确的四阶段循环的举措，为代理提供真正需要的约束机制。
It's a move towards a more intentional four stage loop that gives agents the guardrails they actually need.
四个阶段分别是：先指导，代理需要理解它们被要求创作的背景，以确保输出符合开发者和组织的需求。
The four phases being: Guide First, agents need to understand the canvas on which they are being asked to create so that the output fits with what the developer and organization require.
生成，基于大语言模型的工具会生成它认为能在正确上下文中实现预期结果的代码。
Generate The LLM based tool generates the code it believes will achieve the desired outcome within the right context.
验证。
Verify.
接下来，代理被明确要求检查其工作，确保它真正实现了预期目标，并且可靠、可维护且安全。
Next, the agent is deliberately required to check its work, ensuring it actually achieves the desired outcomes and is reliable, maintainable, and secure.
解决。
Solve.
最后，任何发现的问题都会交给代码修复代理进行修复。
Finally, any issues identified are provided to a code repair agent to fix.
为了支持这一框架，Sonar 显著增强了这一产品线，推出了 Sonar 上下文增强、SonarQube 代理分析、SonarQube 架构分析和 SonarQube 修复代理等功能。
To power this, Sonar has significantly strengthened this offering, introducing products and capabilities like Sonar context augmentation, SonarQube agentic analysis, SonarQubeArchitecture, and SonarQubeRemediationAgent.
前往 sonarsource.com/pragmatic 了解 Sonar 的最新动态，以及它如何助力组织拥抱代理时代。
Head to sonarsource.com/pragmatic to learn more about the latest with Sonar and how it's empowering organizations to embrace the agentic era.
好了，让我们回到 Steve 的指数曲线。作为反方观点，你知道，作为一名工程师，你可以画出各种曲线，但你永远不知道它们何时会结束，或者是否会趋于平缓，诸如此类。
With this, let's get back to Steve's exponential curves of Playing devil's advocate, you know, like one thing about being an engineer is like you you can draw up curves, but, you know, like you never know when they end or if they flatten, whatnot.
我们可以看到它已经走到了哪里。
We can see where it has come.
是什么让你相信这条曲线会持续上升？
What made you believe that this curve would keep going?
尤其是像 LLM 这样的技术，它居然能勉强工作，这在很大程度上出乎了许多人的意料。
Especially that with LLMs, the fact that it even kinda works was a bit of a, I guess, surprise for a lot of people.
而它持续扩展的事实也令人惊讶，现在的问题是，这种扩展还能持续多久。
And the fact that it kept scaling is a surprise, and there's this question of, like, how long they will scale.
是的
Yeah.
所以世界上充满了怀疑者。
So the world is filled with unbelievers.
明白？
K?
那些特别认为曲线是这样的样子的人。
People who specifically who believe the curve looks like this.
像一条S形曲线，先上升，然后变平。
An s, it goes up, and then it flattens.
明白？
K?
他们真的认为我们现在正处于峰值阶段。
And they actually think we're at the hump right now.
是的
Yeah.
而且这会成功的。
And that'll fly.
自从GPD 35推出以来，他们就一直反对这一点，他们说：是的，情况不会再好转了。
And they have fought that ever since the g p d three five came They're like, yeah, it's not gonna get any better.
四零发布了。
Four o comes out.
我们喜欢四零。
We love four o.
人们喜欢四零。
People love four o.
他们现在仍然喜欢。
They still do.
他们根本摆脱不了它。
They can't get rid of it.
是的。
Yeah.
但他们仍然认为这就是最好的了，你知道的。
But they still think that's as good as it gets, you know.
Office 4.5已经发布了，但大多数人还没试过。
Office 4.5 is out, and most people haven't played with it.
大多数人并不了解其中的功能，而这个版本已经推出两个月了。
Most people don't realize what's there, and that thing is already two months old.
据我所知，模型更新的半衰期从去年年初的约四个月，缩短到了今年年初Anthropic时期的两个月。
The half life between model drops, far as I can tell, has gone from about four months beginning of last year to two months from Anthropic at the beginning of this year.
所以随时都可能看到Anthropic发布另一个新模型。
So any day, we're gonna see another model from Anthropic.
等我们这期播客发布的时候，它可能已经上线了。
It'll probably be out by the time we have this podcast out.
对吧？
Right?
而那个新模型将远远领先于当前水平，人们将会因此感到非常震惊。
And that will be so much further up the curve that people are gonna start to be really freaked out by it.
展开剩余字幕（还有 480 条）
当人们看到下一个模型时，他们会感到担忧。
It's gonna it's gonna worry people when they see the next model.
明白吗？
Okay?
因为现在人们抱怨的所有错误和问题都会被反馈回训练数据中，这样下次就不会再犯了。
Because all of the bugs, all the mistakes that they're complaining about right now get fed right back in his training and so that it doesn't make them the next time.
这就是人们没有理解的地方。
And this is what people aren't understanding.
对吧？
Right?
而且，时间还在继续。
And also, time continues.
三年、五年之后依然会存在。
There will be three and five years from now.
太阳不会停止运转。
The sun's not gonna stop.
对吧？
Right?
而且它正在到来。
And it's coming.
所以这些曲线不可避免的碰撞，天啊，将会引发社会动荡。
So this inevitable the collision of these curves, man, it's there will be societal upheaval is what's gonna happen.
而且它已经开始了。
And it's already started.
人们有理由感到愤怒。
And people are justifiably mad.
我和他们一样愤怒，格格。
And I'm mad with them, Gergge.
明白吗？
K?
我对亚马逊裁员16000人却把责任推给没有AI战略的AI感到愤怒。
I'm mad at Amazon for laying off 16,000 people and blaming an AI without an AI strategy for it.
这些人将无法找到新工作，我的朋友，他们只是即将出现的众多之一。
Those people are not gonna be able to find jobs, my enlarge, and they're the first of many to come.
而且没人对此有任何计划。
And nobody has a plan for this.
如果他们没有AI战略，你认为亚马逊为什么要这么做？
Why do you think Amazon did that if they don't have an AI strategy?
因为，不幸的是，人们可能会因为我这么说而讨厌我，但我说了并不意味着它就是真的。
Because, unfortunately, and people are gonna hate me for saying this, but me saying it doesn't make it true.
它本来就是真的。
It was true already.
每个人面前都有一个旋钮，可以从0调到100。
Everybody has a dial that they get to turn from zero to a 100.
你可以把手从旋钮上拿开，但它默认设置就是：为了支付其他人使用AI的费用，你需要裁掉多少百分比的工程师。
And you can keep your hand off the dial, but it just has a default setting of what percentage of your engineers you need to get rid of in order to pay for the rest of them to have AI.
因为他们都在开始用自己的工资购买AI代币。
Because they're all starting to spend their own salaries in tokens.
因此，至少在一段时间内，如果你想让工程师们发挥最大生产力，你就不得不裁掉一半人，以让另一半人实现最大效率。
And so at least for a while, if you want your engineers to be as productive as possible, you're gonna have to get rid of half of them to make the other half maximally productive.
而事实上，你的一半工程师根本不想做提示工程，他们已经准备辞职了。
And as it happens, half your engineers don't want to prompt anyway, and they're ready to quit.
所以现在的情况是，平均来看，每个人都在把这个旋钮调到大约50%，我们将失去大公司约一半的工程师，这令人担忧。
And so what's happening is everybody on average is setting that dial to about 50%, and we're gonna lose about half the engineers from big companies, which is scary.
是的。
Yeah.
这太疯狂了。
That's wild.
这规模远比当年新冠疫情时要大得多。
It's it's way that's way way bigger than we've seen back at COVID.
这规模会更大。
It's gonna be way bigger.
这将会很糟糕。
It's gonna be awful.
但与此同时，另一件事正在发生，那就是人工智能正在让非程序员编写代码。
It's but but at the same time, something else is happening, which is AI is enabling non programmers to write code.
它还让那些已经认清趋势、相信曲线将持续上升的工程师们聚集起来，组成两人、五人、十人、二十人甚至三十人的团队，开始做出足以与那些自乱阵脚的大型公司相媲美的成果。
And it's also enabling engineers who have seen the light and believe the curves are gonna continue to go up to actually get together in groups of two and five and ten and twenty and thirty people and start to do things that rival the output of these big companies that are tripping over themselves.
因此，一场疯狂的创新浪潮正在兴起。
And so we've got this mad rush of innovation coming up
嗯。
Mhmm.
自下而上。
Bottom up.
同时，随着大公司大规模裁员，大量知识工作者正从天而降。
And we've got this mad knowledge workers falling out of the sky as the big companies lay them off.
因为显然，大公司已经不再是合适的规模了。
Because there's clearly the big company is not the right size anymore.
甚至连安迪·贾西都在这么说。
It's not even Andy Jassy saying it.
我们将用更少的人做同样的事情。
We're gonna do the same thing with fewer people.
对吧？
Right?
这意味着我们会有一百万倍更多的公司吗？
And so does this mean we're gonna have a million times more companies?
软件会迎来爆炸式增长，还是人们会彻底离开软件行业，转而去干别的？
Is there gonna be a massive explosion of software or people gonna get out of software altogether and we're all gonna go do other stuff?
我的意思是，我非常好奇这一切最终会走向何方。
I mean, like, I I'm very curious where all this goes.
是的。
Yeah.
拥有正确技能组合、能发现正确商业机会或具备优势的小团队，能做得多得多。
Small teams that have the right skill set or or see the right business opportunity or have advantages can do way more.
所以这其中确实有道理。
So there is something there in that.
确实有。
There is.
所以现在正掀起一场土地争夺潮。
So there's this land rush starting.
我认为，许多来自知识型工作的人只是反对人工智能，而这些人将会很艰难。
I think a lot of the people coming out of knowledge work are just anti AI, And those people are gonna struggle.
对不起。
I'm sorry.
但如果你现在还反对人工智能，那就像是反对太阳一样。
But if you're anti AI at this point, it's like being anti the sun.
你只能去地下生活了。
You're gonna have to go live underground.
对吧？
Right?
但那些支持人工智能的人，我觉得我们会看到工作分配和软件来源的巨大重新分布。
But the people who are like pro AI, like, I I think we're gonna see a big redistribution of who's doing the work and and where you get your software from.
而且我们很可能最终会进入一个美好的状态，那时亚马逊甚至可能不复存在。
And it may we may well wind up from I I I could actually see a happy place where Amazon's not even a thing anymore.
嗯。
Mhmm.
我真的这么认为，因为软件正在发生一些我们还无法命名的变化。
I I really could because software becomes we don't have the words for what's happening.
对吧？
Right?
今年发生了太多我们无法用语言描述的事情。
We're know, so many things happening this year that we don't have words for.
你有没有注意到这一点？
Have you noticed that?
但软件正变得越来越分散。
But software becomes sort of like distributed.
我不知道。
I don't know.
我确实看到非技术人员开始进入软件领域。
I do see non technical people getting into software.
对于工程师来说，会不会出现一个机会，让他们更多地参与维护工作？
Could there be a job there for engineers to come and actually take more maintenance?
是的。
Yeah.
我的意思是，我认为将会有很多机会，会有大量工程师从事软件工程工作。
I mean, I I think there's gonna be plenty of opportunity for there's gonna be there are gonna be a lot of engineers doing software engineering.
我只是觉得我们都会借助AI来做这件事。
I just think we're all gonna be doing it with AI.
对吧？
Right?
是的。
Yeah.
但我认为，在公司完全放心让AI独立编写和部署代码、无需任何人类介入之前，还需要相当长的时间。
But I think it'll be quite some time before companies are comfortable trusting their code to be to full written and deployed by AI without any human being involved at all.
我认为人们忽略了一个关键点，那些反对者和怀疑者忽略的要点是，AI并不是要取代你的工作。
I think the the point that people are missing, the important point that the naysayers and the skeptics are missing is not that it's AI is not coming to replace your job.
它并不是一个替代功能。
It's not a replacement function.
而是一个增强功能。
It's an augmentation function.
它的存在是为了让你更好地完成工作。
It's here to make you better at your job.
对吧？
Right?
这其实并不是坏事。
And that's not a bad thing, actually.
我不明白人们为什么要抗拒这一点。
I don't I don't know why people would fight that.
但是
But
说到开发者这份工作，你说了一些可能让很多人感到不安的话。
Speaking about the job as as developers, you've said something that can be triggering for a lot of people.
你说过，我想是在AI工程师峰会上，如果你现在还在用IDE，那你就是一个糟糕的工程师。
You've said that, I think this was on the AI engineer summit, that if you're still using an IDE now, you're you're a bad engineer.
是的。
Yeah.
你得有点自己的判断。
Well, you gotta be a little prerogative.
对。
Yeah.
你知道，我，嗯，让我这么说吧。
You know, I I I well, let me put it this way.
明白吗？
Okay?
我不会说你是糟糕的工程师，因为我认识一些非常优秀的工程师，他们比我强多了，但在我的等级图里还处于一级或二级。
I'm not gonna say you're a bad engineer because I know some very, very good engineers, better than I am, who are still at, like, level one or two in my chart.
对吧？
Right?
但我为他们感到深深的同情。
But I feel profoundly sorry for them.
我从未像现在这样对这些人感到怜悯。
I feel pity for them like I've never felt in my life.
对于这些成年人，他们曾经是优秀的工程师，现在却说，嗯，用 Cursor 吧，我有时会问它问题，它的回答真的让我印象深刻。
For these grown people who are good engineers or used to be, and they they're like, yeah, you know, use cursor and I I ask it questions sometimes and I'm really impressed with the answers.
但当我仔细审查它的代码，提交后，我会想：天哪，你肯定会丢掉工作。
And then I review its code really carefully and then I check it in and I'm like, dude, you're gonna get fired.
而你是我认识的最优秀的工程师之一。
And you're one of the best engineers I know.
跟我讲讲你的图表吧。
Tell me about your chart.
跟我讲讲你设计的那些级别吧。
Tell me about your levels that you came up with.
是的。
Yeah.
所以我在澳大利亚的黑板上画了这个，给一大群人看，想向他们展示不同阶段发生的情况，因为我看到他们处于各个阶段。
So I was drawing this on the board in Australia for a big group of people trying to show them what happens because I saw them at all different phases.
有些人打开了他们的IDE。
Some of them had their IDs open.
有些人使用的是宽大的编码代理。
Some of them have a big wide coding agent.
有些人使用的编码代理非常狭窄。
Some of them, the coding agent was really narrow.
对吧？
Right?
你知道的？
You know?
于是我心想，好吧，我们把你们都放在一个连续谱上，来展示正在发生什么。
And so I was like, okay, we're gonna put you all on a spectrum just to show what's going on.
对吧？
Right?
第一级，没有AI。
And level one, no AI.
对吧？
Right?
你知道的。
You know?
第二级，就是是或否。
And and and level two, it's it's the the yes or no.
我能做这件事吗？
Can I do this thing?
在你的IDE里，你知道的。
You know, in your in your IDE.
对吧？
Right?
是的
Yep.
然后是第三级，你就会想：人生苦短，及时行乐。
And then level three, you're like, YOLO.
放手去做吧。
Just do your thing.
对吧？
Right?
你的信任度在提升。
Your trust is going up.
对吧？
Right?
第四级，你会觉得，代码开始源源不断地涌出来了。
Level four, you're like, the code you're starting to squeeze the code out.
对吧？
Right?
因为你更关注代理在做什么，而不是再过多地看代码差异了。
Because you're like, you wanna look at what the agent is doing and not so much at the diffs anymore.
对吧？
Right?
所以你现在不怎么审查了？
So you're not reviewing as much now?
你确实不怎么审查了。
You're not reviewing as much.
你让更多的内容通过了，而且你真正专注于和代理的对话。
You're you're you're you're letting more of it through and you're really focused on the conversation with the agent.
嗯。
Mhmm.
到了第五阶段，你就只想依赖代理了。
And then at level five, you're like, okay, I I just want the agent.
我会在IDE里稍后查看代码，但我不再用IDE写代码了。
And and I'll look at the code in my IDE later, but I'm not coding with my IDE.
到了第六级，你会感到无聊，因为你会想，好吧，我的代理正在忙。
At level six, you're bored because you're like, okay, my agent's busy.
我得做点别的事情。
I got I gotta do something.
我闲得没事干。
I'm twiddling my thumbs.
于是你又启动了一个代理，现在你上瘾了。
And so you fire up another agent and now you're addicted.
因为你会很快进入一种平衡状态：每个代理都在等着你，总有一个代理在等你，因为总有别人完成了任务。
Because you'll very quickly get into an equilibrium where every agent's waiting there's always an agent waiting for you because somebody's finished.
对吧？
Right?
只要你数学上启动了足够多的代理。
As soon as you spin up enough of them mathematically.
对吧？
Right?
于是你发现自己在它们之间不断切换，像这样来回折腾，根本停不下来。
And so you find yourself just multiplexing between them, going like this, and you can't leave.
有个实际的问题。
Practical question.
假设我正在同一个代码库上工作，你该怎么分配多个代理，避免它们互相冲突呢？
Assuming I'm working on the same code base, do how do you split up the multiple agents so that they don't get in conflict?
你是打算用，比如说
Is that your are you gonna use, like
是的。
Yeah.
这就把你带到了第七层，天啊。
So that takes you to level seven, which is, oh my god.
我搞砸了。
I've made a mess.
对吧？
Right?
我误发消息给了错误的代理，自己都没发现，结果他们在我这个项目里又做了一个大项目，因为我让他们做的，现在我得收拾这个烂摊子，等等。
I accidentally texted the wrong agent and didn't realize it, and they did a big project inside of this project because I asked them to, and now I gotta clean up this mess, etcetera.
对吧？
Right?
所有这些事情。
All that stuff.
那时我开始想，如果我们能协调这些该多好？
And that was when I started going, okay, what if we were to like coordinate this?
如果Claude代码能运行Claude代码呢？
What if Claude code could run Claude code?
这是每个人都想知道的问题。
That's the question everybody wants to know.
去年所有人都在尝试这件事。
And everyone was trying all last year.
它正在运行Claude代码。
It's going Claude code.
运行你自己。
Run yourself.
它会运行一段时间然后停止。
It would run for a while and it would stop.
对吧？
Right?
是的。
Yep.
所以，正是这个停止的问题，我在这方面下了很大的功夫，最终开发了一些工具来帮助解决它。
And and so it was the whole stopping thing that so, yeah, I I pushed on that really, really, really hard and and wound up building some some stuff to help with it.
但没错，天啊，变化真的太大了。
But, yeah, boy, it's changed a lot, man.
它已经改变太多了。
It's it's changed so much.
回到IDE的话题，你和Zed的Nathan Sobo进行了一场非常精彩的现场辩论，主题是‘IDE的死亡’。
Going back to the IDE, you you had a really good live debate with Nathan Sobo from Zed, and the title was the death of the IDE.
你们俩都阐述了自己的观点。
And both of you argued your view.
你对IDE的看法是什么？
What what is your view about the IDE?
而且，你从Nathan那里学到了什么？他的观点更支持IDE，而你则觉得IDE可能不会永远存在。
And and also, what did you learn from from Nathan on on, like, his take of he was a bit more pro IDE, and you were a bit more like, maybe this is not gonna be around forever.
是的。
Yeah.
我的意思是，以我目前的阶段来看，我认为人工智能最终会为我们完成一切。
I mean, you know, I am where I am in my journey, which is I I think that AI will do it all for us eventually.
所以，我是如何看待IDE的？它们到底做什么？真正的用途是什么？
And so the way I see IDEs is what do they really do and what are they really for?
好吧。
Okay.
它们其实并不是为了写代码而存在的。
It's not really for writing code.
它是用来整合各种工具的
It's for bringing tools together
嗯
Mhmm.
并且把它们变成一个强大的工具
And for making a big tool.
对吧
Right?
没错
Yep.
现在你有MCP之类的工具了
And now you have MCP for that or whatever.
对吧
Right?
没错
Yep.
所以我看到IDE正在回归，我认为Claude Cowork是对IDE形态的一种回归。
And so I see IDEs returning, and I think Claude Cowork is a return to the IDE form.
它是Claude Code在想：哦，我需要为真实用户服务。
It's it's Claude Code going, oh, I need to be for real people.
对吧？
Right?
但我认为Cloud Code的工作形态可能比Cloud Code更适合普通开发者。
But I think Cloud Code works form factor probably works better for the average developer than Cloud Code does.
对吧？
Right?
所以我看到IDE，看到我们正在回归一个IDE主导的世界，只不过这一切都通过对话和监控来实现。
So I see IDE I see us coming back into a world where it's IDEs except it's all conversations and, you know, monitoring.
这是一个非常好的观点。
And this is a really good point.
我哥哥开发了一个叫Craft Agents的东西，它和Cloud Cohort非常相似，只是他们连接了自己公司的数据源。
My brother built a thing called Craft Agents, which is pretty similar to to Cloud Cohort except they connected in in their company their own data sources.
他说一些开发者开始更喜欢这种方式，因为它是可视化的，更容易看到并行的代理，例如，如果你不是高级用户，滚动起来更容易。
And he said that some developers start to prefer that because it's a visual that's easier to see parallel agents, for example, if you're not a power user, it's easier to scroll.
这只是更友好的用户界面。
It's just a nicer UI.
所以你的观点是，也许有些开发者应该试试看，如果你对Cloud Code还不信服，就试试Cloud Code或其他类似的、更视觉化的东西。
So your point on maybe some developers should try out, like if if you're not sold on Clock Code, like try Cloud Code or or any other similar, more visual thing.
它可能更适合你。
It it might be more your thing.
但你知道，像git这样的工具，有些人就喜欢命令行。
But, like, you know, git, some people love the command line.
我其实只是用图形界面，因为我就是不喜欢记命令，虽然承认这一点有点尴尬，但也许现在这已经没那么尴尬了。
I actually just use the UI because I just don't like memorizing the commands, as embarrassing it is to admit, or maybe these days, it's not as embarrassing.
是的。
Yeah.
关键是去尝试。
The key was try.
只要你愿意尝试就行。
As long as you're trying something.
是的。
Yeah.
如今，公司里最重要的单一代理指标可能是代币销毁。
What probably the single most important proxy metric that you can have in a company today is token burn.
因为代币销毁表明你的工程师或非工程师们正在尝试做事情。
Because what token burn says is your engineers are trying to do stuff or your non engineers.
当他们尝试时，他们会失败，也会从中学习。
And when they're trying, they're failing and they're learning.
如果你想尽早发现这些组织瓶颈，想让你的工程师尽早在我提出的八级体系中提升，想提前解决你的业务流程，你就得现在就开始行动，这意味着要去尝试。
And so if you wanna get those organizational bottlenecks discovered early on and you wanna get your engineers leveled up on my eight level spectrum early on and you wanna solve your business processes ahead, you need to start now, which means try.
你尝试什么并不重要。
It doesn't matter what you try.
你用哪个工具也不重要。
It doesn't matter which tool you use.
只要你使用AI并尝试让它完成工作，你就是在做正确的事。
As long as you're using AI and you're trying to get it to do the work, you're doing the right thing.
对。
Yeah.
作为专业人士，我们至少应该尝试一下。
And I I think as professionals, like, we really ought to just at least try.
这样你才能获得第一手经验，然后做出自己的决定。
Like, you get firsthand experience and then you can make your decision.
史蒂夫关于代币销毁的观点非常有趣。
Steve's point about token burn is really interesting.
获胜的公司都是那些最善于实验的公司。
The companies that win are the ones that experiment the most.
如果你想将这种实验心态应用到你的产品上，而不仅仅是AI的使用，那正是我们的赞助商Statsig所打造的。
And if you want to bring that same experimental mindset to your product, not just your AI usage, that's exactly what our presenting sponsor, Statsig, is built for.
Statsig为你提供了完整的工具包，而无需你自己从头构建。
Statsig gives you the complete toolkit without building it yourself.
你可以在一个平台上获得功能开关、实验分析和产品数据分析，并且所有数据都基于相同的用户分配和底层数据。
You get feature flags, experimentation, and product analytics all in one platform and tied to the same underlying user assignments and data.
实际操作中是这样的：你首先将更改推送给1%的用户。
In practice it looks like this: you roll out a change to 1% of users at first.
观察它如何影响你关心的核心指标：转化率、留存率，或与此次发布相关的任何关键指标。
You see how it moves the top line metrics you care about: conversion, retention, whatever is relevant for that release.
如果出现问题，可以立即回滚。
If something goes wrong, instant rollback.
如果效果良好，你就可以自信地扩大规模。
If it's working, you can confidently scale it up.
像Notion这样的公司，从每个季度仅进行个位数的实验，增长到使用Statsig进行超过300个实验。
Companies like Notion went from single digit experiments per quarter to over 300 experiments with Statsig.
它们通过功能开关发布了600多个功能，在快速迭代的同时防止了指标倒退。
They shipped over 600 features behind feature flags, moving fast while protecting against metric regression.
微软、Atlassian和Brex使用Statsig的原因相同：它是实现规模化下速度与可靠性并重的基础设施。
Microsoft, Atlassian and Brex use Statsig for the same reason: It's the infrastructure that enables both speed and reliability at scale.
Statsig 提供慷慨的免费套餐供新手使用，团队的专业定价从每月 150 美元起。
Statsig has a generous free tier to get started and pro pricing for teams starts at $150 per month.
如需了解更多信息并获取三十天企业版试用，请访问 statstig.com/pragmatic。
To learn more and get a thirty day enterprise trial, go to statstig.com/pragmatic.
好了，让我们回到史蒂夫对 Gas Town 现状的看法。
With that, let's get back to Steve's take on the state of Gas Town.
现在很多人不知道该如何尝试，他们说，好吧，我来试试，结果却做错了，因为他们总是这么做。
Now there's a huge problem with people not knowing how to try, and they say, oh, let me do something, and then it does the wrong thing because they always do.
然后他们就惊呼：哇哦。
And then they're like, woah.
这太糟糕了。
This is garbage.
所以，你得教他们，这就像一把铲子，你不能像在《幻想曲》里那样挥舞铲子乱挖。
So, you know, you have to teach them that it's a shovel and you don't go shovel dig like in Fantasia.
对吧？
Right?
就像让扫帚自己走来走去。
Like make the brooms walk around.
不。
No.
你拿起铲子，用它来挖，但这把铲子是你之前用手挖的时候没有的。
You pick up the shovel and you dig with it, but it's a shovel that you didn't have before you were using your hands.
这其实是一个非常非常简单的比喻，但人们就是不明白。
Like, it's a really, really simple analogy, but people just don't get it.
他们就是不明白。
They don't get it.
我想说点可能有争议的话，但这确实是现实。
And I think, I'm gonna say something that's contentious, but in in like, it's it's just the reality of the world.
大多数人不会阅读。
Most people can't read.
我一生中毁掉了太多太多自己的工作。
I've ruined much much of my work in my life.
我完全走错了路，高估了人们的阅读能力。
I've just completely gone down wrong path by overestimating people's ability to read.
我认为，阅读这项技能如今反而变得更难掌握了。
And I think that reading is, if anything, getting harder to come by as a skill these days.
而我们现在所处的情况是，云代码要求你大量阅读。
And and this is the situation that we're in right now is that Cloud Code makes you read a lot.
所以我觉得今年剩下的时间我们都处于一种奇怪的过渡期。
So I think we're in a weird limbo for the rest of this year.
明白吗？
K?
在更好的用户界面出现之前，所有不会阅读的人都会处于严重劣势。
Where until the UIs arrive that are good enough for everybody who can't read, everybody who can't read is gonna be a severe disadvantage.
你能再多说说你观察到的很多人的阅读能力不足这个问题吗？
Tell me a little bit more about your observation that a lot of people or a lot of developers cannot read.
因为你之前在亚马逊工作过。
Because you were at Amazon.
那个地方据说靠着六个警报器和人们实际阅读来运行。
That place supposedly is running on six pagers and people actually reading.
真的吗？
Does
是吗？
it?
我的意思是，大多数人根本不会阅读。
I mean, most dude, most people can't read.
我不知道你知不知道，老兄。
You I don't know if you know this, man.
他们阅读速度真的很慢。
Like, I they just they they read really slow.
明白吗？
K?
还有AI，我的天啊。
And and the AI is I mean, come on.
对大多数人来说，五段就是一篇作文了。
To most people, five paragraphs as an essay.
记得在美国高中时，五段式作文是一种标准，也许在阿姆斯特丹，一篇要写一百段。
Remember five paragraph scenes in high school is a thing we have in America, Maybe I years were a 100 paragraphs in Amsterdam.
但对我们来说，五段已经很多了。
But to us, five paragraphs is a lot.
而这只是AI在清嗓子而已。
And that's like, that's the AI just clearing its throat.
对吧？
Right?
是的。
Yeah.
你知道，你得能读懂大段大段的文字。
You know, you gotta be able to read waterfalls of text.
所以我们正面临一个这样的世界：这种方式行不通了，因此你需要递归式摘要。
And so we're looking at a world where that won't work, and so you're gonna need recursive summarization.
你需要一个工厂。
You're gonna need a factory.
这很有趣，因为这就是为什么尝试用户界面如此重要。
And it's funny because like this is why, I mean, trying UIs is so important.
因为现在Gas Town的情况是，它是一个充满工人的工厂，而你通过电话与它交流。
Because Gas Town right now, the reason I say you can't use it is that it's a factory filled with workers and you're talking to it through a telephone.
你也可以走到窗边，敲打窗户，跟里面的工人说话，但你并没有真正置身其中。
You can also go and look through the window and pound on it and talk to the workers, but it's not like you're in it.
对吧？
Right?
有了用户界面，你就置身其中。
With a UI, you're in it.
你可以看到正在发生的一切。
And you can you can see what's going on.
对吧？
And right?
在 Gesta 中，大部分东西都是不可见的。
It's all invisible in Gesta by and large.
对吧？
Right?
很难看清楚。
You know, hard to see.
所以我真的认为，我要做个大胆的预测。
And so I really do think and I'm gonna I'm just gonna make a bold prediction.
我认为到今年年底，我们会看到相关演示，马上就能看到，但到年底，大多数人将通过与一张脸对话来进行编程。
I think that by the end of this year, and we'll see demos of it, like right away, but by the end of this year, most people will be programming by talking to a face.
脸，指的是屏幕上的一个面孔。
A face as in A face on the screen.
一张脸。
A face.
你的 AI，比如 Gesta 镇长，会是一只狐狸跟你对话。
Your AI, like the Gas Town mayor will be a fox talking to you.
你会说，为什么它不工作？
And you'll say, why doesn't it work?
他们会说，我去看看。
And they'll say, I'll go look at it.
它会像现在这样分派它的工作人员，但你是在和一张脸对话。
And it'll go spin off its workers just like it's doing, but you're talking to a face.
它会说话。
And it will talk
关于是的。
about yeah.
我认为这对大多数人来说是唯一可行的方式。
I think that's the only thing that's gonna work for most people.
很有趣。
Fascinating.
我们把这个预测记下来吧。
Let's let's write this down for prediction.
为什么
Why do
去把它做出来。
Go build it.
我不去做。
I'm not going to.
我们来谈谈Gas Town。
Let's talk about Gas Town.
你提到过Gas Town。
You mentioned Gas Town.
对于那些很多人听说过它的人，Gas Town到底是什么？
What for those that a lot of people have heard about it, what is Gas Town?
Gas Town是一个编排器。
Gas Town is an orchestrator.
所以2023年是完成阶段。
So 2023 was completions.
代码补全。
Code completions.
是的。
Yeah.
自动补全。
Autocomplete.
对。
Yeah.
那就是我们说它的时候
That's when we said it's
补全接受率卡片。
a Completion acceptance rate card.
你还记得吗？
Do you remember that?
哦，对对对。
Oh my yeah.
人们当时在测量它。
People were measuring it.
是的。
Yeah.
顺便说一句，这个指标很蠢。
Stupid metric by the way.
第二个是，但它很接近。
The second one was, but it was close.
它是一个衡量他们是否在尝试的替代指标。
It was a proxy for are they trying?
对吧？
Right?
然后还有2024年的聊天功能。
Then there was chat that was 2024.
对吧？
Right?
然后是2025年的智能体。
And then agents was 2025.
我们知道，你只要看看这个曲线就能明白：如果聊天是停滞在循环里，基本上，而智能体也只是在循环中聊天，那我们就把智能体放进循环里，让它成为编排器。
We knew you could just look at that curve and go, okay, well, if if chat is complacent as in a loop, basically, and agents are basically chatting a loop, well then we're gonna put orc we're gonna put agents in a loop, and that'll be an orchestrator.
对吧？
Right?
然后一堆这样的东西开始出现，我也自己做了一个，是的。
And a bunch of them started coming out, and I built one of my own Yep.
我自己的构想。
My own vision.
但那只是这么回事。
But that's all it is.
就是智能体在运行智能体。
It's agents running agents.
你能详细讲讲软件工程师的架构吗？
And can you talk through an a software engineer to its architecture?
它是怎么组织的？
Like, how is it organized?
我该怎么想象这个架构呢？
How can I imagine, you know, the setup?
是的。
Yeah.
当然。
Sure.
我的意思是，你看。
I mean, look.
Gas Town 非常复杂，这周一直出问题，因为我正在把它迁移到 DOLT，正是在这个过程中，我才意识到它有多复杂。
Gas Town is really complicated, and it's been really broken all week because I'm migrating it to DOLT, and that's where I actually learned how complicated it was.
它有很多功能。
It has a lot of features.
你正在把它迁移到哪里？
You're migrating it to?
迁移到 DOLT。
To DOLT.
它是一个新的数据库。
It's a a new database.
哦，好的。
Oh, okay.
是的。
Yeah.
DOLT 非常棒。
DOLT is DOLT is amazing.
DOLT 是一个基于 Git 的数据库。
DOLT is a Git back database.
它是一个 Git 数据库。
It's a Git database.
Beads 就是把 Git 和数据库强行拼在一起，但其实已经有数据库能做到这一点了。
It's Beads is just Git plus database crammed together badly, and there's actually a database that does this.
所以我在迁移到它。
So I'm I'm migrating to it.
但没错。
But yeah.
不管怎样，Gas Town 应该是这样的：你只跟一个市长沟通，他就是你的对接人，其他所有事情，他就直接派工人去处理。
Anyway, Gas Town is is is what it should be is one one mayor that you talk to, that's your your person, and then whatever else needs to get done, they're just gonna fire off workers.
嗯。
Mhmm.
明白吗？
Okay?
这其实比这稍微复杂一点，因为我觉得有两种类型的工作，人们会来回切换，而且大家在争论哪种才是对的。
It's a little a little bit more complicated than that because there are real I think there are two kinds of work that that that people go back and forth on, and people are arguing about whether they're the right one.
Anthropic 的一些人告诉我，这是最小化上下文的争论。
Some people at Anthropic told me it's the minimaxing context argument.
明白吗？
Okay?
明白吗？
Okay?
有些人认为你应该最大化你的上下文窗口，填满丰富、详尽的上下文，这样AI在对话时就会变得睿智、无所不知，他们只想刚好在上下文的边缘处。
There are people who believe that you should maximize your context window and fill it with rich, juicy context so that the AI is wise and all knowing when it's talking They to wanna, like, you know, just right at the edge of the context.
还有另一些人则认为：任务，搞定它。
And then there are others who are like, task, kill it.
任务，搞定它。
Task, kill it.
我希望
I want
最短路径的上下文窗口。
the shortest path window context window.
因为二次方的成本会增加，你知道的，没错。
Because of the quadratic x, you know, increase in in cost Yep.
再加上随着token数量增加，认知能力急剧下降。
Combined with the dramatic drop off in cognition as the tokens go up.
对吧？
Right?
是的。
Yep.
会迷失方向之类的。
Losing their track and stuff.
那么，哪种做法才是对的？
So so what which one's right?
我们这边有人完全倾向于最小化，也有人完全倾向于最大化。
And we've got people who are, like, full on in the in the in the minimizing and the and the and the the maxers.
我看了看我的工作流程，心想：嗯，Pull Cats 是最小化派，Crew 是最大化派。
And and I looked at my work workflow, and I was like, well, pull cats are the min and crew are the max.
我在 Gas Town 有两个基本的角色分工。
I have two fundamental role worker roles in Gas Town.
所以你那个非常简单的，就是内容很少的 Pull Cats，是吧？
So you have the do you have the the really simple one, which is the small content of the pull you have cats?
如果你有一个非常明确的任务，并且已经分解成多个子任务，那么你就可以找到它，而且它是自包含的。
A really if you have a really well specified task, all broken down into sub tasks, then you can find and and and it's like self contained.
它会明确告诉你该做什么。
It's it says what to do.
对。
Yep.
然后你可以把这个任务交给工人，让他们去完成。
Then you can give it to a worker and have it go do it.
对吧？
Right?
与此同时，你面对的是一个非常复杂的设计问题。
Meanwhile, you have a really difficult design problem.
你必须就这个问题进行一系列的讨论。
You're gonna have to have a series of conversations about this.
我注重最大化上下文。
I maximize context.
我会说，先读完这些文档，然后我们再聊。
I'm like, read all these docs and then we'll talk.
对吧？
Right?
所以这其实就是两种工作流程。
So it's just two workflows.
而且，我喜欢JG。
And like, I I like JG.
我的意思是，这听起来很容易想象，就像一个小镇，你知道的，在这个狂野的西部。
I mean, it sounds like it's I think it's so easy to imagine, like, it's a little town, you know, like, in this wild, wild west.
有市长，还有团队、工人们。
There's the mayor, like, the the crew, the the workers.
每个人都忙忙碌碌，房子正在被建造。
Everyone's buzzing and going around that the houses are being built.
实际上，这又是怎么运作的呢？
In practice, how does this work?
这么说，这对你们来说效果如何？
Like, how has this worked for you?
你听到过别人是怎么成功推进项目的，又是怎么失败的，甚至搞得一团糟的？
How what what are you hearing people get projects done versus not getting it done versus turning into absolute chaos?
关于Gas Town，你学到了什么？
What have you learned with Gas Town?
这是一次很棒的实验。
It's been a great experiment.
我的意思是，我真的
I mean, I've I've really
很享受这个过程。
enjoyed it.
实验。
Experiment.
对吧？
Right?
嗯，是的。
Well, yeah.
我的意思是，对吧？
I mean, right?
我的意思是，我出去做了一个故意不成功的东西。
I mean, I went out and built something that doesn't that deliberately doesn't work.
它太难了。
It's too hard.
对模型来说太难了。
It's too hard for the models.
就连OPUS 4.5都 barely 够用。
Even OPUS 4.5 is barely enough.
有趣的是，Anthropic的人告诉我他们喜欢这个，但有些人有点尴尬，因为这感觉像是我在绕开他们模型的bug，而事实确实如此。
And it's funny because the folks at Anthropic told me they they like it, but they're kind of embarrassed, some of them, because it feels like I've got all these workarounds for bugs in their model, which it kind of is.
对吧？
Right?
但这并不是一个bug。
But it's not a bug.
他们的模型从未被训练去当工厂工人，但很快就会了。
It's their model was never trained to be a factory worker and it will be soon.
所以很多Gas Town都会消失。
So a lot of Gas Town is gonna disappear.
很多复杂性、很多监控角色，他们所做的只是试图让Opus四或五更聪明，而这正是站在了苦涩教训的错误一边。
A lot of the complexity, lot of the roles that are monitoring, all they're trying to do is tell Opus four or five to be smarter and that's being on the wrong side of the Bitter Lesson.
对吧？
Right?
所以很多Gas Town会简化，扁平化为仅仅是最小值和最大值角色。
So a lot Gas Town's gonna simplify and flatten into just min min and max rolls.
嗯。
Mhmm.
你的最大值用 Crew，你的最小值用 Pole Cats，我认为这才是自然的形态。
Crew for your max, and your Pole Cats for your mins, and and I think that's the natural shape.
而且它们只会不断扩展。
And they'll just scale up.
那这些Pole Cats会不会就是呢？
And and could could that be the Pole Cats?
它们在某些时候可能只是子代理？
They might just be sub agents at some point, for example?
比如
Like
子代理啊，我的意思是，Polkads就是子代理。
Well, sub agent I mean, know, the Polkads are sub agents.
嗯。
Mhmm.
只是它们更高级，是第一等的。
It's just that they're they're more they're first class.
它们有自己的身份和收件箱。
You they have their own identity, inbox.
你可以和他们交流。
You can talk to them.
你可以通过分析他们的工作成果来计算技能向量，从而观察他们随时间的表现。
You you can actually see how they performed over time by computing skill vectors on their their work and things like that.
所以这比子代理更进一步。
So a little little bit more than that than sub agents.
我认为子代理的问题在于其不透明。
I think sub agents have the problem of being opaque.
我会启动一大批子代理去完成这项工作。
I'm gonna fire off a bunch of sub agents to go do this work.
然后你就说：好了，完成了告诉我一声。
And then you're like, okay, let me know when you're done.
但通过Gas Town，你可以直接查看他们，然后说：喂，你的邮猫出问题了。
Whereas with Gas Town, you can go look at them and be like, dude, your poll cat's not working.
我要去戳它一下。
I'm gonna poke it.
对吧？
Right?
所以Gas Town让你能更亲身体验，我不知道，更多的掌控感。
So Gas Town gives you a lot of hands on, I don't know, steering.
对吧？
Right?
它并不试图避开你。
It doesn't try to be it doesn't try to get out of your way.
它就在你面前，Gas Town。
It's in your way, Gas Town.
不过真的很有趣。
It's really fun though.
我挺想念它的。
I miss it.
它已经在我这儿停了几天了。
It's been down for a few days for me.
我跟你说，和普通的Claude相比，简直差远了。
And I tell you, working with regular Claude just stinks by comparison.
因为它就像一个创意工厂。
Because it's like an idea factory.
一旦它真正运行起来，所有系统都启动完毕，你就能同时处理很多事情，而且还能相对清晰地跟踪它们。
Once it's actually running and all booted up and everything, you can have so many things going on at once and actually track them reasonably well.
但现在它可能会让你陷入一种状态，让你不睡觉、不吃饭，这对你可不好。
Now it can suck you into a a mode where you don't sleep, you don't eat, and you start it's not good for you.
我其实想跟你聊聊，什么时候有空，说说这个行业现在正在发生什么。
And I actually wanted to talk to you a little bit about what's what's happening in the industry at some point.
但Gas Town本身，我的意思是，一切都是经过精心设计的。
But but Gas Town itself, I mean, like, it was all calculated.
所有的角色，你知道的，包括命名。
All the characters, you know, the naming.
我为什么当初要搞Gas Town呢？
Why did I even do Gas Town?
对吧？
Right?
为什么是
Why is
呢？
it?
为什么？
Why?
因为我想要推动舆论的边界。
Because I wanted to move the Overton window.
对吧？
Right?
因为去年，当我提到编排即将来临的时候，人们都会说不可能。
Because people last year, when I would say orchestration's coming, they'd say no.
代理不会形成群体，也不会有编排，你说的这一切都不对。
Agents aren't aren't no swarms, no orchestration, whatever everything you're saying is just not true.
而现在他们说的是，老兄，你有点太激进了。
And now what they're saying is, bro, you're being pretty aggressive.
对吧？
Right?
这是完全不同的讨论了。
Which is a different conversation.
他们现在说，嗯，你的群体，我不知道。
They're like, now they're like, well, your swarm, I don't know.
也许你的群体做不到那些事。
Maybe your swarm can't do blah blah blah.
这已经完全把讨论从不可能的范畴转变到了可能的范畴。
It's just completely shifted the conversation from the realm of impossibility to the realm of possibility.
所以，可以说你承担了远超你原本认为自己能应付的任务吗？
So is is it fair to say that you took on more than you you reasonably thought you could chew?
你选择了更雄心勃勃的目标，因为你希望检验这些模型的能力
You took on this more ambitious ones because you wanted to both stress test what these models can do
嗯嗯。
Uh-huh.
去发现
Find out
什么是发现？当然，也要玩得开心。
what's find out, and, obviously, just have some fun.
玩得开心。
Have some fun.
去发现接下来会发生什么，我一直在这么做。
Find out what's next, and I'm continuing to do that.
所以我的下一个计划是把100个Gas Town连在一起。
So my next thing is I'm gonna string a 100 Gas Towns together.
我们有一个社区，一个Discord频道，如果MoltBook能让人们为了乐趣贡献代币，就像他们为在MoltBook上运行自己的智能体所产生的推理付费一样。
We have a community, a Discord, and if MoltBook can get people to pitch in tokens for fun, like they paying they're paying you're paying for the inference of your your agent on MoltBook.
对吧？
Right?
所以，如果我把一百个Gas Town连接起来，我们决定一起构建一些东西，我们就会学会联合的机制。
So if I string a 100 Gas Towns together and we decide to build something together, we will learn the mechanics of federation.
我们可能正在重走以太坊的老路，但我们会做到的。
We're probably retracing Ethereum steps, but we will.
而且我们会创造出一些非凡的东西。
And and we're gonna come up with something remarkable.
这就像Molthub的大众版本。
It's like the people version of Molthub.
对吧？
Right?
Malta book，不管它是什么。
Malta book, whatever it is.
那关于Gas Town或它试图实现的目标，有哪些误解是你觉得已经有点跑偏、需要澄清的？
And what what are misconceptions about Gas Town or what it's trying to do that you feel it's kind of, you know, gone off a little bit of rails and is good to clean up?
我的意思是，首先，我不认为人们应该使用它，但他们已经在用了。
Well, I mean, for starters, I don't think people should be using it, and they are.
我真的认真的
And I I really mean it
我觉得人们不应该使用它，除非你在做研究，或者你真的明白这只是一个概念验证。
Well, I'm I'm gonna say people should not be using it, like, should not be using it except if you're doing research or or if you're, like, actually understand that this is just a proof of concept.
我最近和一些非常聪明的人讨论过，他们在寻找自己的问题领域中，哪些子集或类别可以立即让Gas Town在大型企业，比如一家财富500强公司中发挥作用。
So some some very, very clever people that I've been talking to have have been searching their problem spaces for subsets, categories that Gas Town could productively use today at a big company, a big Fortune 50 company, say.
哇。
Wow.
他们已经找到了一些现在就可以部署Gas Town的问题领域。
And they've they've identified some problem spaces that you could put Gas Town on today.
我当时就觉得，这想法真够巧妙的。
And I was like, oh, that's pretty pretty clever thinking.
其中一个例子是我接触过的一家公司，他们能为你在任何你想要的地区搭建定制化的数据中心——这是AWS从来做不到的。
One of them was this company I talked to that sets up bespoke data centers for you, okay, in any region you want, which is something AWS has never been able to do.
谷歌一直试图这么做。
Google's always tried.
他们说，要安装软件并检查一切是否正常，就得连续三个月痛苦地按按钮。
And they say it's just three months of miserable button presses to try to install the software and check that it all works.
而且验收标准非常明确。
And the acceptance criteria are very clear.
这几乎就像一个死循环。
It's, you know, it's almost a Ralph loop.
但他们认为Gas Town可以同时处理这个问题，最终找到一个能正常运行的数据中心，从而帮所有人省去麻烦。
But they think Gas Town could swarm it and and eventually converge on a data center that works, and and save all the people the trouble.
你明白我的意思吗？
You know what I mean?
我当时就想，哦，好吧。
And I was like, oh, alright.
这可能会显著提升他们为更多用户开通这些数据中心的能力。
And this could potentially meaningful move move the needle on their ability to open up more of these data data centers for people.
对吧？
Right?
哇。
Wow.
是的。
Yeah.
真想不到。
Go figure.
同一个人还告诉我，他一直在查看生产事故，发现当系统宕机时，他们的系统其实已经处于一种不确定、未知且损坏的状态。
And the same guy was telling me that he's been looking at production incidents, and he and he's realized their system is already in an indeterminate, unknown, broken state when they're down.
那么AI到底能让情况变得更糟到什么地步呢？
So how much worse can AI actually make it?
我提醒了他，实际上，情况可能会变得更糟得多。
Now I cautioned him and said, actually, it can make it a lot worse.
但他认为，在某些类型的故障中，你们可以进入调查模式之类的，对吧？那样可以加快处理速度。
But he's thinking along the lines that there are certain categories of outages where you could have them in investigation mode or whatever, right, where they could speed things up.
所以人们正在寻找那些模糊的问题。
So people are looking for the fuzzy problems.
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。