Stripe的AI负责人：AI代理将如何买卖和支付

但我觉得，当人类不再需要手动执行交易——不用输入任何凭证，也不用打开网页，一切由智能代理自动完成时，微交易才真正变得可行。

But I think when we enter a world where, like, the human is not doing any work to execute the transaction, you're not typing in any credentials, you're not navigating to any web pages, the agent's doing it for you, then micro transactions become viable.

再结合稳定币余额的自动扣减，我们就可以聊聊Tempo在这方面的工作了。

And you pair that with, okay, and now it's, you know, burning down some stablecoin balance, and we can talk about the work we're doing with Tempo there.

这样一来，微交易就变得非常现实了。

Suddenly, microtransactions become very viable.

我觉得我们很快就会进入一个新世界，尤其是在购买推理服务、代币或AI SaaS产品时，你会用你的链上钱包频繁发起微交易，而不需要逐笔确认。

I think we're going to quickly move to a world where, especially in buying inference or tokens or sort of AI SaaS products, there's gonna be you know, you are going to be using your link wallet to, make a ton of micro transactions and not approve everyone.

不过，如果你现在就想每笔都手动确认，当然也可以。

But yes, if today you want to approve everyone, you can.

但这肯定没法规模化，因为没人愿意为一美分的数据、一点研究结果或一个代币去反复批准交易。

That's not gonna scale because no one's gonna wanna approve the the 1¢ transaction for the little bit of data or the little bit of research or the token, that I needed for for this job.

那我们现在对可能出现的各种问题，有什么想法吗？

So what's our current thinking in terms of, like, everything that can go wrong?

你之前说，跟朋友聊这个的时候，他们会问：要是代理买错了东西怎么办？

You know, you were saying, like, when you talk about this to your friend, they they say, well, what if the agent buys the wrong thing and so on and so forth?

我的意思是，这显然是大家都在探索的前沿，但目前主流的想法是什么？

I mean, presumably, that's the frontier everybody's trying to figure it out, but what's the current thinking?

对。

Yeah.

我们正在从几个角度来思考这个问题。

We're thinking about this on a few dimensions.

一是，针对你提到的——如果传错了东西，或者内容不好之类的，企业必须继续作为交易的记录方。

One is, to your question on, you know, what if what if the wrong thing gets sent or the thing's not good or whatever, businesses have to remain the merchant of record.

这对我们来说是个核心设计原则，也就是说，我们希望智能代理的交易行为能像人类交易一样。

Like, that's a core design principle for us, which means, like, we see our job as having agentic transactions behave the way human transactions do.

但这并不意味着代理必须像人类那样思考或受到同样的限制。

Now that doesn't mean that agency behave like humans or be constrained in the same way humans do.

但从企业的角度看，他们仍然是交易的记录方。

But from the perspective of the business, they remain the merchant of record.

他们才是真正的销售方。

They are selling.

你知道，虽然交易是通过代理完成的，但本质上还是商家在做主。

You know, they happen to have that sale facilitated through an agent, but, like, the business is still, the business.

另一个设计原则，我们之前也提到过一些，就是要提供足够精细、可编程的控制和安全机制，让商业活动能够规模化开展。

Another design principle, which we've touched on a bit in different ways, is always provide appropriate granular programmable controls and guardrails to enable commerce to happen at scale.

也就是说，不管你希望控制得严严实实，还是想放开手脚，都不该被迫换新工具、换新钱包，或者改底层支付系统。

And that's basically saying, like, no matter how scoped in and in control you wanna be or how much you want it to be a free for all, like, you shouldn't have to adopt a new tool, move to a new wallet, change, you know, the underlying payment rails.

你也不该被迫调整产品目录，或者改变展示商品的渠道。

You shouldn't have to change your product catalog or where you expose it.

简单说，这些控制和防护机制，应该能自动适应不同需求，灵活伸缩。

Like, basically, out of the box, those those controls and guardrails should, should ebb and flow.

在消费者场景下，这意味着消费者始终掌握着如何授权代理进行购买的权限。

And I think in the consumer case, that looks like consumers remaining in charge of how Link authorizes agents to buy.

而在商业场景中，我之前简单提过共享支付令牌的事，核心是确保我们对买家及其代理的可信度信息，能完整传递给商家，让商家能据此做出明智决策。

And in the business case, you know, I touched on this briefly in the context of shared payment token, but it looks like making sure that the information we have about the goodness of the underlying buyer and the agent operating on that buyer's behalf are passed fully to the business to action intelligently.

所以，Radar 是我们用来防范欺诈的产品。

And so, you know, radar is our fraud, protection product.

我们已经用了十多年了。

We've had it for, over a decade.

以前主要关注交易欺诈。

It used to be really about transaction fraud.

现在它能检测各种欺诈、滥用、机器人行为，还有多层信誉评估——不只是最终客户，还包括代表客户的代理。

Now it looks at all kinds of, fraud and abuse and bot and multiple layers of goodness where there's not just the the the end customer, but also the agent operating on behalf of the customer.

这些雷达评分已经内置在共享支付令牌里，方便企业用来做判断。

And so those radar scores are actually included out of the box in the shared payment token, for businesses to businesses to reason about it.

但归根结底，商家仍然是交易的记录方。

But but, basically, I think business remains merchant of record.

消费者可以根据需要设置精细的权限和控制。

Consumer has as fine grained guardrails and controls as they need.

而在整个生态里，我们会尽可能让信息对称。

And then in the ecosystem, there is as much symmetric information as as we can create.

我们现在覆盖的交易规模，差不多占全球GDP的2%。

I mean, we now look across just about 2% of of global GDP.

在人工智能的背景下，我们关注的是这部分GDP中非常大的一块，因为几乎所有购买和销售AI服务的主体都使用Link和Stripe。

In the context of AI, we look at a very, very large share, of that of that GDP because basically all AI buyers and all AI sellers are on link and on Stripe.

所以，我们能真正察觉到哪些地方出了问题。

And so, you know, that that allows us to actually understand when something is going sideways.

我们很清楚自己的职责，就是提前提供这些信息，来保护整个生态系统。

And we see very much our job to protect the ecosystem by by providing that information up front.

当深入理解这种以智能代理为核心的商业模式时，真的非常有意思——技术能实现什么，作为负责任的参与者，我们就能做到什么，但归根结底，很多问题最终还是得靠法庭来裁决。

It's really fascinating as one unpacks this entire line of thinking around agentic commerce because this what's the technology can do, that's what you can do as a responsible player in the ecosystem, but like ultimately a lot of this will have to go to court one way or another.

意思是，法律体系必须跟上、做出调整，因为一旦大型交易出了严重问题，到底该由谁负责？

Mean the legal system will need to adapt and evolve because if something goes terribly wrong in a large commerce transaction, you know, who's at fault?

会不会是底层的模型提供商？毕竟，那个智能代理已经失控了。

I mean, could it be the the model provider underneath because, like, the agent went haywire?

我觉得，这一切都需要时间，慢慢在生态系统中理清头绪。

I guess all of this is gonna take time to work its way through the ecosystem.

是的。

Yeah.

不只是在购物方面。

Not just in purchasing.

对吧？

Right?

当有智能代理参与时，谁该为各种不良行为负责，而不仅仅是糟糕的购物行为呢？

Like, you know, who is at fault for not just bad purchasing behavior, but bad behavior of all types when when there's an agent involved.

所以我认为，这方面的情况确实在发生变化。

So so I do think the the landscape is changing there.

说到支付，你刚才提到，过去人们不太愿意把信用卡信息直接留在网上。

When it comes to sort of payments in particular, you you mentioned a bit ago that, back in the day, people didn't really trust just, like, putting their credit card on on the Internet.

我觉得挺有意思的是，像共享支付令牌这样的技术，让代理根本接触不到用户的原始凭证。

I think it's interesting to think, like, with things like shared payment tokens, agent never sees the underlying credentials.

每一笔交易都会由Stripe Radar实时评分。

Each transaction is scored in real time by Stripe Radar.

商家只需处理这些经过验证的交易即可。

Merchants go and handle the transactions.

账户信息根本不会像人类在非Stripe的其他随机网站上手动输入信用卡时那样，经过不可信的渠道传输。

Credentials, like, never travel through untrusted services the way they can when a human types a credit card number into a random website, not Stripe, but other random websites.

所以，我觉得还值得探讨一下：这种模式在支付安全方面，是不是真能带来实质性的提升？

And so, anyway, I I think it's also interesting to ask, like, is there is there sort of meaningful payments trust and safety upside here?

我觉得答案是肯定的。

And and I could imagine, yes.

所以总结一下，代理支付在防范欺诈和减少失误方面，最终应该会比人工输入卡片信息更安全。

So to play back, agent payments, would eventually be safer than than human just typing cards in terms of fraud, in terms of mistake presumably as well.

对吧？

Right?

比如，要是输错了一位数字之类的。

Like, if you type in the wrong number or something like that.

没错。

Yes.

只要做得到位，从长远来看，确实会安全得多。

Done like, done right in the limit, I believe it should be much safer.

我觉得，目前最主要的好处是更方便了。

Today, I think the first order effect is it's just more convenient.

对吧？

Like right?

通常来说，盗刷我信用卡的并不是骗子，

Like, usually, it's not a fraudster on my credit card.

而是我自己输错了CVC码。

It's like me fat fingering my own CVC.

这挺烦人的，也让人挺沮丧的。

It's annoying, and I'm frustrated.

不过说实话，现在我们主要享受到的还是便利性。

But, you know, right now, I think most of what we're getting is is convenience.

但从长远来看，确实可以想象，

But in the limit, yeah, you could you could totally imagine, okay.

现在没人再把支付凭证随意传到互联网上了。

Now nobody's passing these random payment credentials, over the Internet.

顺便说一下，人们其实并不会去细想那些通过十个不同支付方式进来的交易。

And by the way, people also aren't really necessarily reasoning about transactions coming in through 10 different payment credentials they have.

对吧？

Right?

他们就只有一个钱包。

They really have a wallet.

他们只看钱包里发生了什么。

They see what happens in the wallet.

这个钱包在各个商家之间都是以加密、安全的方式使用的。

And that wallet is used in a very tokenized, secure way, across sellers.

还有个特别有意思的话题，我想聊聊，你之前简单提过，就是氛围部署。

So another really interesting topic I wanted to cover and that you mentioned briefly is vibe deployment.

这到底是什么意思？

What does that mean?

就是你把东西做出来之后发生的事，但这方面讨论得太少了。

It's what happens after you build the thing, and it's actually not talked about enough.

不过说到编程这部分，很多人一直在谈AI怎么帮人写代码。

But, like, the coding part, a lot of people talk about AI for coding.

还记得一年前我们聊的时候，这确实说得通，因为那时候还没解决。

You know, when we were talking a year ago, that makes sense because it wasn't solved.

但现在这个问题基本上已经解决了。

Now that's basically solved.

我们的数据也清楚地显示了这一点。

And we actually see that it's solved in our data.

这些数据有点随机，但我觉得挺有意思的。

These are semi random facts, but I I find them interesting.

自从我们一年前聊过之后，AI代理访问Stripe文档的流量增长了十倍以上。

So, like, agent traffic to Stripe's documentation grew more than 10 x since we talked a year ago.

现在，AI代理的访问量已经占到了我们所有文档流量的40%左右。

Agent traffic is now about 40% of all our docs traffic.

也就是说，这些代理在主动研究，会去查Stripe的技术文档来搞清楚问题。

And meaning meaning that the agents are trying to figure it out, and they go to the Stripe technical documentation to understand

也就是说，现在的程序员，很多时候更像是智能代理，而不是人类开发者。

Meaning the coders now are are almost as much agents as they are developers.

在某些领域，几乎全是智能代理在干活，根本看不到人类开发者。

And in some segments, they are basically all agents and and, not at all developers.

另一个例子是我们命令行工具，过去这工具很冷门，只有少数开发者会用。

And another example actually is is our CLI, our command line interface, which historically was, like, a pretty niche tool used by, like, a pretty small group of developers.

但现在它突然爆火了，我们都惊了：这到底是怎么回事？

It now it just, like, exploded, and we were like, what is happening?

现在有70%的API请求都来自智能代理。

And it's now, 70% of its API resource requests are from agents.

所以可以说，今天用我们命令行工具的主体，绝大多数根本不是人。

So, like, you could think of, like, the majorities the majority of entities that are using our CLI today, like, aren't people.

总之，这些只是Stripe的几个小例子，但整个生态里类似的情况多得是，都说明一件事：AI写代码已经成真了。

So, anyway, I think though, you know, those are just two, like, fun Stripe anecdotes, but there's lots of them across the ecosystem you can look at that tells you, like, vibe coding is real.

对吧？

Right?

或者你自己的亲身经历也是这样。

Or just your own lived experience is the same.

对吧？

Right?

那部分是成功的。

That part worked.

但接下来呢？

But then what?

比如，一个智能代理能在二十分钟内给你写出一个完整的可用程序。

So, like, an agent writes you, like, a complete working application in twenty minutes.

太棒了。

Fabulous.

我们真的很喜欢。

Like, we love it.

好吧。

Okay.

但在应用真正上线之前，你还是会遇到挺大的障碍。

But you've still got this pretty big friction before that app is actually live.

对吧？

Right?

你得去搞，嗯，具体怎么做要看你做什么项目。

So you gotta go, like I don't I mean, depends what you're doing.

但你大概得先给数据库服务商注册账号，再搞认证服务，然后是托管平台，来回折腾。

But you gotta probably create an account with your database provider, then your auth provider, then your hosting service, and you're, like, bouncing around.

一堆后台页面等着你处理。

You got all these dashboards.

你自己都不清楚怎么搞的，就手动复制粘贴，管理各种凭证和API密钥之类的。

You're, like don't know how you do it, but copy pasting stuff by hand and managing credentials and API keys and whatever else.

每个服务都有自己的新手引导流程。

And, and every one of those services has its own onboarding flow.

跟我们之前说的支付流程有点不一样，但整体来说，填一堆步骤确实挺低效的。

So like a little different flavor than the payments flow we talked about, but like, it's a pretty inefficient fill out a bunch of steps.

每一个服务的注册流程，都像是专门为像你我这样的人设计的，让我们坐下来一点一点点完这个古怪的设置向导。

And every one of them, just like the payments flows was designed for like you and me as humans sitting down and clicking through this like weird setup wizard thing.

以前我们其实没觉得这有多烦，因为谁也没频繁做这事——真正耗时间的是写代码那部分。

And I don't think it really bothered any of us that much because we weren't doing it that much because the hard part was the coding part.

但现在不一样了，应用二十分钟就能写完，真正拖后腿的反而是部署环节。

But now that, like, the app can be built in coded in twenty minutes, like, okay, the long pole is deploying the thing.

所以，写代码的体验是顺了。

And so, anyway, vibe coding was easy.

但部署的体验，反而成了最大的瓶颈。

Vibe deployment, has become, like, more of the binding constraint.

所以我们最近几个月，其实推出了 Stripe Projects，就是为了解决这个问题。

And so we actually we we recently, like, in the last couple months, launched Stripe Projects for this.

说白了，智能代理应该能直接注册、配置、集成部署应用所需的所有服务，而且这一切都能在命令行里一键完成。

But, basically, it's like agents should be able to, well, sign up for and configure and integrate all of the services they need to deploy an app, and they should be able to do that, like, right from the command line.

这不光是 Stripe 的服务，而是所有相关服务都该支持。

And it's not just Stripe services.

对吧？

Right?

整个生态就是这样的，你随便选，Vercel、Supabase、Cloudflare、Twilio，需要什么就用什么。

It was like a whole ecosystem of, like, you get, like, Vercel and Supabase and Cloudflare and Twilio, like, whatever you need.

Clerk，我们大概一两周前刚宣布又多了16个合作伙伴。

Clerk, I think we announced, like, 16 more partners a week or two ago now.

不过整个生态还是充满热情，因为大家其实都面临同样的问题。

And, anyway, there's still a lot of enthusiasm from the ecosystem because it turns out that everyone kind of has the same problem.

开发者们都有同一个痛点：现在最大的瓶颈是部署。

Developers all have the same problem, which is now, like, the long pole is, like, the deployment.

企业也是一样，他们的注册流程是为人类设计的，现在得改成适配智能代理。

And businesses have the same problem, which is, like, their onboarding was designed for humans, and now they need it to work for agents.

那回到重点，Stripe为什么要在意这个？

And to bring it home, why does Stripe care?

没错。

Okay.

所以我觉得，Stripe关心的很多事，表面上看确实让人疑惑：为什么Stripe会在意这些？

So I think there's a lot of things that Stripe cares about that on paper, you'd be like, why does Stripe care?

说实话，我们关心是因为它已经成了瓶颈。

I'm the the honest answer is we care because it was becoming the bottleneck.

构建应用的门槛没了，但部署的门槛却成了实实在在的障碍。

Like, the barrier to building is gone, but the barrier to deploying is like a real friction.

如果把视角拉远一点，撇开我们是支付公司的身份，我们的使命其实是提升互联网的GDP。

And if you just zoom back, like, okay, separate from being a payments company, our mission is to increase the GDP of the Internet.

这其中很重要的一环，就是怎么让更多公司顺利起步，帮他们更快赚到第一笔钱，然后逐步扩大规模。

A big part of that is how do we get more companies off the ground and how do we help them get kind of their first dollar faster and scale from there.

如果一个人有了点子，能做出应用，却没法部署，那他就卖不出去。

And so if a person with an idea can create an app but can't deploy it, they can't sell.

所以，降低部署的阻力，直接就能推动互联网经济的增长。

And so removing deployment friction directly obviously like expands, the the internet economy.

你可以把它理解为一种协调和整合的工作。

And, you know, you can think of it primarily as, like, just just like orchestration.

我们其实就是做协调工作。

Like, all we're doing is orchestration.

如果别人也想做、能做、真的做了协调，我们完全没问题。

So, if someone else wanted to and could and did orchestration, we'd be cool with that.

我们看着开发者，想让他们把产品上线。

But we were, like, looking at the developers, trying to get the thing live.

我们也看着那些企业，希望他们能被开发者顺利使用。

And we were looking at the businesses, trying to get the thing, like, used by developers.

于是我们就想，好吧。

And we're like, okay.

我觉得我们可以让这个市场变得更顺畅一些。

I think we can we can make this market a little bit smoother.

对。

Yeah.

再回过头说说协调这件事，因为你有这么多合作伙伴。

And to playback orchestration because you have all these partners.

所以最终你们做的是主机监控、邮件、队列和密钥管理这些功能，但都是由不同供应商提供的，你们负责把它们串联起来，确保用户能安全高效地部署自己的代理。

So, ultimately, you do hosting observability, email, queue secrets, all the things, but that's provided by different vendors and you provide the glue to make sure that people can deploy their agents safely and efficiently.

对，就是这样。

Yes, exactly.

没错。

Exactly.

太棒了。

Very cool.

好的。

Alright.

这就是 Vibe 部署的模式。

So that's vibe deployment.

代币当钱用这个话题也很有意思。

Tokens as money is also a fascinating topic.

这是从顶层开始切入的思路。

It's like taking it from the the top.

从Stripe的角度来看，货币化代币意味着什么？

What does it mean to monetize tokens from a Stripe perspective?

这是个大问题，因为从整个世界的角度看，货币化代币到底指的是什么？说白了，

It's a big question because what does it mean to monetize tokens from a world perspective is kinda like, okay.

下一代B2B和部分B2C模式，到底该怎么赚钱呢？

Actually, how's the whole next generation of b to b and some b to c gonna monetize?

你看，

Look.

过去这十多年，SaaS的盈利模式一直相当清晰简单，

Like, the the the last kind of decade plus of SaaS, had, like, pretty beautiful and simple to monetize economics.

对吧？

Right?

尤其是SaaS，你开发一次产品，每多一个客户，服务他们的成本几乎为零，

And in particular with SaaS, like, you build a product once, and then you get one more customer, and it costs you basically nothing to serve them.

边际成本几乎可以忽略不计。

Like, marginal costs are near zero.

所以SaaS的利润率才这么高，这也是为什么固定费用订阅或按席位授权在SaaS领域特别有效。

And that's why SaaS margins are really good, and that's why, you know, the fixed fee subscriptions or seat based licenses work really well, in SaaS.

AI就不一样了，我说的AI是广义的，因为你可能直接卖大模型，也可能卖一个基于大模型的封装产品，或者一个重度依赖大模型、消耗大量令牌的应用。

AI, and I say AI generally because, like, you could literally be selling LLMs, but you could also be selling, you know, some product that's a wrapper on top of LLMs or a product that's heavily powered by LLMs and and requires a lot of tokens.

这打破了原来的模式，因为现在每一次提示、每一次API调用、每一个任务都产生了实实在在的边际成本——这在SaaS里是不存在的。

It breaks that model because, obviously, every prompt and every API call and every task has a a real marginal cost all of a sudden, which it didn't have in SaaS.

推理不是免费的。

The inference isn't free.

于是现在出现了大量这样的企业：客户怎么使用你的产品，直接决定了你是赚钱还是亏钱。

And so you now have all these businesses where how your customers use your product directly determines whether you make or lose money.

这完全是另一种游戏了。

And that's a very different game.

从我们的角度看，这其实意味着，对AI公司来说，按使用量计费变得至关重要。

And, you know, from our vantage point, like part of what that boils down to is the need for usage based billing as really critical for AI companies.

你必须能实时计量客户实际消耗了多少资源，然后以一种能匹配你底层成本的方式收费。

Like, you need to be able to meter what customers are actually consuming in real time and then charge them, in a way, obviously, that that aligns with your with your underlying costs.

从你的角度看，目前AI领域的创业公司是不是基本都用按使用量计费了？还是说仍然有部分在用按座位收费的方式？

And do you see from your perspective pretty much all the players in the AI economy from a vendor's startup standpoint, use usage based billing at this stage, or is that still a mix between per seat and per usage?

我看到的那些已经规模化的AI公司，几乎没人还只靠订阅或按座位收费了。

I see very few scaling or scaled AI companies that are still exclusively subscriptions or seat based.

从我跟他们的交流来看，原因很简单：这种模式的经济账算不过来——有些人用得特别多，有些人却用得很少。

And I believe from my conversations with them that that is for the simple reason that the economics there don't make sense because, you know, you have some people that are, using a ton and some people that are not using very much.

用得多的人成本高，用得少的人成本低，如果没有按使用量计量，根本没法把真正的大客户和小客户区分开来，也就没法合理定价。

And these people cost you a ton and these people cost you not very much, and it is very hard to separate the sheep from the goats and price them appropriately without a usage based meter.

不过，很多公司虽然主打按使用量收费，但还是会搭配一个固定费用的订阅方案。

That said, many of those businesses have a usage based offering, but it is complementary to what you can think of as like a fixed fee subscription.

比如Lovable就是一个很好的例子。

So like a great example is Lovable.

他们刚上线的时候，就用Stripe Billing做了个简单的订阅制。

When they launched, they had like a simple subscription through Stripe Billing.

这完全说得通。

Makes total sense.

他们起步很早。

They were early.

他们进展很快。

They're moving fast.

他们想尽快实现盈利。

They wanted to monetize quickly.

订阅模式对消费者来说也很熟悉、容易理解。

Subscriptions are also very familiar and easy for consumers.

你想想Lovable，他们的部分目标用户其实并不太懂技术。

You know, if you think about a lovable, like, some of their, target users are, like, not very technical.

这些人怎么去理解什么令牌啊、积分啊这些东西呢？

Like, how are they gonna feel about, like, oh, how should I reason about a token or a credit?

对吧？

Right?

明白了。

Okay.

所以他们一开始用的是订阅制。

So so they they started with subscriptions.

但随着公司发展，成本上升，他们的计费方式也得跟着调整。

But then as the as the company grew and and some of their costs grew, their billing needs evolved.

他们需要根据实际的令牌使用量来收费，至少得部分这样算。

And, they needed to charge at least somewhat based on actual, token consumption.

于是他们采用了混合计费模式——现在很多企业，尤其是有B2C业务的公司都在这么干：在订阅基础上叠加按使用量收费。

And so what they did is a hybrid billing model, which which actually we're seeing a large number of businesses, especially businesses that have a b to c component do, where they have usage based billing on top of their subscriptions.

当客户用到某个额度上限时……

So customers hit some threshold.

你懂的。

You know?

顺便说一句，很多人会先提供免费版，超过一定额度后，就收固定月费，比如每月25块或100块，包含一定数量的积分。

And by the way, many people have freemium thresholds, but then above that, they often have, like, a fixed fee, you know, $25 a month or $100 a month, up to some number of credits.

再往上，就转成按实际使用量计费了。

And then above that, you you have usage based billing.

所以一旦客户超过阈值，Lovable 就会根据超出部分的令牌使用量精确收费。

So once customers hit the threshold, lovable, in this case, charges, like, precisely based on the number of tokens consumed above that.

我觉得这种模式挺好的，因为一开始用订阅制能吸引很多人进来，等用量上来之后，你的收入就直接跟客户使用产品的程度挂钩，也跟你们提供服务的实际成本同步增长。

And, you know, I think that that that alignment is good because you get a bunch of people in the door comfortable with subscriptions, but then sort of your at at any volumes that matter, your revenue is scaling very directly with, like, how customers are using the product and correspondingly with, like, how much it is costing you to provide that product.

客户只付他们实际用掉的部分，而你也能确保覆盖掉背后相应的运营成本。

So customers pay only for what they get and you make sure you monetize, for, for the underlying cost that you're gonna have to bear.

Eleven Labs 就是另一个例子。

And eleven Labs is another example.

他们也经历了完全相同的过程，一开始就是纯订阅制。

Like they went through the exact same thing, literally, like started with subscriptions.

最近他们改成了这种按使用付费的模式。

They recently moved to this sort of pay go plan.

这几乎是现在人工智能领域普遍在发生的一种趋势。

And and this is just a pattern that we see playing out pretty much everywhere in the AI space right now.

一旦开始向代理收费，计费方式会有很大变化吗？

Does billing change much, once you start charging agents?

我觉得是这样，主要有几个原因。

So, I think so for a couple reasons.

首先，智能体可以以机器的速度进行消耗。

One is agents can consume at machine speed.

所以，哪怕你是按用量收费，但如果是月底才结算，等到月底一看，一个智能体可能已经花掉了一笔巨款。人类虽然也能通过让智能体帮自己干活来增加消耗，但相比之下，智能体自己产生的费用往往高得离谱。

And so even if you're, like, even if you're doing a usage based thing, but you're like charging at the end of the month, by the time you've gotten to the end of the month, an agent can have spent bet a human can too to some extent with agents working for them, but like especially an agent can have spent like an egregious amount.

因此，我认为在智能体领域，我们会越来越多地看到实时计量和实时计费。这正是我们在 Metronome 和 Tempo 之间合作构建的模式：Metronome 负责复杂模型的实时计量和按量计费，而 Tempo 基于区块链技术，让智能体能够实时消耗代币并即时支付相应成本。

And so actually, I think in the world of agents, what we're going to see more and more of is real time metering, like what have you consumed and real time billing, which is actually what we have co built between Metronome, so real time metering, usage based billing for very complex models, and Tempo, blockchain, where, you know, agents are consuming tokens in real time and, paying down, for the cost of those tokens.

这点很重要，因为智能体可以购买算力速度。

And that's important because agents can buy machine speed.

这是可行的，因为 Tempo 和 Metronome 之间的很多基础设施都支持这一点。

It's viable because, of a bunch of the the infrastructure between Tempo and Metronome.

而且，智能体非常乐意按实际使用量付费，字面意义上的那种。同时企业也需要它们按量付费，这样就不会产生大量账单然后突然停止使用。

And, agents are happy to just, like, pay as they go in a very literal way, plus businesses need them to be paying as they go so that they don't, rack up a bunch of spend and then and then go dark.

对。

Yeah.

这还带来了很多连锁影响，尤其是在财务和会计领域，那些系统也得跟上这种速度。

It's it's fascinating also in terms of downstream consequences for what it means for, like, finance and accounting and, like, systems that need also to move at that speed.

没错。

Totally.

我们有个收入确认的会计产品，主要用在那些有传统订阅模式的企业身上。

I mean, we have a a revenue recognition accounting product, and it's and it's mostly needed by I mean, we we provide it to a bunch of people who a bunch of businesses who have traditional subscriptions.

但真正最头疼的，还是那些靠Excel做传统记账的公司——当微交易大量涌现时，表格根本撑不住，行数直接爆了。

But where there's the most acute pain is actually, like, traditional accounting in spreadsheets, like, does not work when you have this just like, like, proliferation of rows because these are like, micro transactions are are truly happening.

我觉得这也在重新定义会计这个职业。

I also think it's changing what it means to be an accountant.

我就不点名了，但之前我跟一家挺成功的AI公司的财务二把手聊过。

I I I will, avoid naming the company, but I was talking to the sort of number two in accounting at a pretty successful AI company.

有几件事挺有意思的。

And a couple of things were interesting.

他们其实就是个会计兼工程师的结合体，因为要处理的数据量太大了，不得不这样。

One is they were like a hybrid accountant engineer, which they needed to be because of the scale of the data they were dealing with.

其次，他们做的并不是我们传统意义上那种只是结账的会计工作。

And two, they weren't doing, like, accounting in the traditional you and me sense of, like, just close the books.

他们当然也需要结账，这没问题。

Like, they have to close the books for sure.

但他们的工作还包括发现一些异常的、奇怪的情况。

But their job was also to find, like, weird anomalous stuff happening.

我当时找他们聊，是因为他们发现了一些欺诈模式，想请我帮忙分析。

And, I was talking to them because they were, like, identifying some fraud patterns and wanted help with them.

不过说到底，这真的挺有意思的。

But anyway, just just very interesting.

比如说，好吧。

Like, okay.

在这些AI公司里，会计工作到底意味着什么？

What is actually what does it mean to do accounting at, like, one of these AI companies?

肯定需要全新的工具。

Definitely needs new tools.

肯定需要新的系统。

Definitely needs new systems.

可能还需要完全不同的技能组合。

Probably a different different skill set.

而且，你真正该负责的，远不止是结账而已。

And then, like, what you're accountable for really isn't just closing the books.

而是要纵观整个收入确认流程，看看这些数据能反映出我们业务的整体状况。

It's like looking across the the whole rev rec stack and saying, what does this tell me about the health of our business?

这些数据又能揭示我们业务中是否存在欺诈或滥用行为？

What does this tell me about fraud and abuse in our business?

还有，它们能不能暴露出我们产品里的漏洞？

And what does this tell me about breakages in our product?

没错。

Yeah.

所有这些都变成了一个需要实时处理的海量数据问题。

All of which becomes a massive data problem that needs to be treated in in in in real time.

对吧？

Right?

所以才会出现像ClickHouse这样的平台。

Hence, the rise of the click houses of the world.

谁能实时处理海量数据，谁就成为基础设施的核心部分。

Like, whoever can process massive amounts of data in real time becomes the core part of the required infrastructure.

明白了。

Okay.

你之前提到过代币被盗的问题，我想确保我们讲清楚这一点。

You wrote somewhere about token theft, and I wanna make sure that we cover that.

那到底什么是代币被盗？

So what is that?

作为一种新型欺诈手段，代币被盗指的是什么？

What is token theft as a new form of fraud, I guess?

是的。

Yeah.

这是一种新型的诈骗手段。

It's it's a new form of fraud.

我觉得，这可能是目前人工智能领域最受忽视的话题之一，而且被忽视得相当严重。

It's, I think, one of the most under discussed topics in AI right now, maybe by a lot.

骗子们已经发现，在人工智能领域，其实根本不需要去偷钱或者账号信息。

Fraudsters have figured out that in AI, you actually don't really need to steal money or credentials.

你只需要偷走令牌就行。

You can just steal tokens.

而这些令牌是有真实价值的。

And tokens have real value.

对吧？

Right?

你可以用它们来开发各种应用。

You can use them to build things.

也可以在平台上转卖。

You can resell them on marketplaces.

你可以在上面包装一个新产品，一分钱不用花，然后拿去卖这个所谓的全新产品。

You can wrap a new product on top, without paying a cent and go and sell that real that new product.

这有点像转售，但看起来你根本不是在卖那个订阅服务。

So it was kinda like resell, but it doesn't doesn't look like you're selling the subscription to the thing.

别人以为你卖的是自己开发的产品，实际上后台完全靠的是别人的系统在支撑。

It looks like you're selling this other product that you've come up with on your own, but really in the back end, it's completely powered by someone else's thing.

对AI公司来说，也许我们之前谈SaaS时已经隐含了这一点，但这种风险是致命的。

And for AI companies, maybe this is implied by our earlier conversation on SaaS, but like the risk is existential.

对吧？

Right?

如果有人偷了一点你的SaaS服务，可能根本无所谓，因为对你来说边际成本几乎为零。

If someone stole a little bit of your SaaS, like, didn't matter because it didn't really cost you anything on the margin.

但要是有人偷了你的使用额度，只要欺诈比例够高，你的产品经济模型很快就会崩盘。

When someone steals your tokens, if your fraud rate is high enough, the economics of your product actually, break really fast.

所以我们跟几乎所有AI公司都有合作，能近距离看到真实发生的情况和欺诈模式是如何展开的。

So anyway, because we, work with basically all the all the AI companies, we have this interesting front row seat into, what's actually happening and the fraud patterns, playing out.

现在有三种情况。

And right now, there's three.

但我认为这三种不会一直存在，因为过去三个月我们已经基本解决了这三类问题。

And I don't think these will be the three forever because we've already we've already largely burned down these three over the last three months.

但新的问题会不断冒出来。

But new ones will new ones will pop up.

我觉得这三种情况特别有代表性，能帮我们理解这个领域的广度。

I think these three are just, instructive for reasoning about sort of the breadth of the space.

第一种是多账号滥用。

One is multi account abuse.

就是一些不法分子反复注册，不停地刷新用户优惠券。

So this is like bad actors just, like, hammering your sign up again and again and again and again so that they can get those, like, new user credits.

当我看到数据时，这个规模真的让我很震惊。

And the scale of this actually shocked me when I looked at the data.

AI公司里，每六个新注册用户中，就有一个以上是这种作弊行为。

So, like, more than one in six signups at AI companies are this kind of abuse.

顺便说一下，这对公司来说也很让人困惑。

And by the way, that's also very confusing for the company.

这些到底是真正的客户吗？

Like, are these good customers?

这些人到底是谁？

Who are these people?

但对公司来说成本很高，因为我为少数人花了大量算力，他们不断注册大量账号，还消耗了相当多的资源。

But it's very expensive for the company because, like, I've spent a bunch of tokens on these free products for, you know, a relatively small number of people who are spinning up a massive number of accounts and actually, consuming quite a bit of tokens.

明白了。

Okay.

这属于最前端的问题。

So that's like very, very top of funnel.

另一个我觉得有意思的案例是免费试用欺诈。

Another example that I think is interesting is is free trial abuse.

这些是骗子过来钻空子。

So these are like fraudsters come in.

他们注册免费试用。

They create a free trial.

然后填上支付方式。

They put down a payment method.

把所有积分都用光，但根本没打算转为付费用户。

They they drain through, like, all of the credits, but they never have any intent to convert.

免费试用滥用一直存在于互联网各处，不管有没有AI，但过去六个月在Stripe上翻了一倍多。

And, you know, free trial abuse has always existed in various forms across the Internet, AI or not, but it's more than doubled on Stripe in the last six months.

而这翻倍的部分，大部分都来自AI公司。

And most of that doubling is coming from AI and, AI businesses.

举个例子，这事儿利润有多高呢？现在都形成了一整套灰色产业。

And just to give you a sense, like, how lucrative this is, like there's whole cottage industries built around it.

你有没有收到过推广免费试用卡的广告？

So I don't know if you've ever been marketed a free trial card.

我最近就收到了。

I was recently marketed one.

这 basically 就是说，你可以轻松生成免费试用卡。

It's basically like literally, it's like, oh, you can like spin up free trial cards.

这些卡二十四小时后就过期，所以你根本不用付钱。

They expire in twenty four hours so you'll never have to pay.

如果你我有这么一张卡，大概会用它来正经试试服务，省得麻烦去取消。

And if you or I had one, we'd probably use it for some legitimate purpose to try out a service and just not have to go through the pain of canceling.

但真正的骗子会大量生成这些试用卡，疯狂消耗积分，根本没打算真正付费。

But literally fraudsters can just explode these things and spin up a bunch of free trials and spend a bunch of tokens with no intent to convert.

再往下一点，就是我们之前说过的按使用量计费的问题。

And then the third is is a little further downstream of this, which is basically like we talked about the usage based, billing.

有些人用掉几千美元的资源，月底账单来了却死活不付钱。

Folks are racking up, like, thousands of dollars of costs and then being billed at the end of the month, and never paying.

所以呢，也就这样了。

And so, like, whatever.

餐厅里那种吃完不结账就跑路的事儿，现在换成了‘用完积分就跑’。

The dine in dash that happens in restaurants is like a dine in dash, but it's but it's for for tokens.

但不幸的是，到了这一步，这又是AI，不是SaaS，所以公司自己承担了这些令牌的费用。

And, unfortunately, at that point, again, this is AI, not SaaS, So, like, the company has borne the cost, of of those of those tokens.

如果这听起来太明显了，抱歉，不过说实话，我连当个骗子都不够格，因为这事儿对我自己来说都没那么清楚。

And sorry if that's obvious, and I guess I would be a terrible fraudster because it's it's not even super obvious to me.

如果我白得了这些令牌，我到底能拿它们干嘛？

So, if I get tokens for free, what do I actually do with them?

比如说，如果我从像CLOUD或者ChatGPT这样的通用大模型那里拿到令牌，我还能明白能怎么用。

So I guess if I get tokens from a general purpose LLM like CLOUD or ChatGPT, I could see what I do with it.

但如果我去用Cursor、Lovable或者Eleven Labs这类服务呢？

If I if I go to Cursor or Lovable or eleven Labs.

那我到底能拿这些令牌做点啥？

So what is it that I actually do with this token?

嗯，明白了。

Okay.

你到底怎么用这些令牌，完全要看你用的是什么服务。

So what you do with it very much, changes based on, like, what's the service?

这些令牌到底是用来做什么的？

Like, what exactly were these were these tokens, meant for?

你说到点子上了，没错。

You nail on the head for like, okay.

你就是直接从底层的大模型那里获取令牌。

You just, like, get the tokens from the the underlying LLM.

行吧。

Fine.

对于那些在此基础上做了一层服务的公司来说，这成了大量转售滥用的问题。

For those sort of, businesses that are a layer above that, a bunch of resale abuse.

你实际上会把它们拿到别的市场去卖，价格比原价低一点，而你根本没付过这个钱。

So, you literally sell it sometimes in other markets for a slightly discounted price a discounted price on the base price that you should have paid but you didn't pay.

所以存在一个地下市场，有人会说，用 Cursor 或 Lovable 只要两块钱，但我的成本是零。

So there's a dark like, a dark web of marketplaces where you say, like, use a cursor or lovable for $2, but my my cost is zero.

那我岂不是能赚钱？

Therefore, I make money?

是的

Yeah.

没错

Exactly.

所以我直接把我的登录信息给你，随便什么都行。

So so I just I give you my login credentials, whatever.

还有一种情况，虽然看起来是小众的，但实际上存在大量高度特定的欺诈模式。

There's also but but then there's just, like, there's just this, like, long tail makes it sound small, but a very domain specific fraud pattern.

比如，有人偷取令牌来生成内容，然后用各种骗术牟利。

So for example, people steal tokens to create content that they then use to extract money in all sorts of scammy ways.

一个简单的例子就是，有人大量生成音乐曲目，上传到 Spotify 和 Apple Music，刷虚假播放量，再骗取版税。

So a simple example, we see people going in and, like, mass generating music tracks and then uploading them to Spotify and Apple Music and then getting fake streamers and then collecting royalties.

还有，在说唱类业务里，更有一层套一层的玩法——他们不光在淘宝上转售订阅，还会直接克隆 AI 公司的网站，让你能随便搭个页面。

Or then for, like, basically, all of the rapper businesses, there's this whole other layer of wrapper on a wrapper where, like, rather than just resell the subscription on places like Taobao, they'll, like, literally clone the AI company where you can just, like, vibe code a website.

后端只是把从被偷服务那里拿来的结果直接吐出来而已。

The back end is just, like, spitting out exactly what you got from the service that you're stealing from.

然后你把这产品当成自己的卖，还便宜点。

And then you, like, sell the product as yours, but cheaper.

你是打算告诉别人，他们其实很有创意吗？

Are you gonna give it to people that they're creative?

我的意思是，这听起来比真开家公司还难。

I mean, that sounds almost harder than starting an actual company.

是的。

Yeah.

我不太确定。

I I don't know.

我的意思是，确实有不少人这么做。

I mean, a bunch yeah.

他们确实花了不少时间在这上面。

They they definitely they definitely put a lot of time into it.

代币也很值钱。

Tokens are also very valuable.

而且你知道吗，当我们刚开始注意到这些趋势的时候，大概六个月到九个月前，它们还只是零星出现，但后来一下子变得严重多了。

And, you know, what's been what's been interesting as we sort of started seeing these trends, maybe at this point six, nine months ago in various flavors, but then they they escalated a bunch.

跟一些AI公司聊过之后，我是说那些大型AI公司，它们全都高度关注这个问题。

You know, talking to AI companies, AI company I mean, the the large AI companies are all over this.

这简直关系到它们的生死存亡，直接影响利润。

Like, it's like one of the most existential things for their margins.

它们一直和我们并肩作战，一起发现这些问题。

They, have been, like, in the trenches with us identifying the issues.

它们经常说，我们一看到问题，就马上动手解决。

They have been like, you know, we we literally, like, we see a problem.

我们发现问题，就开发一个模型来应对。

We build a model.

比如多账号滥用这种问题，我们就专门建模型处理。

We like you know, multi account abuse.

没错。

Okay.

比如在登录时，会调用一个 API。

Like, at the time of login, there's an API.

你把你知道的信息发给我们。

You send us what you know.

我们回传一个评分给你。

We send you back a score.

如果评分不好，你就直接拦截。

You block if they're bad.

用户开始免费试用时，也是同样的流程。

At the time of free trial start, same thing.

当用户使用量逐渐增加时，你把所有元数据发给我们。

As people are accumulating usage, you send us all the metadata.

我们会告诉你是否涉嫌欺诈，你可以要求他们充值，或者暂停服务，等等。

We send you back whether they're fraudy and you can require a top up or a cutoff service or whatever.

我们每遇到一种情况，大概几周内就能搞定，说实话，这甚至算不上一个完整的产品。

Like, literally, each of those, as we've seen them, we've, like, order of weeks gotten you know, generally, this isn't even like a full fledged product.

这其实就是个API。

It's like an API.

你把一些数据发给我们。

Like, you send us some stuff.

我们再把结果反馈给你。

We send you some stuff.

而且大家的接受度还挺自然的。

And the adoption has just been like, okay.

所以那些AI公司都在积极采用这个方案。

So the the AI companies are are are all over this.

我觉得有意思的是，每家公司未来都会变成一家AI公司。

I think what's interesting is, there's like, every company is going to become an AI company.

但整个行业目前还没意识到这一点，也没真正想清楚：现在大部分欺诈行为其实不是传统的账号或支付欺诈。

And I don't think the industry at large is thinking about this yet or has really even reasoned that, like, actually, most of the fraud that's happening is not traditional, like, credential or payment fraud.

它更像我们过去常说的第一方滥用。

It's like what we would historically have called, like, first party abuse.

对吧？

Right?

比如转售滥用、账号共享、多账号注册、免费试用，这些都属于第一方滥用。

Like, resale abuse or account sharing or multi accounting or free trials, like first party abuse.

我并不是说以前就没有第一方滥用的情况。

And I think it wasn't that first party abuse didn't happen before.

只是在传统的SASE体系里，第一方滥用根本不会让你付出代价。

It's just at least in sort of SASE stuff, first party abuse didn't cost you anything.

所以，像免费获取一点Salesforce这种事，根本没什么价值——就像我们之前举的例子，在大多数SASE场景下根本不适用。

And so, a, it wasn't that useful to get, like, a little bit of Salesforce for free where, you know, the examples we talked about wouldn't be relevant in in most SASE.

而且就算你能找到办法从中捞点好处，作为欺诈者能赚点小钱，但对商家来说，边际成本几乎是零，损失也不大。

And and b, even if it was useful to figure out a way to, like, skim off the top, and you could get a little bit of of profit for it as the fraudster, it didn't actually cost the business that much because their marginal cost was zero.

所以，随着每一家公司都变成AI公司，我们在做Radar项目的过程中，逐渐意识到我们的反欺诈产品必须从关注单笔交易，转向覆盖整个客户生命周期，从传统欺诈转向全流程滥用的防范。

And so, you know, as every business becomes an AI business, I think, we've we've been, in the context of our work on on radar, really reasoning about our, fraud prevention product as moving from transaction to full customer life cycle and moving from traditional fraud to end to end abuse.

但我认为，整个行业还没跟上这个认知。

But I don't think, like, the whole industry is is there yet.

你知道吗，你和我之前聊过很多AI带来的经济收益，但我认为，只有在安全的前提下，这些收益才能真正实现。

And, you know, you and I talked about much of the economic upside of AI, but I think that'll really only be realized, if it can happen safely.

比如，六个月到九个月前，我还在和一些AI公司聊这个事。

So for example, six, nine months ago, I was talking to some of these AI companies.

他们那时候总说：哦，我知道了。

They they'd be like, oh, I know.

我只要把免费试用取消了，就能解决免费试用滥用的问题。

I'm gonna solve my free trial abuse problem by cutting off free trials.

就是说，干脆只做销售驱动，只瞄准企业客户，不再搞产品自驱动增长，对吧？

Like, I'm gonna solve my free trial abuse problem by only having a sales sold motion and only going after enterprises and not having PLG and right?

不过现在，我听到这种说法的次数少多了。

And, like, you know, I hear less of that today.

为什么呢？

Why?

不是因为欺诈问题彻底解决了，而是因为大家都明白，AI代理也得能当买家才行。

Not because the fraud's totally solved, but because, everyone knows they need agents to also be their buyers.

如果代理要成为他们的购买者，那他们就必须提供自助服务的路径。

And if agents are gonna be their buyers, they better have a self serve motion.

他们必须要有产品驱动增长的模式。

They better have a PLG motion.

没人会愿意放弃这么重要的增长来源，只专注于那种高度依赖销售推动的模式。

Like, there's no way they wanna siphon off that source of growth, and sort of only double down on, like, a highly secure sales sold motion.

但最近关于令牌被盗的情况，确实挺让人关注的。

But, it's been it's been interesting to see what's happened with token theft.

我完全同意，欺诈者确实很有创意，但这也恰恰说明这些令牌有多值钱。

I I totally agree the fraudsters are are creative, but I think that's, like, a manifestation of how valuable the tokens are.

所以，我觉得这种创意不会消失，只会越来越强。

And so I actually don't think that creativity is is going anywhere.

Radar 一方面能做到实时响应，另一方面，应该也是完全由人工智能驱动的吧？

And Radar is, one, real time and, two, presumably entirely AI driven as well?

这不就是人工智能在跟人工智能对抗吗？

Mean, it's like AI fighting AI kind of a kind of thing?

是的

Yeah.

它是实时的。

It's real time.

它是AI驱动的。

It's AI driven.

而最重要的是，它的独特之处在于能查看整个Stripe网络的数据。

And then I actually, like, most importantly here, for its, like, differentiation is just looks across the Stripe network.

基本上，我们没见过的优质AI买家几乎没有，不良AI买家我们也见过大部分。

So, like, there's basically no good AI buyer we haven't seen before, and there are very few bad AI buyers we haven't seen before.

所以，这种组合——没错，网络规模很大，全球GDP的2%都流经Stripe。

And so, that combination, you know, it's it's, yes, the size of the network and there's 2% of global GDP, flowing through Stripe.

但说到AI，真正关键的是网络的密度。

But really when it comes to AI, it's the density of the network.

我们之前简单提过Link，举个例子，Lovable就挺有代表性的。

And we talked about Link briefly, but to give you a sense, like, Lovable is a good example.

作为一家AI公司，Lovable的业务量中有58%是通过Link完成的。

As an AI company, 58% of Lovable's volume flows through Link.

Link在AI领域简直是个高度密集的网络。

Like, Link is, like, you know, Link is an extremely dense network when it comes to AI.

所以，你可以据此推断出一些趋势。

And so, know, you you can sort of extrapolate with them.

如果我们已经清楚哪些是优质的买家，也见过那些不良买家在别处的表现，再结合Yes AI和这些实时API，就能构建出非常有效的反欺诈系统。

But if if we know who all the good buyers are, and we've seen the bad guys be bad somewhere else, then combine that with Yes AI and sort of these sort of real time these real time APIs, and you get, pretty good fraud defenses.

你之前提到过Tempo。

You mentioned Tempo at some point.

我们要不要聊聊Tempo？

Should we cover Tempo?

Tempo和代理式商业这个话题有多大关系？

Like, is is how relevant is Tempo to the agentic commerce conversation?

我觉得我们在Tempo上的工作有几个方面特别有意思。

So I think there's there's a couple components of our work with Tempo that I think are, really interesting.

说到商业领域的智能体电商，我们推出了机器支付协议，也就是MPP，和Tempo一起打造的，这是一个开放标准，运作方式非常简洁。

So, when you think about agentic commerce on the business side, we launched the machine payments protocol or MPP, and we built it with Tempo, And it's an open standard, and the way it works is, quite elegant.

对吧？

Right?

比如，一个智能体请求访问某项服务。

Like, an agent requests access to a service.

这个服务可以是任何API、MCP服务器，或者其他形式。

It can be whatever an API or an MCP server or whatever.

然后服务端会返回一个付款请求，智能体直接完成支付。

And then the service responds with a payment request, and then the agent pays.

整个过程不需要注册账户、不需要跳转结算页面，也不需要人工介入——完全不同于我们平时上网的交互方式。

And there's no kind of account creation or checkout UI or human in the loop or sort of the way you and I would traditionally, engage on the Internet.

它就是一种高度标准化、机器可读的机制，让智能体能直接向企业购买服务。

It's just this very kind of machine readable, standardized way for agents to to buy from businesses.

目前来看，MPP正是企业用来支持智能体作为买家的核心方式。

And that is really, MPP is really the primary, mechanism that we're seeing businesses use, for kind of agents agents as buyer.

另一个我和Tempo合作的领域，特别看好的，是跟欺诈防范有关。因为现在AI代理越来越多地成为AI产品的使用者，而这些代理消耗令牌的速度会非常快。

The other collaboration with, with Tempo that I am super bullish on is is more related to to fraud, because, you know, agents are increasingly becoming the users of AI products, and agents can burn through tokens very, very quickly.

这就带来了一个两难局面：企业要么限制自助服务，确保安全但增长缓慢；要么开放给代理使用，但这就可能面临严重的滥用和资金损失。

And so we were talking a little bit about this, like, dichotomy that a business faces where either you can, like, siphon off self serve and be, really safe but grow slowly or open it up, including to agents, but then, like, you know, be putting yourself at risk of for quite a bit of abuse and monetary losses.

这两种方式都不理想。

Neither of those is great.

真正该做的，尤其是当代理是购买方的时候，是实时追踪令牌的消耗情况。

What you actually wanna do, especially when the agents are the buyers, is you wanna, track the tokens as they're consumed.

你刚才提到，需要一套能大规模实时追踪的基础设施，这确实关键。

And you mentioned the infrastructure you wanna track them in real time at substantial scale.

但更重要的是，光追踪还不够。

And then as importantly, you don't just wanna track them as they're consumed.

你还要在令牌被使用的同时，同步完成收款。

You actually wanna, like, collect payments as they're consumed.

我们把这个叫做‘流式支付’，而Metronome和Tempo——这是Stripe参与共同打造的、专为支付优化的区块链——正在共同实现这一功能。

And so, we call this streaming payments, and it's what Metronome and Tempo, which is the blockchain optimized for payments that Stripe helped co build, are are making possible together.

所以，Metronome 的职责就是实时追踪使用情况，而 Tempo 的作用则是支持这种快速、低成本、高并发的微支付，而且能即时结算。

And so just Metronome's job is, like, track the usage in real time, and Tempo's job is just enable this, like, fast, low cost, high volume micro payments, that settle instantly.

当然，结算都是用稳定币进行的。

Obviously, they settle in in stables.

这样一来，AI 公司就能在用户使用令牌时实时收费，而不用在彻底关闭业务和被拖欠账款之间做两难选择。

And so put together, it's like AI companies, can charge generally agent buyers as tokens are consumed, instead of having to choose between kind of, closing off business, you know, or or getting stiffed on on the invoice.

所以我们确实觉得这个方向很有前景。

So, we we've, yeah.

我们对 Tempo 和稳定币在代理经济中的整体应用，尤其是在代理商业场景中的潜力，非常看好。

We're we're very bullish about Tempo and stables in general in in the agent economy at large and for the purposes of, of agentic commerce.

非常有意思。

Super interesting.

好的。

Alright.

那最后一个问题，你们手上应该有不少关于 AI 经济，特别是 AI 初创公司的有趣数据吧。

So, perhaps as the last topic, you guys have all sorts of interesting stats about the AI economy in general, but, like, in particular, AI startups.

我觉得上次我们已经聊过一些了。

I think we covered some of that, last time.

过去一年里，你在AI初创公司、趋势、数据和增长率方面观察到什么了？

What have you seen in the last year in terms of AI startups, trends, and facts, and growth rates?

你都看到了哪些变化？

What have you seen?

这挺有意思的。

So it's interesting.

去年我们聊的时候，谈过AI初创公司的发展，以及它们和传统初创公司的不同之处。

When we when we talked last year, we talked about, the growth of AI startups and, how they looked different than traditional startups.

我觉得现在最明显的是，AI初创公司和非AI初创公司之间确实存在差距。

And what I would say today is, like, know, there's definitely a delta between AI startups and non AI startups.

但更让我惊讶的是，AI正在整体改变创业生态。

But what's more striking to me is how AI is just changing the startup ecosystem generally.

就像现在流行的‘氛围编程’和‘氛围部署’这些趋势，新公司的注册数量其实比大家想象的要多得多，我觉得整体创业活跃度被严重低估了。

And so, you know, in the vein of vibe coding and vibe deployment and all of that, like, new business registrations are, well so I think, in general, the the business formation story is underappreciated.

全球范围内的新企业注册数量都在上升。

So, new business registrations are up basically around the world.

在发达经济体中，荷兰增长了约40%，芬兰高达70%，法国甚至达到了80%。

At least for advanced economies, they're up, like, 40% in The Netherlands and 70% in Finland and 80% in France.

这种活力的爆发，我们在美国确实能感受到，但其实它正在所有发达经济体中同步发生。

And so there's this sort of, like, surge in dynamism that, like, yes, we see and feel in here in The US, but it's happening, across advanced economies.

从Stripe的角度来看，新公司成立的速度相比去年已经翻了一倍。

When we look at this with the Stripe lens, the pace of new businesses launching has doubled since we talked last year.

并不是所有这些公司都属于AI领域，但其中很多能够成立，都得益于AI技术。

And not all those businesses are AI businesses, but, many, many of them were, made possible because of because of AI.

这些公司不只是刚起步。

And they're not just getting started.

它们还在快速扩张。

They're also scaling.

Atlas是我们为创业者提供的公司注册服务。

So, like, Atlas is our, product for founders to incorporate.

2026届的Atlas初创公司，现在才六月，还处于早期阶段。

And Atlas startups from the 2026 cohort you know, it's only June, so it's early in their life cycle.

但它们的收入增长速度，已经达到了去年同阶段的五倍。

But they're tracking to, like, five times the revenue of last year's class at at the same, number of months.

五倍？

Five times.

对，五倍。

Five times.

这其中一部分原因是公司数量变多了。

And some of that is there's more of them.

但更多是因为它们更快赚到了第一笔收入，之后扩张的速度也快多了。

But a bunch of it is they are getting to their first dollar faster, and then they are scaling more up more quickly.

能更快赚到第一笔收入，很大程度上确实要归功于AI。

And getting to their first dollar faster, a lot of that is for sure AI.

而扩张得更快，很大程度上是因为它们从一开始就瞄准了全球市场。

And then scaling up more quickly, you know, the a big part of this is actually how they're going global.

我不确定这其中有多大一部分是AI的功劳，但过去大家的普遍想法是，先做大，等规模上去了，才有资格走向全球。

And I don't know how much of that is AI or not, but, like, sort of the old model people had was you get big, and then once you're big, you deserve to go global.

但过去这一年，我们看到的情况越来越明显：你根本不用等，从第一天上线起，就实实在在地进入了几十个国家——当然不是说每个国家都去，但确实是全球布局。

And what we're seeing with, like, increasingly over the last year is you literally go global from it doesn't litter necessarily literally mean every country, but it's like you're in dozens of countries on day one, like your launch day.

这才是真正做大的方式。

And that is how you get big.

你要想做大，就得先全球化。

You get big by being global.

所以，我们其实聊的还是AI。

And so, I guess we're talking about AI.

我可以拿Emergent Labs做个例子。

So I could use, like, Emergent Labs as an example.

对吧？

Right?

这是一个AI平台，帮你构建和部署完整的全栈应用。

AI platform for, you know, you build and deploy these kind of full stack apps.

他们2024年在美国成立。

So they were founded in 2024 in The US.

他们70%的收入来自海外销售，在16个国家都有实质性的业务。

70% of their revenue, comes from international sales, and they do material business in 16 countries.

他们的收入有相当大一部分来自众多国家。

Like, a substantial share of their revenue comes from many, many countries.

所以，我觉得现在确实出现了很多AI公司，而且整个生态已经从当初只是底层技术提供商，演变成如今到处都是各种应用型公司，覆盖了每一个行业。

So, anyway, I I think there's this sort of, like, yes, there's AI companies and, you know, it's moved from, you know, just being sort of the underlying providers to, like, a lot of wrapper businesses, proliferation of wrapper businesses across every single vertical.

但从宏观角度来看，更有趣的是，过去你得是个开发者才能创业、才能做产品。

But I think what's kind of more interesting from the tops down macro perspective is just you used to have to be a developer to be a builder and therefore build a business.

现在你只需要有一个想法，再加上一点‘氛围编程’、一点‘氛围部署’，再加上基本的商业基础设施就够了。

Now you kind of have to have an idea plus vibe coding, plus vibe deployment, plus reasonable economic infrastructure.

于是我们看到大量新公司涌现，它们的开发进程根本停不下来。

And then, you know, we see this proliferation of, new businesses, and they, aren't, their their development is not easily arrested.

对吧？

Right?

哪怕只有一个员工，他们也能很早就开始盈利。

Like, even with one employee, they, you know, start monetizing early.

发展得特别快。

They, grow quickly.

迅速拓展到多个市场。

They, expand across a bunch of markets.

我觉得，这很大程度上得益于AI能处理大量运营事务，而这些其实并不属于Stripe的核心业务范围。

And, I think that's almost certainly helped by all of the operational work that can also be done with AI, is less than sort of the day to day core wheelhouse of Stripe.

我们在会计、收入确认这些方面会提供一些支持。

We help with some of that on the accounting or rev rec or whatever.

但客户支持和其他运营工作，显然主要是由其他公司来完成的。

But, you know, much of the customer support and other operations are are obviously done by other businesses.

我觉得现在是个挺有意思的时代。

But I think it's an interesting time.

很多人都在讨论，AI会不会让少数几家公司占据绝大部分市场份额，而竞争变得很少？

And there's a lot of discussion of like, is AI going to lead to a small number of firms, with heavy market share and not a lot of competition?

我觉得我看好AI经济的一个原因，至少到目前为止，确实有一些巨头的企业，它们的业务和AI高度互补。

And I think one of the reasons I'm bullish on the AI economy is at least so far, for sure, there are big guys who have things that are highly complementary to AI.

我们没必要一一列举这些公司。

We we don't need to go through that list.

大家都清楚，这些公司正在迅猛发展。

Everyone knows them, that are exploding.

但同时，也有大量小公司涌入市场，不只是被创立出来，而且真的能触达客户、快速成长。

But there's also an explosion of little guys coming in the market and, not just being created, but, like, reaching customers and growing quickly.

所以我认为，这对竞争和经济增长都是好兆头。

And so I think that bodes well for for competition and for and for economic growth.

这一切显然都令人无比兴奋。

All of this is obviously, extraordinarily exciting.

不过，既然我们聊到了token成本、使用量这些事，你有没有担心，买方和使用者这边有点过头了，没太意识到用AI花的钱到底有多少，会不会因此引发对这种超高速增长的反弹？

But since we talked about, token costs and, you know, usage base and all the things, do you worry that, like, a part of this is a little bit people, on the on the on the buying side and on the usage side got a little carried away, quite didn't quite realize the dollar amount that, using AI represented and that there might be some kind of backlash against that, hypergrowth.

我的意思是，我们都读过那些故事，有些公司因为没管好员工，结果无意中在token上花爆了钱。

So, I mean, we've all read the stories of companies who have accidentally gone bananas on token spend because they had, like, no control over what their employees were doing.

真正花得失控的公司，多半是资金雄厚的那一批，但这并不意味着如果这种支出持续下去，对经济就不会造成问题。

I think the companies who have truly gone bananas are by and large the companies with pretty deep pockets, which isn't to say that it's not, like, gonna be a problem in the economy if that spending continues.

不过，也有很多管理得当、运营聪明的公司。

But there are smart companies that are well run, well managed.

它们能扛得住一两个月的决策失误和成本失控。

They can handle a month or two of poor decisions and runaway costs.

说实在的，我不太清楚这些公司内部具体怎么算的，但听说他们花在令牌上的费用占了人力成本的2%到4%，其中三成到四成都是浪费的。

And, you know, by and large, I I don't exactly know what we're talking about at these companies, but we're talking oh, 2%, 3%, 4% of their headcount costs are going into tokens and, 30% of that or 40% of that is inefficient.

我觉得他们很快就能回到高效的状态，这根本算不上什么生存危机。

I think they can, like, pretty quickly get back to an an efficient frontier, and and it's not really anything existential.

至于那些小公司，我觉得情况没这么夸张。

For the little guys, actually, don't don't think that's happening.

很多小公司都是按固定费用购买，还在用基础套餐，没什么大动作。

You know, many of them are sort of making fixed fee fixed fee purchases, and they're still on small plans and and whatever.

所以，这类故事确实不少。

So, I mean, the stories are out there.

顺便说一下，为了明确一点，我觉得我们在AI使用的效率边界上还有很多东西要学习。

And by the way, just to be clear, like, I think we have a lot to learn about the efficient frontier of AI use.

这其中有一部分是关于模型路由，以及不同任务该用什么模型。

And there's a component there around, like, model routing and what models use like for the job.

还有一部分是关于可观测性，我觉得很多公司在这方面都落后了。

And there's a component around just, like, observability, which I think many companies have find found themselves to be behind on.

另外就是一些规范、控制和安全边界的问题。

And then there's just like like norms and controls and guardrails.

我们都希望大语言模型的使用能有高回报。

And, you know, I think we all want high ROI, usage of LLMs.

但同时，我们也希望员工能清楚自己用了多少成本，能主动告诉我们，这些投入是否值得。

But we also want employees to know when they're, you know, rack it up really substantive costs, and we want them to be able to tell us, whether they anticipate the ROI, of those costs is gonna be there.

所以，我觉得可能会有个小调整，但我觉得还没发生什么根本性的变化，长远来看的潜力依然在。

So, yeah, I think there'll be like a little bit of a recalibration, but I don't think anything existential has happened that's gonna annul the sort of long run upside here.

好的。

All right.

最后一个问题，一年后我们再聊的时候，你觉得在代理式商业方面，我们能走到L1到L5的哪个阶段？

So as a last question, when we talk again in a year from now, where do you think we are in terms of agentic commerce, maybe using L1 to L5?

我不会让你对这个预测负责，但从大方向看，你认为未来十二个月真正可能发生什么？

And I I won't hold you to the prediction, but, like, directionally, what do you think realistically is going to happen in the next twelve months?

我觉得未来十二个月最值得关注的是，我们可能会往前推进一大步，但具体是到四级还是五级，我还说不准。

I think the most interesting thing in the next twelve months is I mean, I think we'll I think we'll move up, and I don't know if it's gonna be to four or to five.

不过更有趣的是，我们之前谈代理作为经济主体时，主要集中在购物这一块。

But I I think the more interesting thing is actually that, like we talked about agents as economic actors mostly in the context of buying.

但我认为，我们会开始看到一些更复杂的经济型代理——我不是说会到处普及，但确实会开始出现。

But I think we'll start to see and I'm not saying this will be proliferated everywhere, but I think we'll start to see agents that are, like, multifaceted economic actors.

它们不仅买东西，还会卖东西，搭建基础设施，甚至运营整个业务。

They're buying and they're selling, and they're provisioning infrastructure, and they're running businesses.

它们能从头到尾独立完成整个流程。

And they're, like, doing the thing end to end.

当然，我也不知道这类代理最终会有多少、怎么运作，是彼此协作，还是局限在某些奇特的细分领域里。

And, again, I don't know how many of these there will be or how they'll operate or whether it'll be, like, with each other or, like, in these, like, weird niche silos.