读完一早上的访谈.
主题综述

Agent 代购时代：广告与商业的重写

主题综述

主题页（活文档）· 最近更新 2026-06-12 · 取材 9 篇访谈（含 2 组重复收录，实为 7 期节目，逐字稿全文）

更新日志

主流共识

一、起点没有争议：购物的入口在从搜索框挪向对话框，那笔"税"也会跟着挪

做空方把现状说得最直白。Alex Rampell（a16z，前 TrialPay 创始人、曾是全球最大联盟营销商之一）：

"Like Google, they kind of are a tax on GDP. Consumer spending is a huge part of GDP. They get a percentage of all that spend because they're charging per click. That tax might just shift elsewhere."
「像 Google 一样，它们有点像 GDP 的税收。消费者支出是 GDP 的重要组成部分。他们从所有这些支出中获得一定比例的收入，因为他们按点击收费。这种税收可能会转移到其他地方。」
Alex Rampell · The Death of Search

注意他说的是"转移"，不是"消失"。投资方 Michael Barton（Coatue）描绘的下一步与此完全同构：

"The next step is going to be, I go to Gemini or ChatGPT and say the same thing, but it knows a lot more about me and will suggest better products."
「下一步将是，我去 Gemini 或 ChatGPT 说同样的话，但它更了解我，并会推荐更好的产品。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

而站在品牌与用户之间的 Decagon 走得更远——入口不只替代搜索框，还要替代 App 和网站本身。Jesse Zhang：

"Overall, though, eventually you want this to be a unified, you know, concierge experience. ... And eventually, if it's good enough, most consumers will just interact with the agent instead of even logging into the mobile app or the website."
「但总的来说，最终你希望这是一个统一的礼宾体验。……最终，如果它足够好，大多数消费者只会与代理互动，甚至不会登录到移动应用程序或网站。」
Jesse Zhang (Decagon) · AI Agents Talking to AI Agents

二、基础设施层在抢跑：协议、令牌、目录、银行 API，四个位置都有人在铺路

这一层没有分歧，只有分工。Shopify 把自己定位成"让每家店都能被 AI 读懂"的目录方。Tobi Lütke：

"What we want to do is make sure everyone's plugged into the various chat systems and there's an MCP connected to every Shopify store and a global catalog that's really nicely presented. We're building software specifically for the open AIs and the Claudes and the complexities to make it really, really easy for these products to show up and reason about them."
「我们想要做的是确保每个人都接入各种聊天系统，并且每个 Shopify 商店都有一个 MCP 连接，以及一个非常棒的全球目录。我们正在专门为 open AIs 和 Claudes 以及复杂性构建软件，以便让这些产品能够非常、非常容易地展示出来并进行推理等等。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

Stripe 做支付原语和跨平台协议（与 OpenAI 共建 ACP），明确把"代理"当作一种新客户类型：

"We introduce shared payment tokens, a new payment primitive that lets agents initiate payments without exposing credentials. You don't be flinging your credit cards all around the web."
「我们引入了共享支付令牌，这是一种新的支付原语，允许代理在不暴露凭据的情况下发起支付。你不必在网络上到处乱扔你的信用卡。」
Stripe 2025 年度信（John Collison 朗读） · Stripe's 2025 annual letter
"Autonomous agents themselves are emerging as a new customer type for internet businesses to sell to."
「自主代理本身正在成为互联网企业销售的新客户类型。」
Stripe 2025 年度信 · Stripe's 2025 annual letter

再往下一层，银行牌照方也已经把"agentic commerce"写进产品说明。Jackie Reses（Lead Bank，客户包括 Stripe、Ramp、Affirm）：

"There are core components of our products that deploy AI in the products so that our clients can use them for agentic commerce. ... It's completely programmable such that you could deploy it in a real-time payment network or in a checkout, a checkout modal prompt."
「我们的产品中有核心组件部署了人工智能，以便我们的客户可以将其用于代理商业。……它是完全可编程的，可以在实时支付网络或结账时使用，作为结账模态提示。」
Jackie Reses (Lead Bank) · Inside the Alibaba Boardroom

三、购买谱系的两端免疫，中间地带先被吃掉

a16z 和 Shopify 在这一点上罕见地完全一致。Justine Moore（a16z）给出"杠铃"判断：

"So both ends of the spectrum, I think are harder for AI to disrupt."
「我认为，这两个极端，人工智能都很难颠覆。」
Justine Moore (a16z) · The Death of Search

Tobi 用超市收银台说同一件事：

"It's like the things at the next to the checkout counter in the supermarket. I don't think people are going to use ChatGPT to buy Mars bars or something, or candy or so."
「就像超市收银台旁边的那些东西。我不认为人们会用 ChatGPT 来买 Mars 巧克力棒或者糖果之类的。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

而中间那段"有 UPC、可比价、重复购买"的地带，Rampell 认为消费者本来就在手动跑这个算法，AI 只是补全闭环：

"So like the consumer is the agent and this is like a very very inefficient AI, right? And if you could actually complete the entire circle and say like no longer give information but allow for automatic action on information like people will do that because we have observed that behavior today."
「所以消费者就是代理人，这是一个非常非常低效的 AI，对吧？如果你真的可以完成整个循环，不再提供信息，而是允许基于信息进行自动操作，人们会这么做的，因为我们今天已经观察到这种行为了。」
Alex Rampell（谈 Camel Camel Camel） · The Death of Search

四、互联网的信息层已被联盟佣金腐蚀，干净的"人类判断"成了稀缺资产

这是做空叙事和做多叙事共用的前提。Rampell 把病灶归到联盟营销本身：

"But like most of the things on the internet are crap. And we know that they're crap, but they SEO optimize crap in order to earn affiliate commissions, and summarizing that crap is not helpful. So how do you decrapify that?"
「但互联网上的大多数东西都是垃圾。我们知道它们是垃圾，但它们通过 SEO 优化垃圾来赚取联盟佣金，而总结这些垃圾是没有帮助的。那么，你如何去除这些垃圾呢？」
Alex Rampell · The Death of Search

Coatue 把同一个判断翻译成了一笔可定价的资产——Reddit：

"If OpenAI wants to have a shopping assistant, right? All the reviews in the world are on Google. They're not on OpenAI on ChatGPT today. So where do they go? They have to go to Reddit. What are they willing to pay Reddit to be able to use that data to ultimately build the shopping assistant that's going to take over the market?"
「如果 OpenAI 想要拥有一个购物助手，对吧？全世界所有的评论都在 Google 上。它们今天不在 OpenAI 或 ChatGPT 上。那么他们该去哪里呢？他们必须去 Reddit。为了能够使用这些数据来最终构建将接管市场的购物助手，他们愿意付给 Reddit 多少钱？」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

分歧在哪

分歧一：广告这笔税，是被废除，还是换个名字接着收？

Barton 把购物代理推到逻辑终点时，说出了全场最激进的一句：

"In that world, there's probably no advertising."
「在那个世界里，可能没有广告。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

但他下一段话马上自己拆掉了"无广告"的字面意思——税不消失，只是改道：

"There might not be like an ad per se, but from the merchant's perspective, instead of spending, you know, 20% of my revenue on marketing in the form of ads, and 2% of that is going to Shopify, that split probably changes. So that 20% that I'm spending on marketing ... is now going to like that profit pool is going to be more going toward a Shopify or the actual agent players themselves, OpenAI or Gemini."
「可能不会有广告本身，但从商家的角度来看，与其将收入的 20% 用于广告形式的营销，其中 2% 支付给 Shopify，这种分配可能会发生变化。所以我用于营销的 20%……现在这些利润更多地流向 Shopify 或实际的代理平台，OpenAI 或 Gemini。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

而 Tobi——理论上最该欢呼"广告之死"的人（他的商家是广告费的支付方）——却为定向广告做了全场唯一的正面辩护：

"A personalized ad is a wonderful thing, right?"
「个性化广告是一件很棒的事情，对吧？」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce
"I think the Meta, Shopify, LAU has created more businesses than any government policy in history."
「我认为 Meta、Shopify（这个组合）创造的商业机会比历史上任何政府政策都多。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

更讽刺的是，Barton 自己的账本显示：到目前为止，AI 在商业上的第一笔真金白银恰恰是让广告引擎更强——

"The first use case of AI truly is driving these advertising businesses to grow faster than you would have thought. If you think about two years ago, what did I think Meta's revenues were going to be? Maybe I thought they would grow 15%. Well, they're growing 25%."
「AI 的第一个用例确实是推动这些广告业务比你想象的增长更快。如果你回想两年前，我当时认为 Meta 的收入会是多少？也许我认为它们会增长 15%。但实际上它们增长了 25%。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

Rampell 给 Google 的诊断同样是"半死不死"——丢的是免费信息查询，留下的恰恰是能收税的商业查询：

"So what is currently happening is they are starting to lose some of the free, but not any of the imium, right?"
「所以目前发生的情况是，他们开始失去一些免费用户，但没有失去任何付费用户，对吧？」
Alex Rampell · The Death of Search

但他对 Amazon 广告（纯利润的"最佳 SKU"）的判决毫不含糊：

"So like the best SKU that they sell is the advertising SKU and that's going to be imperiled if they no longer control the presentation layer because like AI intermediates it."
「因此，他们最好的销售产品是广告产品，如果他们不再控制展示层，那么广告业务将会受到威胁，因为人工智能会介入其中。」
Alex Rampell · The Death of Search

分歧二：AI 能不能"创造需求"？——广告位的命根子

这是材料里最干净的一组正面对撞。Rampell 认为需求的诞生地在人类社交场，AI 进不去：

"Also, I think it's going to be very hard for AI to, for lack of a better term, inculcate demand. How do I know that the On shoe is cool? It's like, well, I need to see that Bama."
「此外，我认为人工智能很难，这么说吧，培养需求。我怎么知道 On 昂跑的鞋子很酷？就像，我需要看到 Bama 穿。」
Alex Rampell · The Death of Search
"It's like, once I see, oh wow, I should have that too. I'm in a sorority. I want that shoe. It's very hard for AI to do that."
「就像，一旦我看到，哇，我也应该有那个。我在参加女生联谊会。我想要那双鞋。人工智能很难做到这一点。」
Alex Rampell · The Death of Search

Tobi 想要的恰恰是 AI 主动开口推销——他叫它"私人购物助手的主动推送"，但功能上就是 Rampell 说 AI 做不到的那件事：

"I want the LLM to be proactive and just tell me, hey, this outfit would look good on you and here's a preview. Here's the full cost and you want it. It's the personal shopper outreach kind of component, which I think is just really valuable."
「我希望 LLM 能够主动告诉我，嘿，这套衣服穿在你身上会很好看，这里有一个预览。这是全部费用，你想要吗。这是私人购物助手的主动推送组件，我认为这真的很有价值。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

Barton 明确站队 Tobi，并把"冲动购买"这个范畴整个解构掉：

"Toby at Shopify kind of made this comment. I actually agree with it, which is those were actually not impulse purchases. I actually secretly wanted those things, but no one had ever shown them to me. I would never have gone on Google search and just like, I looked up that interesting steak knife, but when it was showed to me, I bought it."
「Shopify 的 Toby 发表了这个评论。我实际上同意他的观点，那些实际上不是冲动消费。我实际上暗地里想要那些东西，但没有人向我展示过它们。我永远不会在谷歌搜索上搜索那个有趣的牛排刀，但当它展示给我看时，我就买了它。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy
"Now, imagine a world where you're in ChatGPT or you're in Gemini, and instead of asking it for something, it's just telling you, hey, you're going on this trip, I think you need a new ski coat."
「现在，想象一下你身处 ChatGPT 或 Gemini，你不是向它询问什么，而是它只是告诉你，嘿，你要去旅行了，我认为你需要一件新的滑雪外套。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

Tobi 还补了一刀方法论：平台日志里看到的"冲动"，很多其实是多年蓄谋——

"When you look at log files, you're going to get a very incorrect view of the world. It's very often the longest consideration phases. For years, I wanted a travel adapter that can deal with everything to everything. And the fact that it goes in a chat from you or in an Instagram ad doesn't matter. I'll Insta-purchase it at this point."
「当你查看日志文件时，你会对世界产生非常不正确的看法。这通常是考虑时间最长的阶段。多年来，我一直想要一个能处理所有类型插头的旅行适配器。它出现在你和你 Instagram 广告的聊天中并不重要。我现在就会在 Insta 上购买它。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

注意：Rampell 和 Tobi/Barton 说的"需求创造"其实不完全是一个东西——前者指社交模仿驱动的潮流（sorority 效应），后者指对既有潜在欲望的唤起。两边都没把这个区分挑明。

分歧三：新入口之下，赢家是聚合器、长尾，还是"merit"？

同一家 a16z 内部就不齐。Justine Moore 押聚合器：

"AI agents can direct people to things if people start their purchase activities there, which could be an opportunity or a challenge for like the single SKU retailers. My guess is it'll still end up being more of a positive for like the aggregators."
「如果人们从 AI 开始购买，AI 代理可以将人们直接引导到商品，这对单一 SKU 零售商来说可能既是机遇也是挑战。我猜最终这对聚合器来说还是更有利的。」
Justine Moore (a16z) · The Death of Search

John Collison（Stripe）押长尾，类比 YouTube 之于电视：

"You have a much more powerful position as a tail merchant, because, you know, Momax is a tiny brand, but they can just get recommended as the best product."
「你作为一个尾部商家就拥有了更强大的地位，因为，Momax 是一个很小的品牌，但他们可以被推荐为最好的产品。」
John Collison (Stripe) · Tobi Lütke is still captivated by internet commerce

Tobi 当场修正了 Collison 的框架——不是头尾之争，是按质量重新洗牌：

"I think of a slight modification I would just make to this. I don't think the tail part matters. I think merit matters."
「我想稍微修改一下。我认为尾部并不重要。我认为重要的是价值（merit）。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

但对"接入新聚合层是福是祸"，Barton 在 OpenAI dev day 现场看到的是市场的精神分裂：

"You could argue a lot of these companies that are kind of named in the presentation to go and be part of this like agent layer ... it might not actually be a good thing. Like if everyone's using ChatGPT, and now you just got ... an additional layer of like, maybe disintermediate ... But if your name got mentioned, bang, you're up five."
「你可以说很多在演示中被点名的公司会参与到这个代理层中来，这实际上可能不是一件好事。比如，如果每个人都在用 ChatGPT，而你现在只是增加了一个额外的层次，也许（会被）去中间化……但是如果你的名字被提及，砰，你就涨了五个点。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

Stripe 则把赌注押在"不要出现新垄断聚合点"这个前提本身上：

"If all goes well, the little critters are not going to be cooped up in walled gardens, but they'll be zooming down the wide-open protocol highways."
「如果一切顺利，这些小家伙们将不会被困在围墙花园里，而是会在广阔的协议高速公路上飞驰。」
Stripe 2025 年度信 · Stripe's 2025 annual letter

分歧四：什么才是"防 AI"资产——信任策展、产品品质，还是人类数据？

三个阵营各自指认了一种免疫物。Rampell 的答案是 Costco 式信任策展，且他认为这是唯一：

"I think Costco is the greatest company in the world because Costco refuses to sell bad things. They refuse to take a high gross margin."
「我认为 Costco 是世界上最伟大的公司，因为 Costco 拒绝销售劣质商品。他们拒绝接受高额毛利率。」
Alex Rampell · The Death of Search
"Costco is immune to all of this because it's like, they're like the Consumer Reports. ... But it is a very, very unique business model that ... will stand the test of time. And I think it's AI proof."
「Costco 对这一切免疫，因为它就像消费者报告。……但这是一种非常非常独特的商业模式，……它将经受住时间的考验。我认为它是 AI 免疫的。」
Alex Rampell · The Death of Search

Tobi 的答案是产品本身的品质——他甚至把它升格为对消费主义的回答：

"The thing that solves consumerism is quality products that you want to keep using."
「解决消费主义的方法是生产高质量的产品，让你想一直使用下去。」
Tobi Lütke (Shopify) · Tobi Lütke is still captivated by internet commerce

Barton 的答案是稀缺的人类评论数据，而且给出了定价方向：

"Google or OpenAI is willing to pay Reddit for the key piece of data that may determine the success of the entire shopping agentic TAM. My guess is higher."
「谷歌或 OpenAI 愿意为 Reddit 的关键数据付费，这些数据可能决定整个购物代理 TAM 的成功。我猜（这笔费用）会更高。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

这三种答案彼此并不兼容：Costco 模式靠"替你不选"，merit 模式靠"让最好的被选中"，Reddit 模式靠"卖出人类怎么选的记录"。如果代理真的接管选择权，三者只会有一个成为主要瓶颈。

分歧五：进度表——"那个世界基本已经到来" vs "还在 Level 1 和 2 的边缘"

身处 agent-对-agent 一线的 Jesse Zhang 最激进：

"Oh, I think that world is basically here. I mean, you have all these consumer agents that are going out there and they can order DoorDash for you and so on. And at some point, maybe they'll call into an airline to reschedule your flight or something. And then maybe they'll talk to our agent. And then you'll have agents talking to each other."
「哦，我认为那个世界基本上已经到来了。我的意思是，你拥有所有这些消费者代理，他们会出去为你订购 DoorDash 等等。有时，他们可能会致电航空公司来重新安排你的航班或什么的。然后他们可能会和我们的客服沟通。之后你就会看到客服之间相互沟通。」
Jesse Zhang (Decagon) · AI Agents Talking to AI Agents
"We'll just have two agents interacting and they're just like spitting tokens at each other and you can just get something done."
「我们会让两个客服（agent）互动，他们只是互相发送 token，然后你就能完成一些事情。」
Jesse Zhang (Decagon) · AI Agents Talking to AI Agents

最深度押注这个未来的 Stripe，反而给出了最冷的进度读数（他们把 agentic commerce 分为五级：代填表单 → 描述式搜索 → 记忆持续 → 授权代购 → 预判购买）：

"Now, like most in AI, agentic commerce suffers from having been a bit overhyped too early. People paint this utopian picture of autonomous agents running around, planning your whole life, executing commerce without, you know, even talking to you."
「现在，像大多数人工智能一样，代理商务也过早地被过度炒作了。人们描绘出一幅乌托邦式的景象，即自主代理到处运行，规划你的整个生活，在甚至不与你交谈的情况下执行商务活动。」
Stripe 2025 年度信 · Stripe's 2025 annual letter
"But today, the industry is hovering right on the edge of levels maybe 1 and 2. ... At the time, no one knew exactly which protocols or players would win out. There was an AltaVista for every Google."
「但今天，该行业正徘徊在 1 级和 2 级的边缘。……当时（90 年代中期），没有人确切知道哪些协议或参与者会胜出。每个 Google 都有一个 AltaVista。」
Stripe 2025 年度信 · Stripe's 2025 annual letter

中间派的证词支持 Stripe 的读数。Justine Moore 记录了第一波用户的退潮：

"They would find that a lot of the products that recommended did not exist or previously existed, but did not exist in a current form or the amount that they were charging was way different than it said, which I think drove a lot of people who experimented with it sort of back to, you know, I'm going to return my searches to Google or Amazon and wait until ChatGPT figures out this commerce thing."
「他们会发现，很多推荐的产品要么不存在，要么以前存在但现在已经没有了，要么价格与描述的相差甚远，我认为这让很多尝试过的人又回到了，你知道的，我还是回到谷歌或亚马逊搜索吧，并且等待 ChatGPT 搞清楚这个商业模式。」
Justine Moore (a16z) · The Death of Search

Barton 对 OpenAI 首批商务集成的评价同样克制：

"We had OpenAI announce the Shopify and Etsy integration a couple of weeks ago. The product today is not at a place to really be that useful, but you can kind of see where it's going."
「几周前，OpenAI 宣布了与 Shopify 和 Etsy 的集成。今天的产品还没有达到真正有用的程度，但你可以看到它的发展方向。」
Michael Barton (Coatue) · Inside Coatue: $70B Hedge Fund's AI & Retail Strategy

而 No Priors 的 2026 预测则提示：连"谁来做这个入口"都还没定——

"There's so much talk of agents right now, and there has been for a while, but no one has truly created a mass-scale consumer agentic AI. ... And I think this area is not nearly as seeded to the labs as people assume. It really is anyone's ballgame."
「现在有很多关于代理的讨论，而且已经有一段时间了，但还没有人真正创造出一个大规模的消费者代理 AI。……而且我认为这个领域远没有人们认为的那样被实验室垄断。实际上，每个人都有机会。」
嘉宾预测（No Priors 2026 特辑） · The 2026 AI Forecast

都没说透的

1. 归因问题被诊断了，没被治。 Rampell 把末次点击归因骂作互联网上"最具普遍腐蚀性的商业模式"，把 Honey/RetailMeNot 直接定性为"盗窃"（"It went public, big valuation, and it was just theft. / 它上市了，估值很高，但那只是盗窃"），又承认 AI 世界里归因只会更难（"figuring out and disentangling attribution is very, very hard / 找出并理清归因非常非常困难"）。可他自己看好的创业方向恰恰是"我们将成为 AI 时代后的 21 世纪的最后一次点击"（"We are going to be the last click, like the last click of the 21st century post AI"）。他批判的模式和他兜售的机会在结构上是同一个东西，整场对话没有人点破。

2. 代理到底为谁打工，没人正面回答。 Rampell 设想的代理替用户榨取每一分返现与优惠券（"not all of this will be good for merchants / 并非所有这些都对商家有利"）；Barton 描绘的利润池却流向代理平台自身（OpenAI/Gemini）。同一个代理不可能同时最大化用户省的钱和平台抽的成。主持人 Molly O'Shea 借 Reid Hoffman 之口把问题递到桌上——"Advertising was majority of the last one. We don't really know what the AI business model is yet.（广告是上一个时代的主要模式。我们还不太清楚人工智能的商业模式是什么。）"——Barton 的回答转向了"变化太快要多收集数据点"，等于没接。

3. 需求创造如果留在人类社交场，新品牌在代理时代到哪里出生？ Rampell 论证 AI 难以"培养需求"，Tobi 承认 Burton 的品牌光环让直接卖滑雪板的广告打不动、Meta/Shopify 才让利基产品找到"1,000 true fans"。但如果发现环节（Instagram/TikTok/红人）保持人类属性、履约环节（搜索→下单）被代理接管，两段之间的归因和定价怎么接，没人给出图景。Justine 提到的"非赞助 YouTube 测评 → 转成 LLM 可读文本"是材料里唯一接近答案的线头，也只停在一句观察。

4. 开放协议与数据高墙在对撞，双方都装作没看见。 Stripe 的愿景要求"宽阔开放的协议高速公路"；Barton 的 Reddit 论点却意味着最关键的人类数据会被越垒越高的授权费圈起来（Rampell 也指出搜索早已碎成 walled gardens）。代理商务越值钱，墙就越高；墙越高，开放协议愿景越难成立。这组矛盾横跨两篇访谈，材料里无人同时面对。

我的看法

以下是我的判断，不是材料的归纳。其一（把握较高）：这笔"GDP 税"不会消失，只会换收税人和换名目——从按点击付费变成代理层抽成、协议费和数据授权费；三方证词（Barton 的利润池改道、Stripe 的支付令牌与 ACP、Rampell 的"税会转移"）在不同位置指向同一结构。其二（把握中等）：分歧二其实比看起来小——Tobi/Barton 描绘的"主动推送的私人购物助手"一旦规模化，本质上就是重新发明的广告位，Rampell 守住的只是"潮流类需求的社交起源"这一小块；真正的悬念是代理推荐位按什么机制定价、向谁竞价。其三（把握中等偏低，时间表问题）：进度上我更信 Stripe 的"Level 1–2 边缘"而非 Jesse 的"基本已经到来"——基础设施方的保守估计有真金白银背书，而最乐观的证词恰好来自最需要这个叙事成立的卖方。

还想知道什么

取材

继续读
线索
阅读线索：agent 不到 9 nines 可靠，那应该卖给企业还是消费者
线索
阅读线索：如果 coding 是 commodity，AI startup 的护城河长在哪儿
线索
阅读线索：推理成本一年跌 10×，AI app 该按什么定价
线索
摘报
综述
AI Agent 安全与治理
综述
AI 资本周期：泡沫还是 S 曲线