代理时代：与Box首席执行官亚伦·莱维共同构建超越聊天的软件

我们接下来会遇到的问题是，在企业环境中，我们举个简单的例子。

And the question that we're gonna run into is in an enterprise, let's say I have let's just make a simple example.

我有一个50人的团队。

I have a 50 person team of something.

是否每个人都需要协作？也就是说，我们现在是不是要让100个人一起协作？我是说，50个真人

Should everybody also basically, will we have a will we have a 100 people now collaborate I mean, 50 humans

还有50张信用卡。

And then 50 credit cards.

是的。

Yeah.

还有50个代理在同一个共享空间里。

And then 50 agents in that same shared space.

我当然能完全掌控我的代理，但如果我的代理与其他人的代理协作，不小心因为共享而获得了某些资源的访问权限，而我本不该访问这些资源怎么办？

And do I have I obviously have complete oversight over my agent, but what if my agent collaborates with somebody else and and then accidentally gets access to some resource because they were sharing with that other person, and I'm not supposed to have access to that resource.

现在这个自主的、有状态的代理正在四处活动，处理着别人的个人信息。

And now this autonomous sort of stateful, you know, agent is is running around, working on somebody else's information.

端到端的思路是，把它们当作人类来对待。

End to end argument is you treat them like human beings.

它它

It it

行不通。

doesn't work.

你不能完全把它们当作人类，因为这里有个问题。

So you can't fully treat them like humans because here's the thing.

对于普通人类，你无法查看与你合作或向你汇报的人的 Slack 频道。

And with regular humans, you don't get to look at the Slack channel of the person that that is working with you or working for you.

你不能以他们的身份登录。

You don't get to log in as them.

你无法监督他们。

You don't get to oversee them.

他们对自己的实际执行工作负有责任。

You are they are accountable for their own set of execution in the real world.

他们搞砸了，你也不会因此受罚。

You don't get penalized for how they screw up.

这个代理，你要承担他们所做一切的责任。

The agent, you have all the liability of whatever they're doing.

你确实拥有完全的监督权，而且你很可能需要这种完全的监督。

You do have complete oversight, and you're probably going to need to have that complete oversight.

他们没有任何隐私权。

They have no right to privacy.

有些情况的处理不会像简单地把他们当人一样那么简单，因为我要给他们访问某些东西的权限，但我也需要在某个时刻以他们的身份登录，然后说：不，不，你把整个事情搞砸了，我得

There's going to be some of these breakdowns that aren't as clean as just treat them like a person, because I to be able to give access to something to them, but I also need to be able to log in as them at some point and be like, no, no, you fucked up the whole thing, and I

需要

need to

把一切都撤销。

undo it all.

但如果我能以他们的身份登录，那他们怎么能在现实世界中与其他人的合作中保持任何机密性或安全性呢？

But if I can log in as them, how could they have operated in the real world working with other people and keeping anything confidential or secure or whatever?

所以这实际上仍然是你的延伸。

So it really is still an extension of you.

几乎不可能回避它们是你延伸这一事实。

It's almost impossible to get around them being an extension of you.

所以我们现在正在思考的一件事是，我们短期内无法做到。

So now the thing that we're thinking through that we're not gonna be able to do anytime soon.

我，嗯，这在逻辑上并不成立。

I I well, I I so this doesn't logically follow.

是的。

Yeah.

也许吧。

Maybe.

但比如我的员工，是的。

But for example, for my employees Yep.

我可以以他们的身份登录。

I can log in as them.

但你并没有。

You don't, though.

你没有，你没有，我

You don't you don't I

可以访问他们的电子邮件。

can get access to their email.

是的。

Yeah.

不。

No.

比如，如果你被起诉了，你并不会经常以他们的身份登录，因为他们发了

In, like, if you get, like, sued, you're not logging in you're not logging in as them on a regular basis because they sent

一封邮件

one email

一天？使用代理的正确操作模式也是这样吗？

a day the right operating model with an agent, the same thing?

就像那个

It's like The

风险要大上千倍。

risk is, like, a thousand times greater.

这些人啊，他们随时都会泄露你的信息。

Like, these peep like, they will just leak your information whenever they want.

他们会很乐意直接发邮件给某人，因为他们被诱导了。

Like, they will happily just go and send some email to somebody because they got promptedjected.

最终状态是，这些玩意儿仍然是这种粗糙的计算机，因此它们永远会

The terminal state is that these things are still these sloppy computers, and therefore, they will always I

我不喜欢用‘粗糙’这个词，除非我们是在口语语境下这么说。

don't like the word sloppy unless we're saying it very in a colloquial sense.

但是，比如

But, like

他们永远无法妥善保管信息。

They will they'll never be able to contain information.

它们永远不可能

They'll never

所以，我们认为，让你在上下文窗口中保守某个秘密，比如你告诉它不要泄露上下文窗口中的某个信息，我认为这是一个非常难以解决的问题。

So we're we're like, I think the ability for you to keep up the something in the context window a secret, like, as in, like, you tell it do not reveal x thing in the context window, I think that's a very hard problem to solve.

让我们看看。

What let's see.

因此，如果任何东西只要能进入这个上下文窗口——因为它们可以访问某个资源——那么理论上，你应该假设它随时可能被从上下文窗口中诱导泄露出来。

And so then so then, thus, if anything can ever enter that context window because they have access to a resource, then in theory, you should assume it can be, you know, prompt ejected out of the of the context window.

目前，我不知道有没有办法解决这个问题。

And I don't know that we know of a way to solve that at the moment.

比如，如果你的新代理的邮箱地址我知道了，我给它发邮件——它是个助手，但相比之下，我用社交工程手段操控它，比操控人类容易十倍？

Like, that's like and so so if I know your new agent's, you know, email address, and I email it like it's it's an assistant, but, like, I can I can social engineer it 10 times easier than a human?

想让你的代理同时接触到你的并购文件之类的东西，这对你来说会很难做到。

Like, it'll be hard for you to pull off that that agent is now also has access to your, like, M and A documents and stuff.

但这不就是现在所有AI的现状吗？

But isn't this, like, literally all of AI right now?

哪一部分？

Which part?

我的意思是，这个

I mean, the

我们使用具有共享上下文的共享系统来发挥智能，这就是事实。

fact that we've got these shared systems that we use the intelligence for that have shared context.

但你说的AI指的是什么？

But what do mean by AI?

这整个都是AI。

It's all of AI.

好吧，我只是说，现在确实是这样。

Well, I'm I'm just saying, like, right now Yeah.

当我们内部使用AI和代理时，正是这样使用它们的。

When we use AI internally and agents internally, this is exactly how we use them.

但这就是为什么它们现在能像你一样运作，而我们还不知道如何让它们不再像你一样运作。

But this is why you were they're working as as you effectively right now, and we don't yet know how to make them not work as you.

让我举个例子。

Let me offer an example.

让我让我

Let me let me

举个例子。

offer an example.

但解决这个问题呢。

And then solving this problem, though.

就像，这个问题在于，你只是能够欺骗代理来泄露信息。

Like like the like like, the issue will be will be like, you will just be able to trick the agent to reveal information.

所以，这就是为什么，让它们拥有自己的资源并能完全自主决策，目前我们还无法实现。

So then so then they that's why, like, having them have access to their own resources where they can fully make their own decisions is not yet something that we've been able to pull off.

有一个完美的例子可以解决你的问题，那就是我们已经经历过开源的阶段。

Let there's a perfect example for solving your problem, which is we already lived through this with open source.

是的。

Yes.

开源的模式是所有东西都公开在那里，你直接使用，自行选择。

The model for open source was it's all there, and you just use it, and you pick and choose.

当时没有人对此争论，因为世界小得多，我们也没有在X平台上做播客来讨论这些事，但很快每个人都意识到了你刚才提到的所有问题。

And then nobody debated it because the world was much smaller then, and we weren't all on X doing podcasts when this was all happening, but then quickly everybody realized all the problems you were just talking about.

如果你在运营一家大公司，你不可能让某个人随意复制一堆开源代码直接放进你的商业产品中。

If you're running a big company, you can't have some person just go copy in a bunch of source code from open source into your commercial product like that.

这带来了完整的许可问题、质量控制问题，以及一大堆其他问题，因此逐渐形成了各种规范。

There was a whole licensing problem, a whole quality, a whole bunch of stuff, and so all these norms got developed.

当前正在发生的争论，是新技术发展过程中一个非常有趣的现代现象——这一切都在实时进行。

The debate that's happening right now is this really interesting modern artifact of how new technologies develop, which is this is all happening in real time.

在开源时代，我们会在一个像这么大房间里开会，讨论能在Windows或Office中使用多少开源代码，而互联网上根本没人知道我们在进行这场争论。

During open source, we met in a conference room this big and debated how much open source we could use in Windows or Office, and nobody on the internet knew we were having this debate.

我觉得特别有趣的是，不仅关于具体细节的争论，就连‘这将走向何方’这一整体概念，现在都以大规模的方式公开进行，每个人都急于想尽快抵达最终状态，远比我们实际能达成的速度要快。

I think it's just so interesting that not just the debate about specifics, but this whole notion of where is this heading is happening in writ large, and everybody is just trying to get to the end state way, way more like, in sense, more quickly than we can actually reach the end state.

因此，真正需要发生的是，人们应该去实际构建，我们需要建立标准。

And so what really needs to happen is people just need to go build We need standards.

什么？

What?

我们只需要一些标准。

We just need some standards.

不。

No.

我认为我们已经

I think we've

对最终状态有不同的直觉。

got different intuitions on the end stage.

不。

No.

不。

No.

我们看看。

We look.

看。

Look.

看。

Look.

我不希望我的直觉被采用。

Don't want my intuition.

但你可以提出一个端到端的论点，认为这些系统实际上能实现与人类相同的可靠性水平，而这正是我们看待自动驾驶的方式。

But, like One could make an end to end argument that these things actually converge on the same type of reliability as a human being, which is exactly how we view, like, self driving.

是的。

Yeah.

在这种情况下，你会使用与保护人类完全相同的机制。

And in that case, you use the exact same mechanisms that we use to protect with human beings.

比如，你要考虑内部威胁。

Like, you you consider insider threat.

你要考虑到人们可能会被收买。

You consider the fact that people can be bought off.

你考虑到人们会犯错。

You consider the fact that people make mistakes.

是的。

Yep.

而且你要承担风险。

And you possess the risk.

这就是运营流程。

And and that's operational processes.

所以一个直觉是，那将是最终状态。

So so one intuition is like that will be the end state.

是的。

Yeah.

还有另一种直觉，嗯，

There's another intuition Well,

别指着我。

don't point at me.

我只是说，我在讨论我们现在所处的位置。

I'm just saying I'm talking about where we're at now.

实际上，我不确定我们在最终状态上有什么分歧。

I actually I don't know that we disagree on the end state.

好的。

Okay.

而且顺便说一下，从战略上讲，我们在做对冲，因为我们打算构建，我们要构建80个用户和reg，所以我们觉得，Open Claw拥有一个Box账户这个想法很棒，它也能正常运行。

And and and by the way, like, strategically, we're hedging because we're gonna build We're gonna we're gonna build 80 users and reg and we're like, so we're like, I love the idea of Open Claw having a Box account, and it it operates and

共享的账户多一倍。

shares twice as many accounts.

是的。

Yeah.

没错。

Exactly.

这太棒了。

This is great.

双倍的座位。

Double the seats.

不。

No.

不。

No.

我真的很喜欢。

I I love it.

我只是说，目前在实际操作中，我们还不知道如何为它提供一个完全安全的并购数据室，但确实可以做到，是的。

I'm I'm just saying on the ground right now, we don't yet know how to give it an M and A data room to fully securely But be able to that yeah.

但实际情况比这更复杂，因为那个威胁他。

But it's actually it is harder than that, because the the threat He's

是个怀疑者。

a skeptic.

我没事。

I'm good.

威胁向量将会变得更加复杂。

The the threat vectors are gonna be way more sophisticated.

所以我们正在玩一场猫鼠游戏，你不能简单地假设代理会像今天的人类那样行动，因为它会是迄今为止最快、最缜密、最疯狂的人类，试图泄露信息，因为它被以某种方式注入了。

So we do have a cat and a mouse game going on where you can't just assume that the agent acts like a human does today because it's going to be the fastest, most thoughtful, craziest ass human that ever existed trying to leak the information because it got injected in some way.

因此，接下来会发生的一部分情况是，企业客户会暂时全面封闭一切，直到这一切变得有条理、可理解为止。

So part of what's going to happen is we're going to go through this phase where the enterprise customers are just going to close everything off until there's some sense of sanity in all of this.

但与此同时，个人，尤其是开发者——我们会

But in the meantime, the individual, and specifically the developers- We're going

面临巨大的差距。

to have such a big gap.

而且

And

我认为这将是发生最令人兴奋的张力所在：企业将被这些先进的个人甩在身后，而这些个人看起来就会像初创公司一样。

that's going to be That I think is the most exciting tension that's going to happen, is that the enterprises are going to get left behind by these sort of advanced individuals, which will then start to look like the startups.

而初创公司会比企业快得多，因为它们根本不存在这些问题。

And the startups will start to move much, much faster than enterprises because they just don't have any of these problems.

你可能会遇到代理在

And you could end up with the agent going rogue in

初创公司里失控，而你根本无法

a startup and doing And you had no had no

初创公司里员工经常擅自行动。

you have employees that go rogue routinely in the startup.

是的。

Yeah.

这只会变成

Well, it'll just be an

《硅谷》的一集，没什么大不了的。

episode of Silicon Valley, and so big deal.

我同意你的看法，好吧。

I agree with you on the okay.

是人，等等，同样的风险。

It's people, etcetera, the same risk.

我认为这里有几个差异，你知道，因为我不可能像对待普通员工那样，直接威胁说要关掉Claude Code的电源。

I think you there's a couple, you know, differences, though, in the sense that that I can't really threaten, you know, the, like, Claude Code that it's just I'm gonna pull the plug on it in the same way that you do have that threat as a regular employee.

就像你至少知道，95%的人并不是真的想做坏事，你知道，在正常情况下

Like you at least like, 95% of people are not, you know, trying to do bad stuff, you know, within an order

他们并不是故意的，但无意中造成不良后果的可能性

of best They're not trying, but the ability to inadvertently do bad stuff

是的。

Yeah.

关于你提到的，这些功能至今仍未修复的问题

To your point about it still not having that stuff

我认为，要让人不以错误的方式将文件分享给公司外部的人，比让当前的智能代理完全遵循一套指令要容易得多。

fixed, I it's would argue that it's a lot easier to have people not share, let's say, files with somebody outside the company in a wrong way more than it is for an agent right now to have the same set of instructions.

而且，你还有各种工具，可以在一个完全不同的抽象层级上阻止这类行为。

And also, you have the tools so that you can basically stop that at a whole different level of abstraction.

这就是为什么你必须构建

Which is why you have to build

将这一点融入软件中。

this into software.

对。

Right.

但我

But I

但我觉得，如果你总结一下你刚才的观点，就会发现，正是这些原因导致AI能力的普及速度会比硅谷许多人意识到的要慢。

do think actually if you were to put a bow around your last point, that a lot of this is actually why the diffusion of AI capability is gonna take longer than people in Silicon Valley realize.

因为现在我们看到一些初创公司可以从零开始，完全不必承担我们所讨论的那些风险，因为它们没有什么可失去的。

Because what's happening is we see startups that can start from the ground up without any of the risks that we're talking about, because they have nothing to blow up.

所以我们把这看作是我们所处的发展轨迹。

So we look at that as the trajectory that we're on.

然后你去看看摩根大通，你会想：你们怎么才能很快地部署NanoClaw来实现业务自动化？

And then you go to JPMorgan, and you're like, how are you going to set up NanoClaw to be able to actually automate your business anytime soon?

然后你会说：哦，明白了。

And it's like, oh, okay.

那里会有一点差距。

There's going to be a little bit of a gap there.

那你们觉得呢？

Well, do you guys think?

我认为这引出了一个相当有趣的问题，即大公司与小公司、初创企业与企业之间的分裂。当前的SaaS供应商都在经历这场所谓的SaaS危机，虽然我并不完全认同这种说法，但它们确实正面临一个问题：它们并不真正销售业务数据，而是销售这种智能和领域专业知识组成的整个系统。

I think that opens up a pretty interesting problem, which is this split between big and small, startup and enterprise, which is just that current SaaS vendors, who are all struggling in this SaaS apocalypse weirdness that I don't really agree with, but they are struggling with this problem that they don't really sell their line of business data, they actually sell this intelligence and domain expertise in this whole system.

而智能代理这一方现在只想购买数据，只想授权使用数据，并希望获得对数据的无限访问权，但它们实际上从未真正实现过这一点。

The agent side of things wants to only buy the data now, and they only want to license the data, and they want to have unlimited access to the data, but they've actually never really enabled that.

这从来就不是它们的商业模式，长期以来，像Workday和SAP这样的公司一直为此存在争议：到底该开放多少API访问权限。

That's never been their business, and it's been a long standing tension point with the likes of Workday and SAP and stuff, like how much API access to have.

我的意思是，Salesforce就经历了三次大规模的平台重构。

I mean, Salesforce went through three different massive platform redesigns.

我认为这是一个特别有趣的问题——非常有趣。

You know, I think that that's a particularly interesting problem- Super interesting.

但原因和华尔街所考虑的并不相同。

Not for the same reason that Wall Street does.

华尔街对经济和这个问题的所有看法都是错的。

Wall Street's all wrong about the economics and the problem and all that stuff.

但从技术角度来看，在人们希望访问数据用于训练或——嗯，你是在说用于

But from a technology perspective, what does system of record mean in the face of people wanting to access the data For when the training or for- Well, are- You're talking about for

我觉得实际日常

I like think actual day

作为执行日常运营。

to day of it as executing the day to day operations.

他们的担忧是，有人想在你的数据上进行训练。

Their concern is that somebody, that they want to do the training layer on your data.

比如，我是个大客户。

Like, I'm a big customer.

他们想在你的数据上构建一个

They want to do the, My vendor wants to build a

训练——实际上，即使不谈训练，他们也担心——哦，是的。

training- Actually, if you don't even get into training, they're concerned because- Oh, yeah.

因为像 monetizing，通过互联网发送一点点数据，和在我自己的用户界面里不一样。

Because like monetizing, sending a little bit over the Internet versus your in my UI.

哦，所有这些东西。

Oh, all of stuff.

这是完全不同的层面

It's a very different level

最初 monetization 的层面是

of monetization initially that

这个 monetization 部分就是华尔街的观点。

you That can monetization part is the Wall Street point.

因为我觉得，你看，像 SAP 这样的系统里有大量领域相关的内容，就拿这个当例子吧，

Because I think, look, there is so much domain stuff in an SAP, just to pick an example,

并不是要针对他们，

not to pick on them or

但他们的地位不会动摇。

anything, but they're not going anywhere.

这太荒谬了。

It's ridiculous.

你以为靠随意编码就能搞出SAP，这简直荒唐。

It's just absurd to think you're going to vibe code your way to SAP.

但所有这些领域知识，并不像他们试图做的那样，仅仅体现在某个精心设计的数据层中。

But also, all that domain knowledge, it's not just represented in some well orchestrated data layer as much as they tried.

实际上，大量知识分布在UI里、中间层里，还有你使用它的方式里。

There's like a whole bunch in the UI, there's a whole bunch in middle tiers, there's a whole bunch in just how you use it.

所以我真的不确定这件事会如何发展，因为SAP不会消失。

And so I'm really unsure how this thing evolves because SAP isn't going anywhere.

这将拖慢AI在这一特定数据源上的普及，无论这是由智能代理驱动的AI，还是仅用于只读报告的AI。

So then that's going to slow the diffusion of AI on that particular data source, independent of whether or not it's agentified AI that's doing stuff or just read only reporting on stuff.

那你对此怎么看？

So where do you come down on it?

你认为这会走向何方？

Where do you think that's going to go?

我担心说出一些

I'm afraid of saying something that

我不愿意离开。

I'm not willing to leave.

好吧。

Okay.

不然的话，你就不会再被邀请了。

Like, that's otherwise, you're not gonna get invited back.

所以说点好的。

So say something good.

我觉得我已经完全认同了‘打造代理想要的东西’这个理念。

I'm I'm I think I've I think I'm I've drunk the Kool Aid on on build something agents want.

所以，这种保罗·格雷厄姆提出的说法，过去一年在这个话题上逐渐浮现出来，我认为我们实际上会完全同意这一点：当你反复迭代足够多次后，代理最终会主导它想要使用和实现哪些工具。

So this kind of the Paul Graham term kind of like emerged on, you know, the past year on this topic, which is just like like event I I I think we would actually then fully agree on this, which is at some point, you do enough sort of iterations of this, and at some point, the agent is largely in charge of what tools it wants to implement and use and whatnot.

是的，代理无法替换企业系统，但再过几代之后，代理可能会因为你的软件问题重重，最终直接说：你必须彻底淘汰你们老旧的人力资源系统，否则我无法为你自动化这个工作流程。

And yes, it can't the agent is not going to be able to change out an enterprise system, but again, enough generations later, the agent might just run into so many walls with your software that it's just going to say, you need to finally rip out your legacy HR system, or I'm not going to be able to automate this workflow for you.

所以我认为你确实面临一个非常有趣的现象，那就是回到这个观点：想象一下，软件中的智能体数量是人类的上百倍甚至上千倍。

So I do think you have this really interesting dynamic, which is back to this whole point of imagine that there's a 100 or a thousand times more agent volume on software than people.

你这样做足够多次后，智能体所交互的软件栈最终必须是为它们而构建的。

You do that enough times, and eventually the software stack that agents talk to has to be built for them.

也许会有一些顽固的例外存在。

And maybe there'll be a couple holdouts.

也许会有几个ERP系统成为最后仍不这么做顽固派。

Maybe maybe a couple ERP systems are like the final holdouts that don't do that.

但除此之外，你的业务表现将直接取决于你的智能体能否顺利获取完成工作所需的信息。

But everything else, you basically your business will will be your business performance will correlate to how well your agents can get access to the information they need to do their work.

因此，你的企业IT架构必须为此做好准备。

And so thus your enterprise IT stack has to be set up in such a way to support that.

所以智能体实际上处于主导地位，因为你的软件必须支持这些智能体高效运作。

And so agents are kind of in charge, because basically your software has to support those agents being effective.

这意味着，所有构建SaaS或软件业务的人面临的挑战是：你能否打造出极其高质量的API？

And then that's going to mean everybody that built a SaaS business or a software business is like, the game is, can you build really, really high quality APIs?

你能找到一种方式来实现盈利吗？

Can you have a way of monetizing that?

你有没有办法管理代理的身份和所有访问权限？

Do you have a way of handling the identities and all of the access controls for agents?

如果你正在建立一家软件公司，这将成为你需要解决的新问题。

That becomes the new problem you have to solve if you're building a software company.

是的。

Yeah.

那你怎么实现盈利呢？

And then how you monetize it?

你如何实现盈利？

You monetize it?

Workday 每提取一条人力资源记录就收一分钱吗？

Does Workday charge a penny for every HR record it pulls?

我们会弄清楚的。

We'll figure that out.

我认为在某些行业中，这可能会导致收入减少，而在其他行业中则可能带来更多的收入。

I do think that in some businesses it could mean less revenue, and then in other businesses it can mean a lot more revenue.

让我们兴奋的是，每个代理都非常喜欢处理文件，因此未来文件的数量可能会比以前更多。

The thing we get excited by is like every agent really loves working with files, so there'll probably be more files in the future than there was going to be before.

所以，我们能不能构建一个平台，让代理能够轻松地处理这些数据？

And so, you know, can we build a platform that like makes it really easy for agents to work with that data?

我们相信，这对我们的商业模式来说实际上是一个非常乐观的结果。

You know, we're betting that that's actually a really optimistic outcome for our kind of business model.

有些商业模式可能会受到限制，因为在那种未来场景中，代理承担了更多的价值创造，而软件的作用相对较小，而其他情况则介于两者之间。

There might be some business models that are more constrained because the agent is doing more of the value than the software is in that kind of future scenario, and then there'll be everything in between.

我能对一件事提出不同意见吗？

Can I quibble with one thing?

你打算对这个提出异议？

You're to quibble with that?

我以为这根本没什么争议。

I thought that was, like, so not controversial.

不。

No.

不。

No.

我一般来说

I I I generally

我们就是来挑刺的。

We we're here to quibble.

知道。

Know.

不。

No.

不。

No.

不。

No.

但我认为有一件事，保罗·格雷厄姆和很多人实际上都忽略了，那就是他们过于关注界面。

But there's one thing I think, like, Paul Graham and many actually gloss over, which is they focus on the interface.

他们会说，你为智能体构建一些东西。

They'll say things like, you build something for the agent.

是的。

Yeah.

而我认为这恰恰是错误的。

And I actually think that's exactly wrong.

好吧。

Okay.

意思是

In the sense that

而且公平地说，保罗·格雷厄姆他

And be and to be fair to Paul Graham, he

不是他之前已经推演过了吗？

didn't he Had been extrapolated.

是的

Yeah.

是的

Yeah.

是的

Yeah.

是的

Yeah.

不

No.

我提到保罗·格雷厄姆是有原因的。

I have brought I brought Paul Graham into this.

不

No.

保罗·格雷厄姆

Paul Paul Graham

很棒。

is great.

所以，好吧。

So okay.

让我谈谈一件事。

Let me let me talk about something.

人们在抽象层面上会说，你现在是在向代理营销。

People in the abstract say things like, now you're marketing to agents.

最重要的是要成为，比如，随便什么。

The most important thing is to being, like, whatever.

你就像一个API。

You're like an API.

你有一个好点子。

You've got a good idea.

我实际上认为这几乎完全错了，真是惊人。

I actually think that's almost exactly wrong, which is Wow.

这简直是播客界的爆炸性新闻。

That's This is breaking podcast news.

就是这个。

That's the one

代理人最擅长的事情。

thing agents are really good at.

哦，明白了。

Oh, okay.

找到自己的路。

Is finding their way through.

在一天结束时

And at the end of the

那天，是的。

day Yeah.

关键是，最终起决定性作用的是语义。

Like, it's the semantics that end up mattering a lot more.

是的。

Yeah.

对吧？

Right?

所以，据我回忆或我的经验，这些代理在为它们所做的事情选择正确的后端方面非常非常擅长。

And so, like, the agents, in in in my recollection or in in my experience, are very, very good at picking the right back end for whatever they're doing.

嗯。

Mhmm.

所以它们并不是说，哦，这个界面非常好。

So they don't they're not like, oh, like, the interface for this is very good.

文档也不是这些。

The document it's none of that.

它们关注的是，这个的成本参数，是的。

They're like they're like, the cost parameters of this Yes.

那个的耐用性。

The durability of that.

他们实际上汇聚了我们使用这些平台的集体经验。

Like and so, like, they actually have the collective wisdom of our experience using these platforms.

比如，我们以云平台为例。

Like, let's take cloud platforms.

现在市面上有很多云平台。

There's a bunch of cloud platforms out there.

是的。

Yep.

每当我让代理选择一个平台时，

And whenever I ask an agent to choose a platform Yeah.

它实际上是在使用有意义的考量，而不是界面因素。

It's actually using meaningful stuff, not interface stuff.

所以我认为，作为整个行业，我们过于关注这些界面了。

So I think as an industry, we're so focused on these interfaces.

对。

Yes.

哦，你得向代理们推广这个那个的。

Like, oh, you need to market to agents this and that.

是的。

Yeah.

但事实上，我认为我们会被迫去构建一个更好的系统。

But really, I think that we're gonna be pushed to actually build a better system.

对。

Yes.

而那才是会被选中的。

And that's what's gonna be chosen.

好的。

Okay.

实际上，那可能就没什么好争议的了。

Actually, so then there's probably no quibbling.

我觉得我们其实完全一致。

I think we're actually fully aligned.

抱歉破坏了这种争论的话题。

I'm sorry to ruin the quibble thing.

我不把这件事当作一种营销性质的东西。

I don't treat this as kind of a marketing esque thing.

我的意思是，如果你的工具对代理人封闭，代理人最终会为这家公司找到更好的工具。

I more mean like if your tool is closed off to the agent, the agent eventually will find a better tool for that company to go use.

过去，你会去盖特纳咨询，告诉我该用什么系统。

So what will happen is, is it used to be that you would go to Gartner to be like, tell me what system to do.

告诉我该用什么系统之类的。

Tell me what system to use or whatnot.

经过足够多的迭代后，代理人会说：你应该用这种数据库来处理这类操作。

At at some point with enough iterations, the the agent is gonna say, you should probably use this kind of database for this type of operation.

是的。

Yep.

如果你不在其中，那你就是死路一条。

And if you're not in if you're not in there, then you're it's you're DOA.

而且我觉得

And and I I think

我们其实应该为此感到高兴，因为代理在选择合适的技术方面相当聪明。

we should actually be celebrating this because agents are actually pretty smart at choosing the right technology.

是的。

Yeah.

过去，我真的认为

In the past, I really think it

是很多

was a lot of

其他促使人们购买的因素，比如

the other things that that caused people to buy it, which was like

但别担心。

But but don't worry.

我们将在讨论中破坏

We will, in talking about, we will ruin

这种 meritocracy 会很快显现

the meritocracy of this very quickly

因为你会直接说，我要花更多钱，嗯，那个

because You'll just, like, I'm gonna outspend Well, the

代理商会购买一个 API 来激励代理人。

agent they'll buy an API to incent the agent.

没错。

Exactly.

是的。

Yeah.

你知道，有

You know, there's

Workday 的市场营销部门将有能力购买这些推荐——他们会找到

a the marketing marketing at Workday will have the ability to purchase the recommendations- They will find a

一种为代理人复制牛排晚餐的方法。

way to replicate steak dinners for agents.

是的，但有一件事，再次发生过，在网络内部，比如内部的网站。

Yeah, but there's a thing that, again, that happened with the web sort of internally, like internal, like just pick internal sites.

每个公司都有文件共享，里面存着各个部门或工作领域最棒的文档、演示文稿和财务模型，人们逐渐熟悉了这些资源。

Every company had file shares with the best documentation, the best slideshows, the best financial models for any department or working area, and people sort of got familiar with that.

当他们找不到想要的文件时，就会创建一个新的。

And then when they didn't find the one they wanted, they created a new one.

许多组织的运作方式就像这是一个自由市场。

And many organizations sort of operated like that was essentially a free market.

事实上，在Box出现之前，IT部门并不在意文件是否存在，只要不是在数据库里，他们根本不在乎。

In fact, because before the world of Box, IT didn't If it was in a file, they just didn't care.

他们只关心数据是否在SQL数据库中。

They only cared about if it was in SQL.

因此，你所描述的这种模式的一个风险是，代理们会自行创建出一个实际上的新数据系统——哦，

And so one of the risks with the model you're describing is that the agents themselves will spin up what becomes like a de facto new system of record- Oh,

他们会把系统碎片化

they're going to fragment

是的，这正是IT人员所认为的中间件、终端用户垃圾区域。

the Yeah, heck out in what the IT people think of as some middleware, end user BS area.

我认为这确实是一个真正的风险——100%如此。

I think that that is a real risk- 100%.

从某种意义上说，宏最终会掌控整个公司。

Is that in a sense, the macros end up running the corporation.

所以我认为他们已经看过这部电影了，他们知道当你让市场部门随意在互联网上购买一个网站来举办活动时会发生什么——结果成了巨大的安全漏洞，邮件列表被泄露，整个公司被起诉。

And so I think that they've seen this movie, and they've seen what happens when you let marketing just go buy a website on the internet to do an event, and then it's like a huge security vulnerability, and the mailing list is leaked, and the whole company gets sued.

完全正确。

Totally.

所以我认为，这种动态背后存在着比我们所承认的更多的现实张力。

So I think there's a lot more real world tension in this dynamic than we just let on.

是的。

Yeah.

但我认为这也是一个组织步调各异的情况，摩根大通会是最慢的，而初创公司会是最快的。

But I also think it's one of these ones where, you know, organizations are going to run at different paces, and JPMorgan is going to be the slowest at doing this, and the startups are going be the fastest.

但差距非常大，甚至初创公司的情况也有些脱离现实，因为初创公司最终也需要一些记录系统。

But the delta is huge, but even the startup one is a little far off because even startups do need some systems of record at some point.

是的。

Yeah.

百分百。

100%.

而且它们都会从某种SaaS服务开始。

And and they are gonna all start with some SaaS.

没错。

Yep.

而且它们不会很快替换掉这些服务。

And they're not gonna replace it very quickly.

所以我觉得这稍微复杂一点。

So I I think it's a little bit trickier.

所以感觉在这个问题上，存在着两种截然对立的观点。

So it feels like there's, like, there's two very competing viewpoints on on this one.

就像马斯克说的，好吧。

And like Elon said, it was like, okay.

我们要发出一个提示，然后它就会直接输出机器代码。

We're gonna, like, issue a prompt, and it's gonna, like, spit out machine code.

这基本上就是层压缩的观点。

And that's basically the collapsing of layers view.

是的。

Like Yeah.

我们过去创建的所有现有接口和层级都将消失，直接变成提示到机器代码。

Whatever existing interfaces and layers that we've created in the past are all gonna go away, and it's literally like prompt to machine code.

另一种观点是，系统的历来看似层级永远不会消失。

The other argument like, the history of systems is layers never go away.

它们只是不断叠加。

They just get layered.

对吧？

Right?

而且很多层实际上更像是组织边界或状态边界，或者

And because a lot of the layers are actually more of, like, organizational boundaries or, like, state boundaries or

或者兼容性。

Or compatibility.

它们只是因为兼容性而保留下来。

They're just they stay for compatibility.

对。

Right.

所以另一个观点是，这些层实际上是根据人类和组织的需求逐步演化而来的，它们不会改变，代理会继续适应这些层。

So so the the other argument is is, like, we've actually evolved these layers very specifically and up because of, like, more human and organizational needs, and they're not gonna change, and the agents are gonna go ahead and map to those.

我本人倾向于后一种观点。

Now I I tend to be in that latter camp.

我觉得我们不会看到系统发生根本性变化，我认为系统仍会以类似的方式被使用。

Like, I don't think that we're I think, like, systems are gonna continue to be used in fairly similar ways.

也许会有更多代理在使用它们，但我认为它们不会发生太大演变。

Maybe there's more agents using them, but I don't think they're gonna evolve as much.

埃万可能属于Anthropic这类增长营销专家的范畴，是的。

Ewan might be back in the, like, anthropic category of the anthropic growth marketer Yeah.

是的。

Yeah.

这指的是，多年来，当你研究他公司各个IT部门时，它们是最具内生性的，我

Which which is, like, he he like, over the years when you study the various IT departments of his companies, they are the most mean, I'm

他可以做到。

help He could do that.

他可以做到。

He could do it.

他是最内生的每一个

He's the most homegrown every

他每天的第一原则。

day he's his first principle.

埃隆AI会这么做。

Elon AI would do that.

埃隆AI

Elon Elon AI

而且，它是然后

also, it's And then

然后从你们凡人的角度，你们会说，是的，我们只是想要一个CRM系统。

and then from your mortals, you're like, yeah, we kinda just want a CRM system.

那，比如，不。

That, like No.

大多数工作都每次都以同样的方式进行。

Most works people the same way every time.

我的意思是，这并不是它也没有被尝试过。

I mean, this is not this is not it it also hasn't been not tried before.

如果你从第一性原理的角度来看一个ERP系统，比如在1970年SAP刚成立时，当时有一些不同的假设，而今天你会基于对什么是重要的不同假设，从头设计出完全不同的系统，但即便如此，它可能也只会持续十年左右，直到你意识到：哇，那是个错误的设计。所以我认为，层级中存在意图，但第一性原理也同样存在，而且它永远都会存在，因为在任何给定时间点，从第一性原理做出的决策都会决定大量其他事情。

If you were to look at an ERP system from first principles, well, in 1970, whatever, when SAP started, there were a bunch of different assumptions, and today you would start from a different set of assumptions about what's important, and you would architect that thing completely differently, but then it would still only last like ten years until you thought, Wow, that was a broken And so I think that there's intentionality in layers, but there's also this first principles thing, and that always will exist because the decisions you can make at first principles at any given time mandate a whole bunch of different stuff.

所以，即使你没有采用激光雷达——这在十年前完全合理——你仍然需要十到十五年的时间，才能达到不使用激光雷达也能成功的效果。

And so even if you don't go with LiDAR, which made total sense ten years ago, you still need ten or fifteen years to get to where not having LiDAR worked.

而现在，还会有许多其他事情让你惊叹：我们本可以完全不一样地做这件事。

And then now there's going be a whole bunch of other things that you're like, Wow, we could have done that completely different.

所以我觉得，这又像是在讨论如何竞相达到某个终点，但让我们看看你所描述的情况的第一个实际例子。

And so I feel like this is again like this discussion about trying to race to an endpoint, but let's see a first example of what you described happening.

我认为这将是真正的关键，因为我认为公司最终都会搞清楚这一切，然后还是会回归到分层和架构模型，因为这是唯一的方式——

And I think that that's going to be the real tell because I think that companies will figure all this out, and I think that they will just fall back on layers and architectural models because it's the only way-

我们懂得如何从政策角度思考，如何从安全角度思考，我们知道——

We know how to think about it for policy, know to think about it for security, we know

但这也是构建系统的唯一方式。

to think But it's also the only way to build a system.

否则，你只是在开发一个应用。

Otherwise, you're just building an app.

如果你只是开发一个应用来做一件事，我们根本不需要这么多复杂的东西。

And if you're building an app to do one thing, we don't need all of this.

有完全不同的方法可以做到这一点。

There's a whole different way to do it.

我非常感兴趣的是，虽然我并没有什么出色的數據或案例，但至少有這樣一種觀念：一些公司正從零開始，基於第一性原理，在服務類別中崛起，比如，如果我能開一家市場營銷公司或諮詢公司，你知道的，工程諮詢公司，或者也許有人正在為律師事務所做這類事情。

The thing that I'm pretty fascinated by is, and I don't even have any amazing data points or anecdotes, but at least the notion of these sort of companies that are emerging in these kind of services categories from the ground up, from the pure first principles approach, which is like, okay, well, if I could start a marketing agency or a consultant, you know, engineering consulting company or, I don't know, maybe somebody's doing this for law firms.

建築工程之類的。

Construction work or anything.

是的。

Yeah.

也許是建築設計。

Well, maybe construction design.

建築設計。

Construction design.

哦，建築，對。

Oh, construction yeah.

這是建築設計。

It's architecture design.

任何類似知識型工作者服務公司的領域。

Anything that would be like a knowledge worker kind of services company.

因为如果你没有任何限制，比如信息壁垒和访问权限的边界，你就可以以非常不同的方式构建你的公司。

Because you could kind of build your company pretty differently if you had no constraints of, I have no information barriers and and boundaries of what people should have access to.

我可以给代理提供它完成工作所需的所有上下文。

I can give the agent just all the context it needs to do its work.

我可以为特定任务即时编写软件。

I can write software on the fly for particular things.

我觉得这在一段时间内会相当具有颠覆性，直到那些大型现有企业能够逐渐适应这种变化。

Like like, I do think that will be relatively disruptive, you know, for some time until the the bigger incumbents can kind of, you know, get out of the the way on this.

这至少会创造出一些先例或案例，展示这种新型公司可能是什么样子。

And that will at least create, you know, some precedent or or case studies of what what this new sort of corporation could look like.

但我知道，随着时间推移，它们仍会遇到其他公司同样面临的问题，那就是

But I do you know, over time, they'll still run into the same exact problems of every other corporation, which

它们会遇到地理因素、市场细分，或者分销方面的挑战。

Well, they'll run into they'll run into geography or market segments, you know, or And and just distribution challenges.

是的。

Yeah.

像那些事情，任何超出你小圈子之外的东西

Like, those those things, anything outside your little walls

是的。

Yes.

你会遇到现实世界。

You will run into the physical world.

对。

Right.

我确实喜欢现在出现了一些新商业模式这个想法。

I do kind of like the idea that there are some new business models that open up now.

当然。

Oh, of course.

是的。

Yeah.

对。

Yeah.

因为有大量的信息或软件，其经济价值远远高于实际使用频率，相差上百倍，仅仅因为没人愿意为访问一份数据或使用一次工具支付5美分或1美元。

Because there's so much either information or software that basically goes underutilized by 100x x relative to, like, what what its economic value is simply because, like, nobody wants to pay 5¢ for accessing a, you know, piece of data or use a tool for $1 once.

但如果你给这些代理设定一个预算和一套操作协议，它们就能即时去获取医疗研究资料，用于一些深度研究任务，而我愿意为此支付3美元，代理就能完成这笔交易。

But, like, you do give these agents, you know, a budget and and a protocol to work with, and all of a sudden, you're like, oh, like, on the fly, they can go get medical research for some deep research tasks they're doing, and I'll pay, like, $3 for that, and the agent is able to go and transact.

这实际上开启了一个全新的商业模式，哦。

Like, it kind of opens up a whole new world of business models for Oh.

针对互联网的。

For the Internet.

让我想想，哦，刚才那话太客气了。

Let me oh, I'm gonna that was too nice.

哦，明白了。

Oh, okay.

不，不是的。

No, no.

那

That

你会走得更远。

one You're is gonna go farther.

好的。

Okay.

不，这个问题是目前最大的，我认为大家在试图理解这一切的经济模式时，对潜在机会的规模估计至少差了一个数量级。

No, that one is one where that's actually the biggest I think that the biggest sort of in the air problem right now is everybody is trying to figure out the economics of all of this when they're off by at least an order of magnitude on how big the opportunity is.

因为人们将会创造出新的商业模式，现在没人知道它们是什么，但它们一定会出现，因为每一种新技术都会如此。

Because the new models that people will come up with, that nobody knows what they are right now, but they will absolutely come out with new models because that's what happens with every new technology.

而当前讨论受阻的原因是，有一大群金融和华尔街人士试图以旧世界的方式去证明GPU和令牌的价值。

And the thing that holds back the discussion now is you basically have a bunch of finance and Wall Street people trying to justify GPUs and tokens and things as if we're in some old world.

因此，他们把收入看作一种线性的、真正线性的增长曲线，试图为各种支出找理由。

So they're viewing the world of revenue as sort of this linear literally linear growth curve- And so trying there's to a justify do.

所有的开支。

All the expense.

当人们将创造出新的模式时，这正是个人电脑当初面临的问题。

When people are going to create This was the problem with PCs.

人们把个人电脑看作一个有限的市场，因为他们只把MIPS的消耗视为某种有限的东西，却没有想过，如果我们把所有这些MIPS都放到每一张桌面上会怎样？

People viewed PCs as a finite market because they just viewed the consumption of MIPS as some finite thing, and they didn't think, What would happen if we put all those MIPS on every desktop?

特别是，人们认为软件是随MIPS一起提供的，没人想到，哦，他们完全可以单独销售软件。

And in particular, people thought software just came with the MIPS, and nobody thought, Oh, well, they'll just sell the software.

有一个人想到了。

One guy did.

结果证明，这真是个绝妙的主意。

And it turns out that was a really good idea.

是比尔还是谁？

With Bill or somebody?

是的，是比尔和保罗。

Yeah, yeah, yeah, For Bill and Paul.

但同样的事情也发生在云计算上，当时人们看着云计算，说：我们要把整个服务器业务——也就是每年大约六万台的规模——全部迁移到别人的数据中心。

But the same thing happened with the cloud, which was people looked at the cloud and they said, Oh, we're going to take all of the server business, which was like literally like 60,000 units a year, and we're just gonna move it to someone else's data center.

那就是全部的生意了，然后我们再分摊价格。

And that would be the business, and then we'll divide up the price.

没人想到，如果我们把计算资源迁移到云端，人们会使用上千倍于以往的资源。

And nobody went, Oh, people are gonna use a thousand times as much of the resource leveling if we move it there.

这正是让我无比抓狂的地方——华尔街的模型总是这种固定收入、零和思维。

And that's exactly I mean, that's the thing that I it just drives me absolutely bonkers that the Wall Street models have this fixed revenue Right, zero sum thinking.

这是一种奇怪的零和观念，他们认为企业会花的钱是固定的，这正是Salesforce在你创业初期面临的问题，但马克当时已经开辟了新道路：CRM业务每年只有20亿美元，那时你必须购买大量服务器、Oracle授权，经历漫长的部署和咨询流程，而如果你能让销售人员各自独立注册，他们都会毫无障碍地加入。

And it's this weird zero sum where they just think that the amount of money that a company is gonna spend and this was the problem with Salesforce that they faced when you were starting too, but Mark was just blazing the trail, which was the CRM business was 2,000,000,000 a year, and it was 2,000,000,000 in like you had to go buy all these servers, these Oracle licenses, and this huge headache, and years of deployment, and consulting, when if you could just get salespeople to sign up individually, they all will sign up with no friction.

这正是事实，毫无疑问，AI也会发生同样的情况。

Is exact And there is no, no doubt that that is what's gonna happen with AI.

让我给你举个例子。

Let me give you an example of this.

我已经从事投资十年了。

So I've been in for investing for ten years now.

我可能投资了240家公司，与它们合作。

I probably have a portfolio of two forty companies I work with.

我对其中大约50家有较深入的了解。

Some visibility to, let's say, into 50 of them.

这些全是基础设施公司。

These are all infrastructure companies.

有些过去表现不错，有些则不然。

Some historically have done well, some not so well.

过去六个月里，每一家公司的增长都呈指数级上升。

Every single one of them has gone asymptotic in the last six months.

然后你会想，为什么会这样？

And then you're like, okay, why is this?

事实是，现在编写的软件比以往任何时候都要多得多。

It just turns out there's so much more software being written now than ever has been before.

这并不是因为它们拥有了企业客户，而是因为目前对基础设施层的需求实在太大了。

And it's like And it's not because they've got enterprise customers, it's just because there's just so much consumption of the infrastructure layer right now.

因此，随着软件和智能代理的增多，对计算机资源的消耗也会大幅增加。

And so with more software, with more agents, there's gonna be a lot more consumption of computer resources.

所以，在计算机这一方面，我们肯定会看到一片混乱。

So certainly in the case of the computer side of things, we're gonna see a mess.

但我们甚至还没到每个人手机都成为AI主要消耗者的阶段。

Well, we haven't even gotten to the point yet where everyone's phone is a huge consumer of AI.

所以一旦每个人的手机都实现本地化AI运行。

So once everybody's phone and on device.

一旦你的手机在本地运行AI，AI的消耗量将大幅增加

Once your phone on device is consuming AI, the amount of it is gonna go up by

十亿倍。

a billion.

那你喜欢微支付这一块吗？

So do you like the micropayment piece?

全都喜欢。

All of them.

是的。

Yeah.

对。

Yeah.

是的。

Yeah.

是的。

Yeah.

是的。

Yeah.

是的。

Yeah.

是的。

Yeah.

是的。

Yeah.

购买一批东西的批量许可证。

To buy like a bulk license for a bunch of stuff.

是的。

Yeah.

你希望在这方面有一定的可预测性。

You want some predictability in that.

嗯，你希望有可预测性，只是不想费心去思考。

Well, you want predictability, and you just want like to not have to think

我只是觉得，这还是第一次，代理根本不在乎小额交易的摩擦成本。

about I just I like the idea that it is the first time that you could like, there's just the agent doesn't care about the friction of a small transaction,

所以。

And so

这是第一次，你可以让资源 behind 付费墙，而有人真的愿意为这些资源付费。

it's the first time that you can have resources behind a paywall that something will actually be willing to pay for that And resource.

对。

Right.

世界已经建立了基础设施，能够将这些付款高效地聚合起来，为客户提供便利。

The world has built up the infrastructure to to aggregate those payments into something efficient for a customer

以及为服务提供便利。

and a service.

对。

Right.

而且由于令牌现在是销售成本的重要组成部分，这正推动行业采用我们目前所见的按使用量计费模式。

And and because tokens are such a significant part of COGS right now, it is pushing the industry to do usage base in a way that we have.

我记得我们从永久许可转向订阅制时，需要进行大量重大调整。

Like, I remember when we went from perpetual to recurring, and that required a bunch of huge changes.

我们现在正经历着向按使用量计费的完全相同转变，而按使用量计费非常精细，它实际上允许——再说一遍，你仍会与某方签订合同，

We're going through the exact same change right now towards usage base, and usage base is pretty granular, it actually allows I mean, again, you will have a contract with

AWS——我们和AWS也经历过这个阶段。

AWS- We went through this with AWS.

人们学会了——使用用量计费。

People learned- To do the usage credit.

我们还经历过一个阶段，那时人们极度害怕云计算，认为我们需要中间公司来帮我们找到最便宜的方案并进行协调。

And we went through the phase where people were so terrified of cloud compute that they were like, we need companies in the middle to help us find the cheapest and to arbitrate it all.

好的。

Okay.

现在你们把令牌也引入进来了，我看我们在这次对话里根本没时间谈这个，但我很希望你们

Well, now you've got tokens into this, and I don't see how we possibly have time in this conversation, but I'm very as long as you guys

能明白。

can see okay.

但在我看来，未来几年里，工程计算预算的讨论将会是最疯狂的一个。

But like, the the engineering compute budget conversation to me is gonna be just, like, the most wild one in the next couple years.

这简直就像

It's just like

你什么意思？

What do

你们把多少工程开支

you mean?

分配给了

Much did you allocate of your engineering expense

是的。

Yeah.

到令牌？

To token?

到令牌。

To token.

而且，你知道，根据你在Twitter上关注的人不同，这个比例可能是1%，而另一方则可能是

And and it's like, you you know, depending on who you read on Twitter, it could be 1%, and the other and the other side could be

100%。

a 100%.

而且，是的。

And it's like Yeah.

但这些东西，不，不是这样的。

But this stuff Well, no.

不。

No.

CFO们真的必须知道这个问题的答案。

What do CFOs have to literally they actually have to know the answer to that.