克劳德代码的创造者：全球最热门的软件

如果它能多跑几轮，检查自己的输出结果，不断迭代，那效果肯定会好很多。

If it can loop a few times and it can check the output of its work, it can iterate, then it's just going be much better.

所以，如果Claude Code写的代码真像你说的那样，比你还要漂亮，那你们这些软件工程师到底还在干什么呢？

So if Claude Code is writing beautiful code, as you say, that looks better than yours, what are you and every other software engineer in the world actually doing here?

你觉得自己在这个过程中应该扮演什么样的角色？

Like, what do you envision as your role in this process?

编程这门学问挺特别的。

Programming is this kind of weird discipline.

它已经以某种形式存在了差不多八十年了。

It's been around in some form for, well, like eighty years maybe.

我祖父当年在苏联就搞过编程。

My grandfather actually programmed in the Soviet Union.

哦，真的吗？

Oh wow.

是的。

Yeah.

他那时候用打孔卡编程，因为当时写代码根本不是现在这种软件方式。

And he programmed punch cards, because back then, the way you write code, it wasn't software.

就跟今天差不多。

It's like today.

你是在纸上编程。

You programmed in paper.

然后把纸卡放进一台大机器里，它算完之后，几盏灯亮起来，答案就出来了。

And then you fed the paper into a big machine and it did some calculations, then a few lights lit up with the answer.

我妈妈小时候常讲，我爷爷会把一大摞打孔卡带回家，她就拿蜡笔在上面乱画。

My mom, growing up, she would tell this story about my grandpa bringing back these big stacks of punch cards home, and she would draw all over them with her crayons.

以前编程是实实在在的体力活。

Programming used to be physical.

在打孔卡之前，完全是机械式的。

Before punch cards, it was purely mechanical.

那会儿算是电子设备了。

It was kind of electronics.

如果你想想苹果一号电脑，它完全是电子元件组成的。

If you think about the Apple One computer, it was all electronics.

史蒂夫·沃兹尼亚克用芯片把它做出来的。

Steve Wozniak built it as chips.

虽然也有软件，但所有的逻辑其实都体现在芯片里。

There were some software, but really all the logic was expressed in chips.

后来情况变了。

It changed.

六十年代的时候，人们开始意识到：也许我们没必要非得把代码写在纸上或者硬接在硬件上。

Sometime in the sixties, people realized, Okay, I think we can write code and it doesn't have to be paper or hardware.

我们可以把它做成软件。

We can probably put it in software.

再后来，有人突然反应过来：等等。

Then at some point people realize, Oh, wait.

我觉得我们还能做得更进一步。

I think we can go beyond this.

我们可以把整个操作系统都放进去。

We can take the entire operating system.

操作系统没必要非得用芯片实现。

The operating system doesn't have to be chips.

它也可以完全是软件。

It can be software also.

这其实是一个重要的领悟。

That was a realization.

苹果II和七十年代初那一代电脑，就是从那时开始的。

That was the Apple II and that generation of computers in the early seventies that started that.

过去五十年里，我们运行的操作系统和内核软件，全都是软件形式的。

For the last fifty years, the operating system, the kernel software that we run, it's all in software.

根本不是硬编码在硬件里的。

It's not really in hardware.

我们发布QuadCode之后，开发者就不再像过去五十年那样直接写代码了。

What changed when we released QuadCode is developers stopped writing the software directly the way that they've been doing the last fifty years.

他们开始和模型对话，由模型来编写软件。

They started talking to the model, and the model writes the software.

现在我们又往上走了一层，有了循环、例程和QuadTag。

Now we're actually going up one more level, and now we have loops and routines and quad tag.

这些变化意味着，我们再次往上提升了一个层级。

What's happening with these is we just went up one more level.

你跟模型说话，模型再去和其他模型交流。

You talk to the model, the model talks to other models.

那些模型会生成源代码。

Those models write the source code.

这太不可思议了，我们五十多年都困在同一个模式里，结果两年内接连发生了两次飞跃。

This is crazy because we've been stuck in this one place for fifty years, and we just had two leaps in two years.

事情就是这么发展的。

That's what's happened.

回看我过去的工作，我总能进入深度专注状态，花上几天甚至几周时间写一段软件。

When I look at my work, I used to have this deep focus mode, and I would spend days or weeks on writing one piece of software.

现在我做的就是跟Claude交流。

Now what I do is I talk to Claude.

任何时候，我都会运行几个到几百个、甚至上千个协作单元，它们一起合作开发软件。

At any point, I have a few quads running, sometimes hundreds, sometimes thousands, they're collaborating on building software together.

这让我能腾出精力，去想更多可以让它们做的事。

This frees me up so I can think of more things for them to do.

有趣的是，我根本想不完该让它们做什么。

And the funny thing is I just never run out of things for them to do.

我听说，早在Claude Code出现之前，甚至在AI编程出现之前，软件工程师的职业生涯中都会有一个节点，那时他们就不再写代码了。

I've heard even long before Claude Code, even long before AI coding, my understanding is that in the career of a software engineer, they hit a point where they stop coding period.

他们可能更多地在白板前讨论，或者花大量时间招聘之类的。

And maybe they're like on some whiteboards or they spend a lot of time hiring, etc.

每个软件工程师最终都会脱离亲手敲代码的阶段。

But every software engineer sort of graduates out of typing out code.

所以这个问题可能根本就不适用于你。

But so this question may not even apply to you.

今天Anthropic还有人写代码吗？

Is there anything Anthropic today?

还有人在手动敲代码吗？

Is there anyone typing out?

现在还有什么事是需要人工写代码的吗？

Are there any things for which someone is typing out code?

说来有点好笑。

So, you know, it's funny.

我职业生涯中有一段时间，确实没再写代码了，因为被推去搞管理、写文档那一套。

I in in my career, there was a point where for a little while I stopped writing code because I was pushed to the same thing, like to go to management and writing documents and stuff.

作为工程师，我那时候特别不开心。

And I just felt as an engineer, I was so deeply unhappy.

因为他们都讨厌那样。

Because they all hate

啊。

it.

是的

Yeah.

对

Yeah.

因为作为工程师，我

Because I as an engineer, I

我想说，做记者也是这样。

wanna I wanna say, a journalist too.

一旦你成了编辑，基本上就不再写了。

Like, once you become an editor, you basically stop writing.

没错

Right.

没错

Right.

没错

Right.

你知道吗，对有些人来说，这简直太棒了。

You know, for some people, that's amazing.

这正是他们最拿手的地方。

Like, that's the thing they're really good at.

但对我来说，我就想动手做点东西。

But for me, like, I wanna build.

我想写代码。

I wanna code.

这才是我真正喜欢做的事。

That's what I like to do.

是的。

Yeah.

所以当我看看Anthropic这边，就我个人而言，从去年十一月开始，我写的每一行代码都是由AI生成的。

So when I look across Anthropic, for me personally, a 100% of my code has been written by quad code since November of last year.

明白了。

Okay.

现在所有Quad Code、所有协作成果，以及我们的所有产品，都是用Quad Code写成的。

This is now true for all of quad code, all of co work, all of our products are written using quad code.

我们的基础设施和研究代码中，用Quad Code编写的比例也在持续上升。

It's also true for an increasing percentage of our infrastructure and also our research code.

在Anthropic内部，平均下来大概有98%的代码都是Quad Code生成的，差不多就是这样。

And so across Anthropic, think the average is something like 98% quad code or something like that.

那剩下的2%是什么内容呢？

And that 2%, what is this?

比如优化芯片之间通信的代码？

Like code that optimizes the way chips talk communicate?

到底哪些部分还是让人来手写更合适？

What what's the 2% that's still it's better to have a human typing it out?

是的。

Yeah.

还是存在一些特定的领域。

There's still, a few pockets.

比如有个经典例子，配置文件里有时候就改两个字符，还不如自己动手来得快。

Like, one classic example is, like, configuration files where, you know, it's like a two character change or, you know, or something, and it's faster to just make it yourself.

明白了。

Okay.

说实话，这种地方很快就会消失。

Honestly, think this is going to go away really fast.

我们已经开始在客户那边看到同样的趋势了。

We're starting to see this with our customers also.

当初我们刚推出Claude Code的时候，跟谁解释都费劲，现在人人都在用。

At the beginning when we started Claude Code, was really hard to explain to anyone what is this thing, but now everyone uses it.

我经常给硅谷的Y Combinator加速营做分享。

Like I do this talk for Y Combinator batches, the startup incubator in Silicon Valley.

我刚开始做这些分享时，会问大家：请举手，有多少人用过Quad Code？

When I first started doing the talks, asked everyone, please raise your hand if you use quad code.

当时只有寥寥几只手举起来。

There was a few hands that went up.

后来有一次我做这些演讲时，每个人都举起了手。

At some point I did these talks and just every hand goes up.

所以我就不问这个问题了。

So I stopped asking this.

现在我问的是：谁是100%用Claude Code写代码的？

Now the question that I ask is, who writes a 100% of their code using quad code?

我第一次问的时候，大概有四分之一的人举手。

And the first time I asked this, maybe a quarter of their hands went up.

现在超过一半了。

Now it's a little more than half.

我打赌下次问的时候，所有人都会举手。

And I bet the next time I ask, it's gonna be everyone.

我们的客户规模各不相同。

Our customers range in size.

有Airbnb和Ramp，也有最大的那些公司。

There's Airbnb and Ramp, and then also the biggest companies.

还有Salesforce、德勤和埃森哲。

There's Salesforce and Deloitte and Accenture.

这些大型公司都在使用Claude Code，而且也看到了同样的趋势。

All these very big companies also use Claude Code, and they're seeing the same thing.

越来越多的代码是由Claude Code自动生成的。

A bigger and bigger percent of the code is being written by quad code.

不过我想再深入问一下，如果现在你招聘工程师，除了写代码的能力，你更看重哪些具体的技能？

Just to press you on this point, though, if you're hiring engineers nowadays, like, what are the specific skill sets that you're looking for if it's not necessarily the ability just to write code?

我开始觉得，把工程、设计、产品、用户研究和数据科学分得这么清楚，这种思维方式已经过时了。

I've started to think that this idea of engineering versus design versus product versus user research versus data science, I think this is the old way of thinking about it.

我现在觉得，因为每个人都能写代码了，角色的分工也发生了一些变化。

My feeling now is because everyone can write code, the roles shift a little bit.

比如在Claude Code团队里，我就看到这种情况——团队里的设计师、产品经理、工程主管，每个人都会写代码。

And I'm seeing this on the Claude Code team, for example, because on the Claude Code team, everyone writes code, including our designers, product managers, engineering managers.

大家都写代码，因为现在写代码太容易了。

Everyone writes, because it's easy.

现在做起来容易多了。

It's much easier to do now.

这其实太棒了，因为我的设计师不用再每次发消息问我：

It's actually awesome because my designer doesn't have to message me every time by, hey,

能

can you

把按钮往右移一个像素吗？

move the button over by a pixel?

她自己就能搞定。

She can just do it herself.

对大家都好。

It's great for everyone.

所以我开始觉得，这些角色其实是在朝相反的方向分化。

And so I've started to think that the roles are actually segmenting in kind of the opposite way.

我注意到人们正在分成两类：原型制作者。

And I've started to see people kind of split into prototypers.

这些人特别擅长快速把最初的想法摸索出来，然后迅速迭代成实际产品。

These are people that are amazing at just figuring out, like, what is that first idea and, like, very quick iteration into builders.

一旦有了新点子，他们就知道怎么把它做出来，推向市场。

So once there's a new idea, figuring out how do you actually build this and bring this product to market.

还有一类是维护者。

Then there's maintainers.

当软件达到一定规模后，就是这些人来负责长期维护。

These are the people that once the software is at scale, they can maintain it.

还有一种人，我称之为增长者，或者叫扩张者。

There's something that I call growers or maybe scalers.

他们接手已经验证了市场的产品，把它进一步做大。

These are people that take an idea and this product that exists that has product market fit and then scale it up.

让产品规模扩大十倍、百倍。

Scale it 10x, 100x.

顺便说一句，现在在Anthropic，这类人特别吃香。

By the way, these people are very popular at Anthropic now.

接下来我认为最后一个角色是收尾的人。

Then I think the final role is sweepers.

我不确定你们有没有更好的叫法，但我管这叫收尾者、清洁工之类的。

I don't know if you guys have a better idea for the name, but I call it a sweeper or janitor or something.

这其实是个非常重要的角色。

It's actually a very important role.

就是打磨产品、优化基础设施、精修代码，把所有粗糙的地方都处理干净。

It is about polishing the product, polishing the infrastructure, polishing the code to get rid of all the rough edges.

因为作为用户，当你用上特别精致的软件时，你一定能感受到。

Because as a user, when you use really polished software, you feel it.

完美主义者。

The perfectors.

完美主义者。

Perfectors.

完美主义者。

The perfectors.

是的。

Yeah.

他们去把产品打磨到完美。

They go and make the product perfect.

没错。

That's right.

没错。

That's right.

他们尽力了。

They try to.

对。

Yeah.

既然我们谈到设计，还有我之前说过的一个观点——工程师某种程度上也得成为产品经理和专家，你以前说过，Claude Code 的命令行界面其实只是权宜之计，因为模型进步得太快了，根本没必要专门为它设计一套完整的用户界面。

Since we're on the topic of design and this idea that I guess engineers are also gonna have to become in some ways product managers and specialists, you've said before, I think, that the command line for Claude Code was basically a stopgap measure because the models were improving so quickly that it didn't make sense to design, like, a whole user interface around it.

现在还是这样吗？

Is that still the case?

然后，你能想象将来会有一个更……我不该说传统用户界面吧，毕竟命令行本身就已经挺传统的了。

And then, you know, could you envision at some time having, like, a more I don't wanna say traditional user face because in some ways, the command line is, like, the traditional, yeah, user face.

我记得九十年代中期在MS-DOS里敲命令的时候，那种感觉，真觉得自己像个编程天才。

And I have very fond memories of, you know, entering commands in MS DOS in, like, the mid nineties and feeling like an engineering genius at the time.

但你觉得，未来会不会对这种界面做出重大的改变？

But could you imagine, like, a a substantial change to that interface at some point?

我有点不敢肯定，因为刚才我在彭博办公室走了一圈，每个人桌上都摆着彭博终端。

So I'm hesitant to say because I was walking around the Bloomberg office area, everyone has their Bloomberg terminals.

对。

Yeah.

是啊。

Yeah.

彭博终端确实很受欢迎。

Bloomberg, definitely a fan the kind Terminal.

很多人可能不知道，Claude Code最初是在终端里起步的，但很快就走出了终端环境。

Something that a lot of people might not know about Quad Code is we started in a terminal, but very quickly we actually got outside of the terminal.

Quad Code 为所有主流 IDE 都提供了插件，你可以直接用它们，不用再开终端。

Quad Code has extensions for all the popular IDEs that you can use instead of the terminal.

我们还有一个桌面应用，特别受欢迎，聊天、写代码、协作，全都整合在一个地方。

We have a desktop app that's also very popular, and it has chat and code and co work, it's all in one place.

我们还有 Android 和 iOS 的移动应用。

We have mobile apps for Android and iOS.

其实我现在用 Claude Code 最多的方式，是通过 Slack。

And actually the way that I use Claude Code the most nowadays is through Slack.

就跟跟同事聊天一样，直接跟 Claude 在 Slack 里对话。

And it's just talking to Claude and Slack like a like I would to a coworker.

在我转用 Slack 之前，我主要是在手机上用 Claude。

And before I moved over to Slack, I was actually using Claude mostly on my phone.

那时候我基本就靠 iOS 应用，随时跟它聊天。

So I was mostly on the iOS app just talking to it.

我偶尔还是会用终端，但说实话，现在基本不用了。

I I you know, I use terminal sometimes, but overwhelmingly, I I actually don't nowadays.

挺有意思的。

Interesting.

你提到 Slack 机器人这点很好，这让我想到另一个我一直挺好奇的问题，因为 AI 模型和 AI 框架跟传统的企业软件不太一样。

I'm glad you brought up the Slack bot because this gets into a different sort of line of questioning that I've been curious about because AI models, AI harnesses, they're a little bit different than traditional enterprise software.

比如，你会听到有人说，哎呀，我的上下文窗口满了，接下来两小时都没法继续写代码了。

For example, you see people talk about like, oh, I ran out of space in my window and I'm not going be able to code again for another two hours.

然后他就得去散个步什么的。这对用过 Slack 或者成千上万种其他企业软件的人来说，肯定是个挺特别的体验。

I'm going to go take a walk or something, which is not anyone who's used Slack or a million other enterprise software, that's got to be a unusual experience for them.

不过我从商业角度有个问题想问。

But here's a question I have from a business perspective.

随着 Fable 的发布，首次出现了不是所有人都能直接升级到最新模型的情况。

With the launch of Fable, for the first time, not everyone was just able to like, now I'm upgrading to the newest model, etcetera.

之前有个类似白名单的机制，比如 Project Glasswing，再加上像白宫出口管制这类问题，后来都解决了。

There was sort of like a white list with Project Glasswing and then some of these questions about like, obviously with the White House and export controls, etcetera, that got resolved.

但抛开监管问题不谈，我们是不是正走向一个局面：最先进的模型不会同时向所有人开放？

But even setting aside the regulatory questions, are we heading into a world in which each most advanced model will not be distributed to everyone at the same time.

从商业角度看，这就像是某家公司决定全面采用Anthropic的产品。

From a business perspective, it's like, okay, some company wants to be an Anthropic shop.

这该让他们感到担忧吗？

Should that be a source of anxiety for them?

你有没有发现，他们其实担心最强大的模型不会同时对所有人开放？

Or have you seen it as a source of anxiety for them that the most performant models may not go to everyone all at the same time?

总的来说，我们一直努力为所有人提供最强大、最智能、最高效的模型，因为这符合我们的利益。

In general, we try to give everyone the most performant models we can, the most intelligent models, and the most efficient models, because we are incentivized to do this.

没错。

Yeah.

对吧？

Right?

我们的业务就是模型，自然希望把最好的模型带给用户。

Like our business is models, and so we want to give people the best models we can.

比如说，我每天都会用Fable。

And so, you know, for example, I use Fable every day.

我们的客户用的也是同样的东西。

That's the same thing that our customers use.

对。

Yeah.

你提到模型 rollout 不是同步向所有人开放，我觉得你可能把像 Mythos 这类本身就更危险的模型，和日常用的这些模型搞混了。

When you talk about the rollout of the model that's kind of not even, that doesn't go to everyone at the same time, I think you might you might be thinking of like Mythos and models are that are inherently more dangerous than these kind of day to day models.

像 Mythos 这种模型比较特殊，因为它存在 Fable 没有的极高风险。

And something like Mythos, it's a bit of a special model because it it has hyper risks that Fable doesn't.

所以我们才推出了 Glasswing。

And so this is why we had Glasswing.

我们之所以在发布上格外谨慎，是因为如果第一天就让所有人都能用 Mythos，大家只会到处乱试。

This is why we have been thoughtful about the rollout because if we just gave everyone Mythos access on day one, everyone would just kind of be hacking.

原因在于，Mythos 在发现零日漏洞和利用方式方面实在太厉害了。

And the the reason is that Mythos is just very, very good at finding zero day vulnerabilities and exploits.

因此在发布这类模型时，我们特别重要的是先让靠谱的用户用上，给他们一点时间领先，再向大众开放。

And so for us, like, in that rollout, it was just really important to give it to the good guys first and to give them a head start before we give it to everyone.

你看到的正是这种极其谨慎的发布方式的延续。

And you're seeing kind of the continuation of that very careful rollout.

这是一种颠覆性的能力，所以我们必须谨慎行事。

It's a step changing capability, so we have to be thoughtful.

与此同时，还有Fable，这是我个人使用的Mythos版本。

At the same time, there's Fable, which is the version of Mythos that I use.

这个模型没有那些强大的黑客能力，现在所有人都能用它。

That's the model that doesn't have all these kind of same hacking capabilities, and that's the thing that everyone has access to now.

我担心的是，假设我不是Anthropic最大的客户之一等等。

Here's what I would worry about, which is like, let's say I'm not one of Anthropic's biggest customers, etcetera.

我们知道算力是稀缺的。

We know that compute is scarce.

对吧？

Right?

否则Fable早就24小时开放了，而不是只在一段时间内默认提供。

Otherwise, Fable would be on for twenty four hours as opposed to like, it's only gonna be in the model as a default for, like, some period of time, etcetera.

我担心的是，如果我不是那种长期重度使用Claude的客户，会不会因为这个原因，我的Fable访问权限就比那些死忠用Claude的公司少？

What I would be worried about is that, oh, if I'm not a sort of, like, heavy and consistent Claude shop, do I have to worry that my access to Fable, set aside Mythos, will not be as much as a company that is like a ride or die Claude shop?

不会的。

Oh, no.

每个人都能用。

Everyone gets access.

而且你看看那些公司，他们通常不会用那种有使用限制的订阅方案。

And also, when you look at companies, they're not using subscription plans typically that have rate limits.

大多数公司更愿意按使用量付费，这样他们能更好地掌控成本。

Usually companies prefer to pay per token because that way they can control it.

这样他们能更准确地预测开支，工程师也不会遇到使用上限的问题。

They can forecast a little bit better, and also their engineers don't hit rate limits.

所以他们能掌握更多的主动权。

So they have a little bit more control I that

我其实想问一下这方面的问题。

wanted to ask about this actually.

我觉得现在大家都清楚，比如那种Claude Code的超级用户，或者有点AI妄想症的人，天天忙着搭一堆网站和各种程序。

So I think at this point we all know, you know, like a Claude Code super user or someone with AI psychosis who's like setting up a bunch of websites and different programs on a daily basis.

然后还有公司也在用Claude Code。

And then you have companies that are using Claude Code.

我猜如果一家公司有两千名员工都在用这个工具，再加上有风控团队、各种规章制度，他们产生的输出肯定和单个超级用户不一样。

And I imagine if you have 2,000 employees that are using this tool and you have, you know, risk management committees, rules, that sort of thing, the output is going to be a bit different to the individual superpower user.

你觉得这两种使用场景之间，最明显的区别是什么？

What are the key differences you've noticed between those two?

另外，公司真正要采纳这些工具时，最大的障碍在哪里？

And I guess, what are the big sticking points when it comes to companies actually adopting these tools?

我通常看待公司采用Claude Code的方式，是把它看成一个阶梯，得一级一级往上走。

Usually the way that I think about companies' adoption of quad code is I think of it as this ladder that you have to go up one step at a time.

不可能一上来就让所有人都用它干所有事。

You don't just jump straight to the top of everyone using quad code for everything.

最终是能走到那一步，但得一步一步来。

You get there, but you get there a step at a time.

第一步是先用某种AI工具，慢慢把它引入进来。

The first step is you use some sort of AI and you kind of start to bring this in.

通常一开始是通过IDE或者其他程序使用Claude。

And usually it's like Claude through an IDE or through some other program.

这就是你使用Claude的方式。

And this is how you use Claude.

第二步是给所有人配Claude Code，还有CoWork，现在还有Tag。

The second step is you give everyone Claude Code and co work, and nowadays tag also.

刚开始的时候，通常是一个工程师对应一次Claude Code会话。

The way that it usually works at the very beginning is one engineer, one Claude Code session.

要么是一个市场人员对应一次CoWork会话，一次只进行一个会话。

They're just running one session at a time, or one marketer, one co work session.

就是一对一的模式。

It's just one to one.

你每次只和一个AI助手对话。

You're talking to one quad at a time.

在这个过程中，你得考虑设置一些安全边界。

As you do this, you want to think about guardrails.

当然，系统本身已经自带了不少防护机制。

Obviously, there's a lot of things that comes out of the box.

我们有按用户设置的使用预算控制。

We have per seat spend controls.

我们还提供了顾问型模型。

We have advisor models.

在企业层面，你可以调整使用强度的级别。

You can pick effort levels at the enterprise level.

总之，控制方式有很多种。

So there's just all sorts of ways to control this.

同时，你也得关注安全性方面的问题。

Then you also should think about the safety side.

比如沙箱隔离这类措施。

So this is sandboxing and things like this.

一般来说，我们默认就把所有安全设置都调好了，你根本不用操心。

In general, we try to make all the safety settings correct by default, you don't have to think about it.

它自己就能正常运行。

So it just kind of works.

但你觉得有什么障碍吗？

But do you see an impediment?

我不确定。

I don't know.

挑几个例子说说。

Pick a couple.

我不确定。

I don't know.

比如说，辉瑞公司这样的客户。

You know, like, oh, Pfizer.

我们可以向他们推销一些Claude或Claude Code的使用席位。

Let's sell some Claude or Claude Code seats to them.

刚开始时，用户真正理解并接受这个东西，这个障碍到底有多大？

How much is just that, like, initial sticking point of them literally figuring out?

我们知道，大公司对让员工在电脑上安装任何软件都特别谨慎，更别说那种功能越强、越需要完全系统权限的软件了。

We know that big corporations are very anxious about letting users download any software to the computer, let alone software whose maximum capability comes when it has the deepest root access to the entire file system and everything.

从商业角度看，像‘我们不敢让这么强大的软件放在员工电脑上’这种顾虑，到底有多普遍？

How much of a sticking point business wise are you seeing in just companies like, we do not feel comfortable with this such a powerful piece of software sitting on employee desktops?

我觉得几年前，大家确实有点不适应，因为这还是个全新的概念。

I think a couple years ago, there was some level of discomfort because this was a really new idea.

但随着时间推移，员工用得越来越熟练，公司也慢慢建立了信心，接受度就越来越高了。

But I think what's happened over time is as employees usage gets more sophisticated, as companies build up their confidence, they get more comfortable with it.

这背后也有原因，因为我们一直在安全、对齐、防护和隐私上投入了大量精力。

You know, it helps because we spend so much effort on safety and alignment and security and privacy.

这些对我们来说至关重要。

It's just extremely important to us.

所以我看那些早期就采用的公司，都是循序渐进的：从每个工程师一台设备，到十台，再到一百台，现在有些已经扩展到每人一千台了。

And so, like, when I look at companies, the ones that adopted it kinda early on, they've gone up this kind of adoption ladder, and they went from one quad per engineer to 10 quads to a 100 quads, now some to a thousand quads per engineer.

每个人都是一步一步慢慢适应的。

And everyone kinda makes it up one step at a time.

现在你看看纽约最大的那些银行，一些顶尖的制药公司，连NASA都在用Claude Code。

And so, now you look at all the biggest banks in New York, you look at some of the biggest pharma companies, NASA uses Claude Code.

现在这东西已经无处不在了。

Now it's everywhere.

出于好奇，你有没有发现不同公司在设置权限和安全策略上有差异？

Out of curiosity, do you see differences in how different companies, I guess, customize permissions, safety permissions?

我知道你说过你们尽量标准化，让用户一上手就能用，但我猜还是有不少客户会自己调整。

I know you said you try to standardize them so that they're easy to use from the get go, but I imagine you still have customers that will change things up.

是的，当然有。

Yeah, absolutely.

Claude Code 的配置灵活性非常高。

There's so quad code is just very, very configurable.

天哪。

There's gosh.

我记不清具体数字了，但肯定有好几百项可调的设置。

I I know the exact number, but it's gotta be, like, many hundreds of different settings that you can change.

现在估计得有四五百项了吧。

There's, you know, probably four or 500 at this point.

厉害的是，你直接让Quad帮你搞定就行，根本不用去翻文档。

The cool thing is you can actually ask Quad to to do it for you, so you don't even have to read the documentation.

Quad自己清楚所有的设置。

Quad knows its own settings.

全球增长正翻开新篇章，而很多进展都发生在非洲。

A new chapter in global growth is being written, and much of it is happening in Africa.

非洲人需要加大投资。

Africans need to invest.

这里有大量机会等着去把握，生意等着去争取。

There are deals to be done and business to be won.

我是詹妮弗·扎巴萨贾。

I'm Jennifer Zabasaja.

每周的《下一代非洲》播客，我们都会追踪影响非洲未来的资本流动和政治变化。

Every week on the Next Africa podcast, we track capital flows and political shifts shaping the continent's future.

非洲的数字化进程将推动它的增长。

The digitalization of Africa is gonna power its growth.

应对像艾滋病这样的全球性挑战是完全有可能的。

Riding the world of something like HIV is possible.

非洲的人口增长规模实在太大了。

Population growth is so enormous in Africa.

在苹果、Spotify 或你常用的任何播客平台收听《下一代非洲》。

Listen to Next Africa on Apple, Spotify, or wherever you get your podcasts.

总的来说，现在网上到处都是AI生成的代码，很多开源库和数据库里都充斥着这些内容。

With coding in general, the internet is now awash in AI generated code, and a lot of the open source libraries and databases filled with that.

几年前，这些数据还是干净的训练素材呢。

A few years ago, this was pristine training data, etcetera.

你有没有注意到，他们说的那个叫‘模型坍塌’的现象？

Do you see, what do they call it, model collapse or something?

现在有没有出现什么问题？

Are there issues that are arising?

就算不谈Claude Code，单说AI生成代码被用来训练代码模型这件事，会不会改变技术进步的节奏？

Even setting aside Claude Code, just coding capabilities from essentially code learning from AI generated code, and does that change progress curves at all?

说到AI的扩展，人们经常提到的都是扩展定律。

Look, when you think about AI scaling, the thing that people talk about often is the scaling laws.

对。

Yeah.

如果你不了解，扩展定律是八年前、十年前左右的一篇论文提出的。

And for people that don't know, the scaling laws, it was this paper that was written maybe eight years ago, ten years ago or something.

那是第一篇系统阐述模型智能如何随训练规模增长的论文。

It was the first paper that described how model intelligence scales as a function of training.

谈到训练，其实涉及几个方面。

When you think about training, there's a few pieces.

一是投入的算力，二是训练数据，三是神经网络的规模，还有推理时的计算量，也就是模型能花多少时间来思考。

There's the compute that you put into it, the data that you put into it, and then the size of the neural network, and also the test time compute, so the amount that the model gets to think.

有意思的是，看完这篇关于缩放损失的论文后，最初的几位作者在发表后就分出去创办了Anthropic。

What's interesting is when you look at the scaling loss paper, actually the first few authors after writing the paper, they branched off and they started Anthropic.

达里奥和萨姆都在这篇论文里署名了。

This is actually Dario is on the paper and Sam is on the paper.

杰瑞德也在上面。

Jared is on the paper.

他们就是我们的创始人。

These are our founders.

原因在于他们看到了这一点

And the reason is they saw that

哦，我没想到你们这儿也有个萨姆。

this Oh, I didn't realize you guys had a Sam too.

是的。

Yeah.

他当初是我们第一位首席技术官。

He was our he was our first CTO.

明白了。

Got it.

哦。

Oh.

是的。

Yeah.

关于规模扩展，有趣的是它的表现非常平稳。

And the thing about the scaling was is they're remarkably smooth.

更奇怪的是，它实际上似乎还在加速。

And what's also kind of weird is it actually seems to be accelerating a bit.

这比我们八年前的预测还要超出一些。

It's a it's a bit beyond what what we guessed, eight years ago or whatever.

所以，它还在持续扩大。

And so, yeah, it just continues to scale.

总会有瓶颈存在。

There's always bottlenecks.

总会遇到各种问题，但总能一一解决，然后继续扩大规模。

There's always issues you hit, and you always work through it, and then you keep scaling.

Fable 这边似乎还在持续发展。

It just seems to be continuing with Fable.

你知道吗，开头我们提到过今年早些时候那场关于大型软件和 SaaS 的恐慌。

You know, in the intro, we talked a little bit about the big software SaaS scare earlier this

今年。

year.

末日论。

Apocalypse.

是的。

Yeah.

SaaS 末日论。

SaaS apocalypse.

现在看来情况缓和了一些，但大家心里还是隐隐担心，会不会人人都得自己写程序了。

And it seems to have died down a little bit, but there is definitely this lingering anxiety about whether or not everyone's just gonna be coding their own programs.

在你看来，未来会有多少人直接自己设计软件呢？

Can you weigh in on the extent to which people are gonna be just designing their software, their own software, in your view?

另外，我很好奇，总的来说，在硅谷，你现在是不是挺受欢迎的？

And also, I'm very curious, just in general, in Silicon Valley, are you, like are you a popular guy at the moment?

一方面，你处在人工智能这种热门技术的前沿，

There's a bunch of you know, on the one hand, you're on the cutting edge of AI, the hot technology.

但另一方面，也可能有人觉得你在抢走一些SaaS专家的饭碗。

But on the other hand, there might be a sense that you're putting some SaaS experts out of their out of their jobs.

我是这么看的：你们听说过‘七大优势’这个框架吗？

The way I would think about it is do you guys know this, like, seven powers framework?

没听过。

No.

我这个人特别喜欢历史，也特别依赖各种分析框架。

It's it's like I I'm, like, a big kind of history person and, a big framework person.

我就是爱把工作放到更大的背景里去看，这样才清楚什么重要、什么不重要。

I just, I love anything that puts my work into context to help me understand kinda what matters and what doesn't.

这七个优势是一个特别棒的商业框架，我特别喜欢一个播客，经常聊这个。

So the seven powers is just this, like, amazing business framework, and there's this other podcast that I love that talks about it a lot.

所谓优势，其实就是讲企业有哪些护城河。

Powers, they essentially talk about what are the moats in business.

大概有七种吧。

There's seven of them, roughly.

其中一种护城河是规模经济。

One moat in business is scale economies.

你规模越大，单位成本就越低。

As you scale, your marginal cost goes down.

这是一种天然的优势。

This is a natural mode.

另一种是网络效应。

Another one is network effects.

用你产品的人越多，每个用户能获得的价值就越大。

The more people that are using your product, the more value any individual person using the product gets.

另一个优势是转换成本。

Another mode is switching costs.

如果你深度依赖某个软件，换起来特别麻烦，那这可能就是一个优势。

If you're super locked into some software and it's really hard to switch, that potentially is a mode.

像这样的优势还有很多。

So there's a bunch of modes like this.

我觉得未来几年，随着Claude Code这类产品的出现，其中一些优势的重要性会下降。

The way that I think about what's happening is some of these modes are gonna get less important over the next couple years because of products like Claude Code.

如果你想从厂商A迁移到厂商B，只要问Claude一声：‘你能帮我迁移吗？’

So if you wanna port from vendor A to vendor B, you can ask Claude, hey, can you, like, port me?

它就能直接把代码写出来。

And it'll it'll just write the code.

它会自己分析，然后搞定一切。

It'll figure it out and do it.

但当我观察那些最大的企业和顶级SaaS公司时，发现它们往往不止拥有一种优势。

But when I look at kind of the biggest businesses and the biggest SaaS companies, they don't just have one moat.

他们是在经营生意。

Like, they're running businesses.

既然你在做生意，自然希望积累护城河，增强实力，把生意做得扎实。

And if you're in a business, you kind of want to accumulate moats and you want to build strength, and you want you want to, like, build a good business.

他们很少只靠一种护城河，比如切换成本——我觉得这个越来越不重要了。

And very rarely do they just have one moat, like switching costs, which I think matters less.

通常情况下，是切换成本加上网络效应，或者切换成本加上资源垄断。

Usually, it's something like switching costs and network effects or switching costs and a corner to resource.

当你把这些护城河结合起来，力量就大得多。

So when you combine these moats, you get a lot more power.

所以，我会从这家公司的角度这么来看待这个问题。

And so this is the way that I would think about it from this company's point of view.

有些护城河会变得没那么重要，但大多数其实和以前一样强劲。

Some moats will matter less, but actually most of them are still just as powerful as they were before.

现在出现了一种新的说法。

There's this emerging narrative.

我搞不清楚这是认真的观点，还是纯粹的营销话术，但有些公司——比如跟Anthropic这种前沿企业比起来还差一截的——却在跟客户说：你知道吗？你要是用Anthropic，等于把狐狸放进鸡窝了。

I can't tell whether it's serious or a marketing spiel, but some of the companies that I would say are not quite at the frontier the way, say, Anthropic is, have been making this push that saying to customers, you know what, if you use Anthropic, you're letting the fox into the henhouse.

如果你是律所、银行这类机构，用Anthropic的话，它们会把你公司的方方面面都摸得一清二楚，最后甚至能替你做你的生意。

If you're a law firm or a bank or something like that, by using Anthropic, they're going to learn so much about your business and one they'll be able to do your business.

别用Anthropic或者OpenAI了，不如让我们给你定制一个开源模型。

Instead of using Anthropic or OpenAI, let us customize an open source model for you.

这个模型会直接融入你的专属数据，部署在你自己的服务器上，这样东西就是你自己的了，诸如此类。

It will bake in your own data, it'll be hosted on your servers, and then, like, you own it, etcetera.

客户凭什么放心让Claude、让Anthropic深度嵌入他们的业务流程？

Why should customers feel comfortable letting Claude, letting Anthropic be so plugged into their business workflows?

我觉得我得先问问，这话是谁说的，他们的动机是什么。

You know, I would probably ask who who's who's saying this and what are

他们的利益诉求是什么？

their incentives?

就说微软吧，他们的CEO在推特上发了一篇长文，虽然说得挺含糊，但明显就是在暗示这一点。

I'll just say Microsoft, for example, is, like, very the CEO of Microsoft put out a long post on Twitter, and it was a little bit, like, vague, but this was clearly the insinuation that they were pushing.

几周前，亚历克斯·卡普在CNBC的一次采访火了，他其实也在暗示同样的事情。

Then there was an Alex Karp interview on CNBC that went viral a couple weeks ago, and he was basically making the same insinuation.

你这是在犯错。

You're making a mistake.

你把钥匙交给了这些大公司，它们一旦深度接入你的业务，可能干出更多你想不到的事。

You're handing over the keys to these big companies that could potentially do a lot more things if they're like plugged so deeply into your business.

为什么不选一个开源模型，自己部署在云端呢？这样东西就真正归你所有了。

Why not use an open source model that you host on your own cloud and so forth, and then you just own

对。

it.

是的。

Yeah.

所以我觉得最该问的是，这些人在背后有什么动机？

So I think the biggest thing I would just ask is like, what are the incentives of these people talking about?

没错，这正是我想说的。

No, that's what I'm saying.

我刚才说了，这主要是营销手段之类的。

I said it was marketing, etcetera.

但我相信，我们很清楚他们的动机是什么。

But I believe, I'm sure the, we know the incentives are clear.

但如果我是企业方，我觉得这种担忧并不夸张。你拥有这么多能力，还有这么多资金，这听起来确实是个合理的顾虑。

But if I'm a business, that doesn't seem crazy to me that you have all these capabilities, all this capital, That does not seem like a crazy fear.

比如，我不仅要把所有信息都交给 Claude，还要以各种方式给它提供访问权限，至少在很大程度上开放我的基础设施。

It's like, oh, I'm going to not only put all of my information into Claude, I'm going to give it access in various ways, at least to a significant degree to my infrastructure.

然后有一天，Claude说，你知道吗？

Then one day Claude says, you know what?

咱们成立一家律师事务所吧。

Let's spin out a law firm.

再成立一家银行之类的。

We spin out a bank, etcetera.

而且我们手里已经有足够的信息了，关于这些工作流程的，所以其实没必要再卖软件了。

And we know there's enough information that we have about these workflows that we don't have to sell the software anymore.

我们可以提供一种服务，以前人们是通过我们的软件来实现这个功能的。

We can sell the service that people were previously using our software to build.

是的。

Yeah.

我通常会这么想：我们在隐私、安全和保障方面极其严谨。

The the way that I would probably think about it is we take privacy and security and safety extremely seriously.

甚至到了这种地步：当用户在使用Claude Code时遇到问题，作为工程师，我最需要的其实是能看看他们的对话记录，这样我就能搞清楚问题出在哪，然后说，啊，原来bug在这里。

It's actually to the point where when a user has a bug in Claude Code, the most useful thing to me as an engineer that needs to debug it is I'd love to see their conversation so I can see what happened and I can be like, oh, there's the bug.

我们就能直接修复它。

We can just go fix it.

但我根本看不到这些数据。

I cannot see that data.

确实存在这样的情况，

There there is

从客户的角度来看，他们可以确信，自己拥有一个实例或账户，而且能证明，Anthropic的任何人根本无法查看他们的对话内容。

And from the customer's perspective, it is provable that they can have an instance or an account that is provable, that there is no way for anyone at Anthropic to see that conversation?

是的。

Yeah.

我的意思是，这确实是我们一贯的政策。

I mean, is our this is our policy.

我们服务了大量客户。

Like, we power a lot of customers.

我们支撑了众多企业。

We power a lot of businesses.

对我们来说，信任至关重要。

And to us, the trust is very important.

这正是我们做事的方式。

This is just the way that we operate.

不过我得说，我觉得更值得关注的是模型的持续进步。

I I gotta say though, I think the bigger thing that I would think about is model progress continues.

对。

Mhmm.

如果模型还停留在今天的水平，智能不再进步，那确实可能有道理说你得自己掌控基础设施，从商业角度看这也有其合理性。

If models were stuck in the world of today and the intelligence was static and it was not improving, there might be actually some merit to this argument of you want to control your infrastructure, and this might make sense from a business point of view.

如果你愿意承担运行模型的成本，没错。

If you want to pay the cost of running the model Yeah.

还想自己搞定调试推理失败的问题，做这些事可不轻松，而且完全是种非常专业的技能。

And you want to, you know, figure out how to debug when inference doesn't work and kind of do all these things, which by the way is a lot of work, and it's a it's a very niche expertise.

但技术一直在进步。

But progress continues.

所以我认为，对大多数企业来说，紧跟前沿、享受智能提升带来的好处，好处大得多。

And so I think actually for most businesses, there's a really big upside of staying on the frontier and benefiting from that intelligence.

这正是我们在Anthropic内部看到的情况。

And this is what we're seeing internally at Anthropic.

我们的所有客户也都看到了同样的趋势。

This is what all of our customers are seeing.

如果你只需要很小的模型，那用开源的也行。

Maybe if you need just only tiny models, go use an open source model.

也许这样挺好。

Maybe that's great.

但如果你需要前沿的智能模型，而前沿还在不断前进，那我们正好能帮上忙。

But if you need a frontier intelligence model and the frontier continues to move, then we're here to help.

既然乔提到了银行，而你也说你喜欢历史，鲍里斯，我们能聊聊COBOL吗？

Since Joe mentioned banks and since you said you like history, Boris, can we talk about COBOL for a second?

所以，Claude Code 现在能处理COBOL了，对吧？

So Claude Code can do COBOL now, right?

如果我是大型银行，主系统的问题基本上就解决了，我终于可以升级、改进，把我的

So the mainframe issue is basically solve if I'm a large bank, I can finally upgrade and improve and integrate Bring my

70年前的代码库迁移到现代标准。

my 70 year old code base into modern standards.

别搞错了。

Make no mistakes.

确实有不少银行正在用Claude Code来做这种迁移。

There there are actually a lot of banks that are using Claude Code for exactly this kind of migration.

等等。

Wait.

再说详细点。

Say more.

是的。

Yeah.

COBOL 这话题在好多期节目里都出现了。

This is COBOL has come up on so many episodes.

是吗？

Oh, yeah?

对。

Yeah.

我们总听说，如果你是 COBOL 工程师，那确实挺赚钱的。

And we always hear, like, if you're if you're a COBOL engineer Yeah.

像大家说的，在银行里干这行，收入相当可观。

You can make bank at the banks, as they say.

是的。

Yeah.

克劳德在迁移代码方面真的很厉害。

Well, Claude is really good at migrating code.

这其实是核心技能之一，而且这些能力一直在持续提升。

This is one of the actually the skills, like the core skills that's just been improving over time.

对。

Mhmm.

举个例子，我们刚发布了一篇博客，讲的是邦德团队的杰瑞德，你知道，邦德是驱动克劳德代码的JavaScript引擎。

One example, we just published a blog post about how Jared on the the Bund team you know, Bund is the JavaScript engine that powers Claude Code.

他把整个代码库从Zig语言迁移到了Rust语言。

How he migrated the entire code base from one language to another language, from Zig to Rust.

整个过程只花了十一天，真惊人。

And it took about eleven days Wow.

就一个人完成的。

For one person.

他用了四重代码配合动态工作流来完成这个任务。

And he used quad code with dynamic workflows to do this.

以前这得好几位工程师花上一年左右的时间，这种事我们根本不敢想。

In the past, this would have taken, like, a few engineers, like, a year or something, and it's something we never would have done.

哦，我看过那篇文章。

Oh, I saw that piece.

是的。

Yeah.

结果就花了他们大约十五万美元的积分什么的。

And it just cost them, like, $150,000 in credits or something

差不多吧。

like that.

差不多吧。

Like that.

还不到雇工程师成本的一小部分。

Fraction of what engineers would cost.

以前啊，我们根本不可能这么做，因为你得停掉一年的开发才能完成。

And back in the day, like, we just never would have done that because you have to stop development for a year to to do it.

这种成本，哪家公司都负担不起。

It's just like no business could actually pay that cost.

但说实话，当时的经济环境真的在发生变化。

But, yeah, like, the economics were really changing.

过去如果你有个庞大的代码库，停掉开发或者全部迁移到 Java 都不划算，但现在你完全可以这么做。

And so, you know, if in the past you had this big code, well code base, and it wasn't cost effective to stop development, or it wasn't cost effective to just migrate everything to Java, you you can now just do this.

你只要给 Claude Code 下个指令，它就能帮你搞定。

You can just prompt Claude Code, and it can do this for you.

将来编程语言还会重要吗？

Are computer languages gonna be irrelevant in the future?

是的。

Yeah.

我觉得，其实今天编程语言就已经没那么重要了。

You know, I I think they're largely irrelevant today.

这事儿挺有争议的，因为如果你去问不同的工程师，每个人看法都不一样，而且——

And I you know, this is like a this is a spicy thing because if you talk to different engineers, they're gonna have all sorts of views, and

对。

Yeah.

我也不确定哪种看法才是对的。

I don't necessarily know what's the right view.

作为工程师，我总是习惯从利弊角度去思考问题。

You know, as an engineer, I think about everything as kind of pros and cons.

我个人特别痴迷编程语言。

To me, I'm a big languages nerd.

我真的很喜欢编程语言。

I love programming languages.

我也特别喜欢类型系统。

I love type systems.

我甚至还写过一本关于我特别喜欢的一种语言的书。

I actually wrote a book about a language that I really like.

随着大语言模型的发展，我觉得这越来越不重要了，因为模型根本不在乎这些。

Increasingly with LMs, I think it matters less and less because the LM doesn't really care.

某些语言特性确实能带来一点帮助。

There are some things about a language that helps a bit.

如果一种语言效率高、有类型检查，还具备良好的静态分析，那确实能帮模型生成更好的代码。

If the language is really efficient, if it's type checked and it has a good static analysis, then this helps the model generate better code.

但随着模型越来越智能，这一点的重要性反而越来越低。

As the model gets more sophisticated, this actually matters less.

因为就算模型直接写原生汇编，它也可能第一次就写得很棒，而且未来只会越来越强。

Because, you know, even if the model is writing just raw assembly, it can probably just do it really well the first shot, and that'll only get better over time.

你觉得我们会走向一个统一标准代码的世界吗，还是说我们会朝着

Do you think we could move to a world where there's one standardized dominant code, or are we heading in a

另一个方向发展

world

毕竟像Claude Code这样的平台已经能做这么多事了，我们会不会反而看到更多小众语言出现？

because Claude Code and other platforms can do so much of this where we get even more niche languages?

你知道吗，我觉得Claude的出现正在引发一场创新热潮。

You know, I think that with Claude, what is happening is there's an explosion in innovation.

在商业领域，我们正看到各种全新的挑战涌现。

We're seeing this on the business side with all sorts of new struggles.

还是那场Y Combinator的演讲，有个初创公司用Claude来发现新材料。

Again, one of these Y Combinator talks, there was a startup that was using Claude to discover new materials.

材料发现。

Material discovery.

当时他们说——

Were like-

材料科学。

Material science.

对，材料科学。

Material science.

没错。

Yeah.

他们的观点是，当年因为硅材料引发了一场革命。

Their thesis is like there was a revolution because of silicon.

那下一个硅材料会是什么？

What's the next silicon?

我们该怎么找到它？

How do we discover that?

我们该怎么发现这种新材料？

How do we discover that material?

他们正用Claude来寻找答案。

They're using Claude to search for it.

现在商业和产品领域正发生一场变革。

There's this revolution happening in business and in product right now.

我觉得这背后还有很多延伸影响，同样的事情也可能发生在语言和计算领域。

I think there's just a lot of coroaries to this where the same thing might happen to languages and computing.

我甚至能想象，未来会出现语言和计算方式的爆发式创新。

I could see a world where there's just a Cambrian explosion of new languages, of new ways to think about computing.

我想再回到命令行和图形界面这个话题。

I'm gonna go back to this sort of command line versus graphical user interface question.

自从我开始用Claude Code的终端，我就觉得再也不想用网页了，太笨重了。

Once I started using the terminal at Claude Code, I was like, I don't wanna use the web anymore because it feels clunky.

我只想在终端里直接说一句：给Tracy发封邮件，内容是这样，而不是非得打开Gmail，点来点去，特别麻烦。

I wanna just be able to say like, send an email to Tracy saying this in the terminal rather than going to like Gmail and then you click out a button and it just feels very clunky.

还有另一件事，我几年前就注意到了，年轻时用电脑，特别在意文件的管理。

Then there are other things like, and I noticed this years ago, for example, that when I was younger and using computers, really cared about my files.

你看，这是个文件，点一下打开，整个结构是层层嵌套的。

Here's a file and I click on it and I open it and then there's this very hierarchical thing.

但后来搜索功能普及了，这种需求就没那么重要了。

But then when search became a thing, that became less necessary.

你根本不需要把邮件分门别类地放进文件夹里。

It's like you don't need to organize your emails into files.

我直接搜人名或者关键词，就能找到想要的内容。

I just search the name of the person or I search a keyword and I find the files.

我们还需要保留视觉文件系统的位置吗？

Are we still going to have room for visual file systems?

既然打几个字就能看到结果，视觉界面还有什么作用呢？

What is the role of the visual framework when it's just so easy to type something and see the words and get the output right there?

我给你演示一下好吗？

Can I show you an example?

当然可以。

Yeah, sure.

好的。

Okay.

我们截个图吧。

And we'll get a screenshot of this.

这样音频听众也会有兴趣去YouTube上看一看。

So this will be a good will be a reason for the audio listeners to check out the YouTube.

太棒了。

Awesome.

太棒了。

Awesome.

好的。

Okay.

我来给大家演示一下。

So so let me show you guys this.

我们在 Slack 上设了一个反馈渠道。

So this this is we have this feedback channel on Slack.

好的。

Okay.

我发了一条反馈。

And what I did was I posted this feedback.

你们有没有注意到那两个音频图标？我老是分不清它们各自代表什么。

Like, have you guys seen the there's these, like, two audio icons, and I'm always confused which one means what?

哦，对。

Oh, yeah.

这种情况挺多的。

Is Many such cases.

对。

Yeah.

是的。

Yeah.

真的特别让人困惑。

It's just, super confusing.

我就问了，嘿，有人有同感吗？

And I ask, like, hey, like, does anyone agree?

这难道不让人糊涂吗？

Is this confusing?

结果Claude Tag就主动插话进来了。

And so what happens is Claude Tag jumped in to the conversation.

我可没点名问他。

I didn't ask it.

它只是偶然看到这个帖子，就主动插了进来，还回了话。

It just kinda noticed this thread, it jumped in, and it and it responded.

我让它深入查一下，它就找到了人们使用这些按钮的频率数据。

And I asked it to dig in, and it found data about how often people use each of these buttons.

它还整合了两个不同的数据源。

And it created across two data sources.

它同时查看了Datadog和Google BigQuery里的信息。

It looked at both Datadog and Google BigQuery.

所以我把这两边的数据都看了，然后整合成一个挺清晰的回答。

So I looked at both, and then I combined it into this, you know, pretty coherent answer.

它还提出了一些替代方案。

And it suggested some alternatives.

我就跟它说，好啊。

And I asked, okay.

你能帮我设计几个方案吗？

Can you make some designs?